UNN-184 j The eL t4 Cu, C .asqua1eL Scamcd iz Pt A World Bank Research Publication - 3 rt >r ~~~FILE LA)UtY 30hns flopkrnS The Agricultural Economy of Northeast Brazil A WORLD BANK RESEARCH PUBLICATION The Agricultural Economy of Northeast Brazil Gary P. Kutcher and Pasquale L. Scandizzo PUBLISHED FOR THIE WORLD BANK The Johns Hopkins University Press BALTIMORE AND LONDON Copyright (C 1981 by The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved .Manufactured in the United States of America The Johns Hopkins University Press Baltimore, Maryland 21218, U.S.A. The views and interpretations in this book are the authors' and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. The map on p. 20 has been prepared for the convenience of readers of this book; the denominations used and the boundaries shown do not imply, on the part of the World Bank and its affiliates, any judgment on the legal status of any territory or any endorsement or acceptance of such boundaries. Library of Congress Cataloging in Publication Data Kutcher, Gary P., 1944- The agricultural economy of northeast Brazil. (A World bank research publication) Bibliography: p. 259 Includes index. 1. Agricultural-Economic aspects-Brazil, Northeast. I. Scandizzo, Pasquale L. II. Title. III. Series. HD1872.K87 338.1'0981'3 81-47615 ISBN 0-8018-2581-4 AACR2 Contents Abbreviations xi Preface xiii 1. Introduction and Summary 3 Development, Decline, and Government Intervention 5 The Approach Followed in This Study 19 Summary and Conclusions 21 Appendix: Zonal Delineation of the Northeast 30 2. Labor, Land, and Capital Resources 36 The Agricultural Labor Force 36 Land: Its Distribution and Quality 41 Agricultural Capital 53 3. Outputs, Inputs, and Incomes 58 Agricultural Products 58 Agricultural Output 60 Inputs and Costs 63 Agricultural Incomes 67 Appendix: The National Study of Family Expenditures 73 4. Farm Sizes, Types, and Performance 82 Land and Labor Use 82 Market Dependency: Family and Nonfamily Farms 89 Farm Sizes and Types: A Multiple Classifiication 94 Some Conclusions 97 5. The Structure of the Programming Model 102 An Overview of the Model 104 The Core of the Farm Models 105 Characteristics of Individual Farm Models 109 Linkages among the Farm Models 116 The Treatment of Demand 118 Appendix A: Aggregation of Farm Types 120 Appendix B: Algebraic Statement of the 1Programming Model 137 Appendix C: Estimation and Testing of Parameters 145 6. Results from the Base Solution 150 Validation Considerations 150 Production and Prices 152 v Vi THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Employment and Labor Utilization 156 Family Reservation Wage 159 Price Elasticities of Demand 160 Optimality Assumptions 162 The Pattern of Constraints 164 7. Policy and Project Options for Northeastern Agriculture 169 The Agricultural Economy Projected to 1978 170 The 1978 Base Case Solution 173 Technical Progress: The Effects of Improvements in Crop Yields 175 The Influence of Risk 177 Employment Policies 179 The Promotion of Demand 183 Development Strategies in the Absence of Agrarian Reform 191 8. The Land Reform Alternative 195 Legislative History 196 The Land Reform Module 197 Potential for Redistribution of the Estate Subsector 198 Performance of the Module Farms 199 Potential Effect of Land Reform 202 Potential Beneficiaries 205 Estimates of Cost 206 Alternative Development Strategies in Conjunction with a Land Reform 208 Conclusions 210 9. The Northeast Problem and the Future of Brazilian Development 212 Miracle Growth and the Rural Northeast 212 Alternative Solutions 213 Potential Role of the Rural Northeast 216 Introduction to the Appendixes 219 Appendix A. Questionnaire Design 220 Design Criteria 220 Field Tests 223 Questionnaire Content 224 Appendix B. Sampling Procedures 228 Selection of Municipios 230 Selection of Farms 233 Supplementary Samples 239 Notes on Sample Expansion 241 CONTENTS vii Appendix C. The Fieldwork 249 Selection and Training of Enumerators 249 Organization and Evaluation of the Fieldwork 251 Appendix D. Data Validation Procedures 253 Bibliography 259 Index 265 Tables 1-1. Principal Characteristics of the Physiographic Zones 32 2-1. Total and Rural Population and Number Economically Active in Agriculture, by State, 1970 37 2-2. Agricultural Labor Force, by State, 1970 37 2-3. Agricultural Labor Force, by Zone, 1973 40 2-4. Intensity of Agricultural Employment, by Zone 42 2-5. Estimated Size Distribution of Farms, by Zone 43 2-6. Average Size of Sampled Farms, by Farm Size and Zone 44 2-7. Total Area in Farms, by Farm Size and Zone 45 2-8. Distribution of Farms and Farm Area, by Size and Zone 46 2-9. Average Land Values, by Farm Size and Zone 48 2-10. Average Distance from Sampled Farms to Municipio Seat 51 2-11. Tests for Independence between Land Quality and Farm Size, by Zone 52 2-12. Agricultural Capital, by Zone 53 2-13. Average Agricultural Capital per Farm, by Zone 54 2-14. Distribution of Equipment, by Farm Size and Zone 56 2-15. Regression Results: Incremental Capital/Output Ratios (ICOR) 57 2-16. Capital/Output Ratios: Value of Housing, Other Structures, and Equipment 57 3-1. Average Yields of Principal Crops, 1973: Alternative Estimates 62 3-2. Percentage Composition of Agricultural Receipts and Costs, by Zone 64 3-3. Average Expenditure per Farm on Inputs for Livestock, by Farm Size and Zone 66 3-4. Average Expenditure per Farm and per Hectare on Other Purchased Inputs, by Farm Size and Zone 66 3-5. Agricultural Incomes, by Type of Agricultural Agent 68 3-6. Average Permanent Workers' Wages, by Zone 69 3-7. A Low-cost Northeastern Diet 70 Viii THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL 3-8. Average Net Income per Farm, by Farm Size and Zone 72 3-9. Average Net Income per Hectare, by Farm Size and Zone 73 3-10. Mapping of Agricultural Households between the sUDENE/World Bank and ENDEF Surveys 75 3-11. Selected ENDEF Expenditure Data: Current Monetary Expenditure 76 3-12. Selected ENDEF Expenditure Data: Global Expenditure 78 4-1. Gross Output per Hectare, by Farm Size and Zone 83 4-2. Relations between Gross Output per Hectare, Farm Size, Percentage of Farmland under Crops, and Land Quality 84 4-3. Man-years of Labor and Capital per Hectare, by Farm Size and Zone 85 4-4. Trends in Capital and Labor Use as Farm Size Increases, by Zone 86 4-5. Cobb-Douglas Production Functions, by Type of Farm and Zone 94 4-6. Comparison of Marginal Products of Labor with Wage Rates, by Type of Farm and Zone 96 4-7. Comparison of Marginal Products of Labor on Family Farms with the Probability of Finding Alternative Employment 97 4-8. Scalar and Dummy Variable Analysis of Effects on Gross Output of Farm Size, Sharecropping, and Three Types of Equipment 98 5-1. Core Matrix of Farm Models 106 5-2. Family Consumption Tableau 110 5-3. Risk Tableau III 5-4. Livestock ''ableau 113 5-5. Sharecropping Tableau 114 5-6. Man-years of Labor per Hectare, by Farm Size and Zone 124 5-7. Northeast Farm Characteristics, by Farm Type 126 5-8. West: Farm Characteristics, by Farm Type 128 5-9. Middle-North: Farm Characteristics, by Farm Type 129 5-10. Sertao: Farm Characteristics, by Farm Type 130o 5-11. Semihumid Southeast: Farm Characteristics, by Farm Type 131 5-12. Humid East: Farm Characteristics, by Farm Type 132 5-13. Humid Southeast: Farm Characteristics, by Farm Type 133 5-14. Agreste: Farm Characteristics, by Farm l'ype 134 5-15. Northeast: Estimate of Number and Size of Farms, by Type 136 6-1. Model Performance: Crop Production and Prices 152 6-2. Regression Results: Goodness-of-Fit Tests for Production Volumes of Price-Endogenous Crops Simulated by the Model 154 6-3. Persons Occupied in Agriculture: Model Results 156 6-4. Labor Requirements: Model Results 157 6-5. Employment Rates, by Month and Zone: Model Results 158 6-6. Effects on Model Results of Variations in the Family Labor Reservation Wage 160 6-7. Output Generation and Labor Use, by Farm Tvpe: Survev Data and Mlodel Results 163 CONTENTS ix 6-8. Shadow Prices of Land, by Farm Type and Zone: Model Results 165 6-9. Shadow Prices of Tree-crop Stocks, by Zone: Model Results 166 6-10. Domestic Resource Cost Coefficients of Food Crops, 1973 168 7-1. Time Trends in Yields and in Area Harvested, Principal Crops, 1961-63 to 1971-73 172 7-2. Percentage Shifts in Demand for Price-Endogenous Crops, 1973-78 173 7-3. Output, Income, and Employment: 1973 and 1978 Base Cases 174 7-4. Technical Progress Experiment 176 7-5. Alternative Emplovment Policies 180 7-6. Results of Promoting Demand 190 7-7. Alternative Development Strategies in the Absence of Agrarian Reform 192 7-8. Responses of the Estate Subsector to Policy Interventions 194 8-1. Potential Number of Module Farms in the Estate Subsector 199 8-2. Characteristics of Land Reform Modules 200 8-3. Results of a Land Reform in the Estate Subsector 201 8-4. Performance of the Estate Subsector before and after Redistribution 204 8-5. Estimated Partial Costs of Establishing Land Reform Modules 207 8-6. Alternative Development Strategies in Conjunction with a Land Reform 209 B-i. Production Patterns 231 B-2. Selected Municipios, by State 234 B-3. Change in the Optimal Size of Two Samples, as a Function of the Ratio between the Degrees of Variance They Are to Measure 237 B-4. SUDENE/World Bank Survey: Farm Sample Size 238 B-5. Municipio Weights for Sample Expansion 244 Figures 4-1. A Farm Classification Scheme Based on Market Dependency 90 4-2. Preliminary Classification of Sampled Northeastern Farms 91 5-1. Demand, Output, and Price Relations: Model Structure 119 6-1. Effects of More Elastic Demand Schedules: Model Structure 162 7-1. Supply Schedules for Price-Endogenous Crops of Different Classes of Farms: Model Structure 184 7-2. Alternative Equilibrium Points for Price-Endogenous Crops: Model Structure 185 7-3. Effects of Augmenting Supply under Alternative Elasticities of Demand: Model Structure 187 7-4. Effects of Augmenting Demand: Model Structure 189 A- 1. Representative Table from Questionnaire I 222 X THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Map Physiographic Zones of Northeast Brazil 20 Abbreviations ANCAR Associacao Nordestina de Credito e Assistencia Rural (National Agricultural Extension and Credit Agency) BB Banco do Brasil (Bank of-Brazil) BNB Banco do Nordeste do Brasil (Bank of Northeast Brazil) CEPA Comissao Estadual de Planejamento Agricola (State Agricultural PlannLing Commissions) CHESF Sao Francisco Hydroelectric Company CVSF Sao Francisco Valley Commission DNOCS Departamento Nacional de Obras Contra as Secas (National Department of Works against the Droughts) EMBRAPA Empresa Brasileira de Pesquisa Agropecuaria (Brazilian Agricultural Research Company) ENDEF Estudo Nacional da Despesa Familiar (National Study of Fami,ly Expenditures) FUNRURAL Fundo de Assistencia ao Trabalhador Rural (Assistance Fund for Rural Workers) GERAN Grupo de Estudo para a Reorganizacao da Industria Acucareira (Study Group for the Reorganization of the Sugar Industry) GTDN Grupo de Trabalho para o Desenvolvimento do Nordeste (Working Group for the Development of the Northeast) IBGE Instituto Brasileiro de Geografia e Estatistica (Brazilian Institute of Geography and Statistics) IBRA Instituto Brasileiro de Reforma Agraria (Brazilian Institute for Agrarian Reform) IPEA Instituto de Planejamento Economico e Social (Institute of Economic and Social Planning) INCRA Instituto Nacional de Colonizacao e Reforma Agraria (National Institute of Colonization and Agrarian Reform) xi Xii THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL INDA Instituto Nacional de Desenvolvimento Agricola (National Institute for Agricultural Development) POLONORDESTE Programa de Desenvolvimento de Areas Integradas do Nordeste (Development Program for Integrated Areas in the Northeast) PROTERRA Programa de Redistribuicao de Terras e de Estimulo a Agro-Industria do Norte e do Nordeste (Program for Land Redistribution and Stimulation of Agro-Industry in the Northeast) SUDAM Superintendencia para o Desenvolvimento da Amazonia (Superintendency for the Development of the Amazon Region) SUDENE Superintendencia do Desenvolvimento do Nordeste (Northeast Development Superintendency) SUPLAN Superintendencia do Desenvolvimento do Plano de Obras do Governo (State Superintendency of Works) 34/18 Article 34/18, which instituted a tax credit scheme to induce industrial investment in the Northeast Preface The research reported in this study began in 1973. In that year, Robert McNamara's Nairobi speech signaled a broadening of emphasis in World Bank policies and operations toward a direct concern with the problems of poverty and the distribution of the benefits of growth. This change of emphasis had particular ramifications for Bank activities in Brazil, one of its principal borrowers. The Brazilian "economic miracle" of the 1950s and 1960s, in which World Bank financing had played a not insignificant role, had left conditions in the Northeast virtually unchanged. Despite several massive developmental efforts, this tradition-bound, predominantly rural region of 30 million people remained the largest concentration of poverty in the Western hemisphere. A team from the World Bank's Development Research Center visited Brazil several times in late 1972 and 1973 to identify researchable questions about the Northeast. The team assumed thiat the Northeast "problem" was not confined to any single sector, and accordingly a study of the agricultural economy was only one component of the multisectoral Brazil Regional Studies Project that was originally formulated. Once the agricultural com- ponent had been embarked upon, however, the issues to be confronted proved so complex that the research on other sectors was not pursued. At the outset information about Northeastern agriculture was so scarce that even the most obvious development-al questions could not be ade- quately addressed. Earlier students of Northeastern problems, most no- tably the Working Group for the Developrnent of the Northeast (GTDN), led by Celso Furtado, had not been deterred by this situation. The Northeast Development Superintendency (SUDENE), the regional authority first directed by Furtado, had long had an agricultural survey planned and budgeted, but it was never carried out, reportedly for lack of technical assistance. ' The result of this mutual interest in data collection was a survey of 8,000 farms conducted jointly by SUD]ENE and the World Bank in late 1973 and early 1974. The present study is only one attempt to analyze the data from that survey and is by no means comprehensive. No panacea for the rural 1. Entitled "The Typical Size of the Unit of Production in the Northeast" ("tamanho tipico"), this survey was authorized during Furtado's term as superintendent of SUDENE. xiii XiV THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Northeast emerges; no single policy prescription; no lending program. It is hoped, however, that the presentation of detailed data on the characteristics of the agricultural economy and the application of new analvtical tools to identify some of the causal influences at work will help shed light on the sorts of measures required to stimulate agriculture and reduce the extent of povertv in the Northeast. The views and conclusions presented in this study are our sole responsi- bility and are not to be attributed to SUDENE, the World Bank, or anv organization or individual. We acknowledge our gratitude for the guidance, assistance, and support of the following: Peter B. Clark, who conceived and directed the project in its early stages and without whose relentless efforts the study would not have commenced; Arlindo da Costa Lima, our research counterpart in SUDENE, without whose expertise on Northeastern agricul- ture and immense energy the study would have also come to naught, and his colleagues, Alfredo Arruda Branco and Joaquim Veloso Maranhao; Yony Sampaio, Quirino Paris, Lee Bettis, and Daniel Loucks, who assisted in the early analytical stages; and Maria Helena de Castro Silva, Vinh Le-Si, and Scott Sirles, who provided expert research assistance, toiling long hours on computers in both Washington and Recife. Various stages of the manuscript were typed by Miriam Bailey, Maria McReynolds, Thelma Rapatan, Wendy Shinn, Leela Thampy, and Elizabeth Trask. Various drafts benefited from the comments of Jock Anderson, Bela Balassa, Wil- fred Candler, Wellington Dantas, John Duloy, Antonio Giles, Vander Gontijo, Peter Hazell, Constantino Lluch, Gerald O'Mara, Roger Norton, Guy Pfeffermann, T. N. Srinivasan, and two anonvmous reviewers. Rachel Weaving and Vivianne Lake edited the manuscript, and Barbara Palmer prepared it for publication. Chris Jerome corrected proof of the book, Pensri Kimpitak prepared the figures, and Ralph Ward and James Silvan indexed the text. The map was compiled byJulio Ruiz and drawn by Larry A. Bowring under the supervision of the World Bank's Cartography Division. The authors, of course, remain solely responsible for anv errors contained herein. More important than all of us are the nearly 8,000 farmers, administra- tors, and sharecroppers who freely gave their time and answers to the survey, and we wish to salute them; their only remuneration was the promise that the survey results would be used to study means of improving the lot of the Nordestino. If we have failed, our apologies are to them, their workers, and their would-be workers. The Agricultural Economy of Northeast Brazil 1 Introduction and Summary The Northeast "problem" is one of massive economic and social disparity, compounded by an apparent intractability. For over a century, the plight of the Northeast has been a source of embarrassment to Brazilians, of unre- lenting frustration to politicians, and an enigma to economic planners. More recently, the problem has become a major concern of other Western Hemisphere governments and international organizations, because it exem- plifies the extreme case of a large and persistent pocket of poverty in an otherwise dynamic, rapidly developing country. Since World War II, Brazil's growth and development have often been termed a "miracle." An agrarian, coastal economy has been transformed into a thriving, industrialized, geographically dispersed economic power. Per capita incomes have recently surpassed US$1,780 (1979) because growth rates in the industrial sectors often exceeded 10 percent. Population has more than doubled since 1945, and its concentration in urban areas has increased from 35 percent to over 60 percent.' Industry, population, and wealth, however, have remained concentrated in the Center-South, and the available evidence suggests that this concen- tration is increasing. By 1970, the Center-South contained 61 percent of Brazil's population and received 76 percent of national income; the North- east, with 30 percent of the country's population, received 11 percent of national income. For the 30 million people in the Northeast, the per capita income of about US$200 in 1970 was less than 38 percent of Brazil's average and less than one-fourth of that in the C'enter-South. Life expectancy at birth in the Northeast in 1970 was forty-eight years, compared with sixty-six in the South and fifty-six in all of Brazil. Within the Northeast, Langoni's landmnark study found that in 1970 the poorest 40 percent of the population received only 10.7 percent of regional 1. World Bank, World Development Report, 1979 (New York: Oxford University Press, 1979), pp. 127, 129, 165. 3 4 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL income, which implies that more than 11 million Northeasterners were subsisting on about US$50 each-an annual income too low to ensure adequate nutrition, shelter, health, or education.2 Indeed, census figures for 1970 showed that fullv 75 percent of rural Northeasterners over the age of five could neither read nor write. Illiteracy rates in the Northeast range from 72 percent in Bahia to 78 percent in Piaui, the poorest state in Brazil. Official concern has not been lacking. Several massive interventions have been attempted throughout this century. Innovative development strategies have been devised, and vast resources-both Brazilian and inter- national-have been spent. Yet there has been little growth or development in the Northeast as a whole. The rural Northeast, in fact, may' not have experienced any per capita growth: It continues to hold the largest concen- tration of povertv in the hemisphere. The Northeast problem remains an embarrassment to Brazilians and an immense challenge to economic and social planners. There are, of course, many dimensions and complexities to the North- east situation. It is not merely an agricultural problem, nor is it solely a rural problem. It is a problem of defining a role for a region that, having become disparate, can seemingly find no way out of its disparity. Many diagnoses have been made, many solutions attempted, and many other solutions may yet be possible. This study focuses on the agricultural sector, not because it represents the Northeast problem in its entirety but because the origins of this regional problem are rooted in the agricultural system; because the most dramatic effects of the problem are reflected in the lives of the 10 million or more people who continue to scratch out a meager existence in agriculture; and because the dismal conditions within agriculture have had spillover effects on other sectors and other regions, thereby contributing to urban slums, social unrest, an inadequately educated or skilled labor force, and nutritional deficiencies traceable to forgone consumption of basic food- stuffs. In taking this focus, we do not wish to belittle other approaches nor to cliscourage other inquiries. Indeed, this study concludes that a comprehen- sive solution to the Northeast problem cannot lie solely in agriculture-but because agriculture lies at the core of the problem, it demands intense scrutiny. The following section begins to examine the Northeastern agri- cultural sector with a brief look at its origins, at the subsequent growing realization that a problem exists in agriculture, and at the inadequacies of the numerous indirect attempts to solve the Northeast problem. 2. Carlos Geraldo Langoni, Distribuicao da renda e desenvolvimento economico do Brasil (Rio de Janeiro: Editora Expressao e Cultura, 1973). INTRODUC1ION AND SUMMARY 5 Development, Decline, and Government Intervention The narrow, humid coastal strip of land extending no farther than seventy miles inland along the easternmost part of Brazil is called the Zona da Mata (forest zone). Tropical rain forests originally flourished on its dark, rich soils, but in the early sixteenth century cultivation began to transform Brazil's northeast coast into the world's most productive source of sugarcane.' Early Agricultural Development Portuguese merchants and explorers arrived in the Zona da Mata in the early 1 500s, establishing intermittent trading posts along the coast and, not long thereafter, more permanent settlements under captaincies granted by the Portuguese crown. The captaincies, appearing as early as 15 30, consti- tuted the first attempt to secure a springboard for colonization, exploration, and commerce in Portuguese America. They provided the base for a thriving brazilwood trade in the sixteenth century and an even more prosperous sugarcane colony in the seven,teenth century.4 The sugarcane crop has been a potent influence in Brazil's economic, political, and social history.5 Indeed, Northeastern sugar provided the first growth stimulus for the entire countrv, drawing large numbers of settlers from Europe and slaves from Africa and yielding export earnings that financed expansion southward and westward. Another major influence in the development of the Northeast region was the Dutch occupation during the first hialf of the seventeenth century. Ironically, the period of their occupation coincided with the most prosper- ous years of the Brazilian sugar economy. Although the Dutch never attempted to reform or contribute to the Northeastern socioeconomic structure, they did make every effort to assimilate, as much as possible, the technological and organizational aspects of the sugar industry. When, after a long struggle, the Dutch were evicted by the Portuguese in 1654, they proceeded to apply this wealth of knowledge to the creation of their own 3. Vivianne Lake contributed parts of this section. 4. Fernando de Azevedo, Brazilian Culture: An Introduction to the Study of Culture in Brazil, trans. William Rex Crawford (New York: Macmillan Company, 1950), p. 44. 5. For a fascinating discussion, see Gilberto Freyre, The Masters and the Slaves: A Study in the Development of Brazilian Civilization (New York: Alfred Knopf, 1946; abridged ed., 1964). 6 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL sugar industrv in their new Caribbean colonies. This new sugar source was equally as productive as the Northeast, and the resulting glut depressed the world price of sugar for the ensuing century. From the second half of the seventeenth century, the sugar industrv in the Northeast began to decline, and the impetus of Brazil's development moved south with the discovery of gold and other minerals. The excellent harbors and more fertile agricultural land in the South attracted settlement and agricultural development, particularlv of coffee, to that region. Even sugar enterprises turned out to be two to three times as productive in the South as in the Northeast. Thus began a long period of stagnation and relative decline. The vestiges of the sugar-producing system hampered the introduction of the economic and social changes required to avoid agricultural stagnation in the region. The sugar plantations with their usinas, or sugar mills, were vast enterprises not unlike city-states, relying on slave and indentured or conscripted Indian labor. Sugar was cultivated to the exclusion of virtuallv all other crops. Workers were frequently prohibited from planting subsistence crops, in an attempt to preserve all the land for sugar or to bind them further to the usina. Production technology advanced little over centuries, relying on manual labor except in the use of draft power to transport the product. The socioeconomic and political structures of the Northeast as formu- lated during the 1600s remained virtually unchanged for centuries as well; nevertheless, each of the factors related to the sugar industry-the planta- tion, slavery, and free labor-exercised a profound influence on the de- velopment of the Northeast life stvle and its dispersion throughout the interior regions. The captaincy, for example, spawned the plantation system, which was further subdivided by the landowner into plots rented out to individual farmers. This system, in turn, was emulated in the Northeast interior in the form of estates and sharecropped lands. Thus, from the onset of coloniza- tion, the pattern was set for a rural social structure based on economic dependence rather than a legal-political form of social organization.6 Slave labor, a factor of great sociocultural significance in the complex network of economic ties, was introduced simultaneously with the sugar plantation as necessary for continued agrarian growth. Its inherent charac- teristics of "patronage and paternalism" reverberated much beyond aboli- tion in 1889 and influenced the very nature of socioeconomic relations in Brazil as a whole.7 6. Shepard Forman, TheBrazilianPeasantr- (New York: Columbia University Press, 1975), p 23. 7. Ibid., pp. 21-22. INTRODUCTION AND SUMMARY 7 At the other end of the agricultural labor spectrum arose a new class of tenant farmers, the result of the subdivided estates. These tenant farmers represented the first form of free labor in Brazil, and when sugar production fell in the second half of the seventeenth century, many unemployed farmers became the driving force toward settlement of the interior. Agricultural development progressed inland slowly, to the Agreste and the drought-prone Sertao, but poor access to markets, inadequate or unreli- able rainfall, and rough topography have hampered the development of commercial agriculture in the interior even to the present day. Such agri- culture as developed in the interior was of three types: slash-and-burn nomadic agriculture; small-scale subsistence farming in scattered areas with adequate water supply, such as river valleys; and large estates and planta- tions. The latter were similar in institutional structure to the usinas and also relied on dependent labor, but their products were more diverse. Such conditions led neither to the development of family agricultural enterprises nor to adaptable commercial farms. Though not conducive to farming, the Northeast interior proved suitable for the expanding cattle-breeding population of the coast. The sugar planta- tions' growing needs for draft power and food led to a disproportionate expansion of the cattle-breeding sector whose nomadic nature, in turn, became increasingly incompatible with the sugar-producing economy. As a result, the Portuguese government finally prohibited cattle breeding in the coastal sugar zone and created, in effect, a new and separate subsistence sector in the interior region.' While the coastal economy was increasingly disturbed by external eco- nomic factors and its dependence on imports of labor and capital, cattle breeding continued to extend farther into the Agreste and Sertao unencum- bered by external dependencies. Its main asset reproduced itself automati- cally, entailing few major costs, and the demand for meat and hides continued to increase as a natural consequence of human population growth.' Cattle breeding was well suited to the climate and agricultural conditions of the interior and thus provided the first major stimulus to the penetration of a region that, otherwise, would have long remained bleak and neglected. The second major incentive to inland migration was the discovery in the early 1700s of gold and other mineral deposits in central Northeast Brazil. Following trails already blazed by cattle raisers, the first mining pioneers, 8. Celso Furtado, The Economic Growtb of Brzzil: A Survey from Colonial to Modern Times, trans. Ricardo W. de Aguiar and Eric Charles Drysdale (Berkeley: University of California Press, 1963), p. 62. 9. Ibid., pp. 67-68. 8 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL known as the "bandeirantes of Sao Paulo" (mestizo sons of white and Indian parents), initiated a new economic utilization of the interior. '0 The respec- tive quests of miners and cattle breeders soon became intertwined and interrelated: The growing opportunities gained through territorial explora- tion and expansion resulted from the respective efforts of both groups, to their mutual advantage." Thus, cattle paths soon developed into well-traveled trails along which towns and villages began to emerge. Cattle and gold not only enhanced the prospects of Northeastern economic growth but also, and perhaps of more significance in the long run, led to increased national unity. 2 The breeding and transporting of cattle and the search for new pasturage and mineral wealth decreased distances between North and South and caused East and West to become more interdependent. The 1800s witnessed the introduction of cotton to the Sertao as another impulse to internal migration and development of agricultural estates. During a relatively long drought-free period (1845-76), the development of a new strain of drought-resistant cotton attracted masses of rural laborers from the Agreste, thus adding an element of agricultural stability to the otherwise transient nature of the Northeast settler. The increasing numbers of farmers, cattle breeders, miners, and mere adventurers who migrated inland represented a growing subsistence sector emanating from the coastal population because of the steady decline of the sugar economy. In the words of Celso Furtado, "the expansion of the economy of the Northeast during this long period consisted of a process of retrograde economic evolution: the high-productivity sector was losing its relative importance, whereas productivity in the stockbreeding sector de- clined in proportion to its own expansion."" Thus, while the perimeters of civilization continued to push farther southward and westward, the tech- nological tools either remained unchanged or reverted to more primitive stages to adapt to the more primitive standards of living of the Northeast interior. Though the nature of settlement in the Northeast, and more especiallv in the interior, was partially responsible for delaving the process of its own development, a number of other factors-climatic, socioeconomic, and political-have further impeded the development process from the 1 800s to the present day. 10. de Azevedo, Brazilian Culture, p. 48. 11. Ibid., p. 50. 12. Ibid., p. 51. 13. Furtado, Economic Growth of Brazil, p. 70. INTRODUCrION AND SUMMARY 9 Insufficient and irregular rainfall is a major characteristic, if not a hand- icap, of the Northeast Sertao. Severe droughts occur about once every hundred years, and milder droughts approximately every ten years. Such periods are characterized by the aggravation of poverty, starvation, agrar- ian unemployment, and migration.'4 Yet, historically, the Northeasterner has been unable to avert the possibility of future disaster, waiting and relying, instead, on government assistance. The government, in turn, faced the same unpredictable climatic situation and thus responded only when the need arose, without any long-term planning or policy objectives. Yet it was hardly the climate alone that determined governmental reac- tion to the Northeast problem. From the mid-1700s, when the sugar industry was declining rapidly in the Northeast, mining was becoming a lucrative alternative in the South, the nation's capital was moved from Bahia to Rio (1762), and Brazil's politico-economic centers of activity began thriving in the Center-South. As a result, the Northeast found itself isolated from the major impulses to growth, not only geographically but, of even more significance, politically, economically, and socially. Further- more, since the creation of the republic in 1889, the Northeast had been divided into many states smaller than those in the South, thus reducing, comparatively, the amount of power the former could wield in the government. "1 During the past century, the impact of droughts has been increasing in direct proportion to the increasing population of the Northeast. Intermit- tent attempts have been made to deal with this climatic problem by indi- viduals as well as by federal agencies, but it is only in the past two to three decades that the government has sought to formulate a more systematic and comprehensive development policy. By the late 1940s, even before Brazil's "economic miracle" had emerged, the plight of the Northeast and its growing disparity from the rest of the country had been recognized as a deep-seated and growing problem, not simply a transitory one overshadowed by the successes of the growth centers of Rio de Janeiro and Sao Paulo.'6 Social disturbances focused attention on the Northeast as conditions worsened. The growing income 14. Anthony L. Hall, Drought and Irrigation in North-East Brazil (Cambridge, England: Cambridge University Press, 1978), p. 1. 15. Albert 0. Hirschman,Journeys toward Progress: Studies of EconomicPolicy-Making in Latin America (New York: Twentieth Century Fund, 1963), p. 17. 16. Roberto Cavalcanti de Albuquerque and Clovis de Vasconcelos Cavalcanti, "Regional Development Policies in Brazil: Precedents and Prospects," Seminar on Regional Develop- ment/Brazil, European Economic Community (Brasilia, October 17, 1972), p. 4. 10 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL gap could not be overlooked: By 1955 the state of Sao Paulo had a gross product 2.3 times that of the Northeast, although the latter had twice the population. " Government Intervention Until relatively recentlv, the underlying causes of the plight of the Northeast have commonly been viewed as environmental rather than eco- nomic or social, and most of the concentrated interventions in this century have been in response to specific droughts. This governmental tendency to take action only in response to impending dire need, together with the presence of Northeasterners in the decisionmaking body, were two kev elements influencing the realization and impact of governmental assistance programs. 18 Government interventions to alleviate the effects of drought date from 1906 (in response to the devastating drought of 1877) and have had two objectives: first, to provide relief through short-term employment in public works; second, to achieve a lasting solution to the drought problem by constructing water storage facilities-the so-called hydraulic solution.'9 Under the auspices of the Department of Works against the Droughts (a newly created federal agency that, having undergone several name changes since then, is known today as DNOCS), the hydraulic program and a serious agricultural research effort were initiated to help the Northeast prepare itself for future droughts. As many as 536,000 workers were employed in temporary relief works in 1958 (when this total amounted to 13 percent of the population affected by drought); such a large-scale mobilization is only one example of the depth of official concern for the region and its poverty. TFhe public and private dams constructed in the Northeast probably num- ber in the thousands. By 1955, public reservoirs had a storage capacity of 7,800 million cubic meters, increasing to 11,400 million cubic meters in 253 dams by 1972.20 After the calamitous drought of 1958, in which thousands starved and tens of thousands more were uprooted, attention turned from the shortage 17. Furtado, Economic Growth of Brazil, p. 265. 18. Ibid., pp. 4-6; Hall, Drougbt and Irrigation, pp. 5-7; and Stefan H. Robock, Brazil's DevelopingNortheast: A Study of Regional Planning andForeign Aid(Washington, D.C.: Brookings Institution, 1963), p. 69. 19. Hall, Drought and Irrigation. 20. Ibid., p. 7. By comparison, Egypt's share of the net gain in storage from the Aswan High Dam amounts to only 7,500 million cubic meters. See H. A. El-Tobgy, Contemporary Egyptian Agriculture (Beirut: Ford Foundation, 1974), p. 28. INTRODUMrION AND SUMMARY ii of water to its distribution and use. The failure of the hydraulic solution and the irrigation program that followed it have been extensively docu- mented by Hall, Hirschman, and others.2 The reasons behind the failure are still a matter of debate. Certainly the hydraulic solution suffered from a myopic, if not naive, view of the drought problem, and the irrigation program from poor management, insufficient technical expertise, and cor- rupt practices in the distribution of aid. It has also become clear that, although water storage may reduce the consequences of drought for people and, especially, animals, the irrigation potential in the Northeast is insuf- ficient to warrant any large-scale program of investment. Hall feels that no level of investment in water resources will overcome the structural bot- tlenecks in the Northeastern economy, which not only prevent resources from being channeled to the rural poor but also impede economic growth. EIarlier, Robock argued that "from the standpoint of formulating develop- ment programs for the region, it is crucial to recognize that the most basic problem of the Northeast is not the periodic drought but continuing poverty."22 Significantly, an economic geographer comes to an almost identical conclusion: "The fact that there is a coincidental relationship be- tween the drought area of Northeast Brazil and the area of greatest poverty does not mean that there is a causal relationship between them."2" The weaknesses inherent in the early intervention attempts came under attack during the administration of Epitacio Pessoa (1919-22), the first and only president from the Northeast. Although his plans for large-scale projects subsidized by foreign capital proved far too ambitious for Brazil's absorptive capacity at the time,24 his efforts underlined a fundamentally misconstrued perception of government policymakers: In a country where a prosperous and an indigent region exist side by side, governmental intervention and spending are generally directed toward the growing de- mands of the former until its thriving economy permits surplus funds to be invested in the poor, stagnant region. Such a tendency might imply that the Northeast problem was not so much a climatic consequence as a result of political and administrative incompetence and inefficiency.25 In the 1940s, Jose Augusto Trindade, the first director of the DNOCS agricultural research program, further emphasized the need for consistent and effective government intervention, particularly in the form of irrigation 21. Hall, Drought and Irrigation, pp. 55-108; Hirschman, Journeys toward Progress. 22. Robock, Brazil's Developing Northeast, p. 8. 23. Kempton E. Webb, The Changing Face of Northeast Brazil (New York: Columbia Uni- versity Press, 1974), p. 178. 24. Elirschman, Journeys toward Progress, pp. 30-31. 25. Ibid., pp. 36-37. 12 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL projects. Because Northeasterners had virtually never before been exposed to modern irrigation technology, they were naturally hesitant to integrate it into the region's hitherto primitive agrarian svstem. The government's responsibility was, therefore, to redirect the inhabitants' socioeconomic behavior by gathering and disseminating information, clarifying their perceptions of the Northeast situation, and reorganizing land distribution. Not until the aforesaid conditions were met, or at least approximated, would government programs actuallv vield positive and long-term results. Trindade's perceptions of the Northeast problem have constituted the fundamental guidelines for intervention programs since the 1940s, but it was not until almost twenty years later, during the period Robock has termed "the new era," that serious efforts began.26 Before then, however, several other attempts had been made to alleviate the Northeast problem. In hopes of providing new growth impulses to the underdeveloped Sao Francisco River Valley, the federal government cre- ated the Sao Francisco Vallev Commission (CVSF) in 1946, followed shortlv thereafter bv the establishment of the Sao Francisco Hvdroelectric Com- pany (CHESF). Though the performance of CVSF was rather ineffectual because of ill-defined objectives and geographically dispersed projects rely- ing on annual funding (as opposed to lump sums), CHFSF proved more successful in fostering growth incentives in the Sertao by providing an abundant new source of inexpensive energy. Nevertheless, CHESF remained primarily an engineering project and, like CVSF, was unable to transform the Sertao economy." Similar criticisms were leveled against the DNOCS bureaucracv. Bv the late 1950s it was channeling most of its funds into dam construction projects, instead of into more needed but controversial irrigation projects, as a result of the growing affiliation between DNOCS directors and engineers and Sertao politico-economic leaders and powerful landowners. Because no efforts were being made to promote self-sufficiency, the Northeast was perceived as a continual sieve for investments that could otherwise prosper in the Center-South.28 Following a severe drought period in the early 1950s, DNOCS authority began to decline as its continued water storage policy, in a time when irrigation was so drastically needed, became increasingly conspicuous.29 26. Robock, Brazil's Developing Aortbeast, p. 88. 27. Martin T. Katzman, Cities and Frontiers inB razil: RegionalDimensions of Economic Develop- ment (Cambridge, Mass.: Harvard University Press, 1977), pp. 134-35. 28. Ibid., p. 138. 29. 1 iirschman, Journeys towvard Progress, p. 76. INTRODUCTION AND SUMMARY 13 A regional development bank, the Bank of the Northeast (BNB), was created about this time, marking "the first official government acceptance of the economic solution as a federal policy for the Northeast. "'0 Unlike most federal development agencies, the BNB was situated in the Northeast to attract individuals dedicated primarily to the advancement of that region and to maintain readily available funds specifically for long-term invest- ment loans that would encourage regional development." The BNB there- fore had the twofold purpose of promoting agricultural and industrial growth in the Northeast and simultaneously helping to prevent the con- tinual loss of financial and administrative resources to the Center-South. Ultimately, however, the BNB'S performance was affected by many of the same socioeconomic and political pressures that had constrained Northeast- ern development in the past. Thus, in response to the calamitous drought of 1958, when disparity in income levels was reaching an all-time high despite the BNB'S limited success in stimulating investment, the Working Group for the Development of the Northeast (GTDN), headed by Celso Furtado, produced a report entitled "A Policy for the Economic Development of the Northeast."'2 The Furtado report was an updated and condensed version of various diagnoses and policy prescriptions that Furtado and his colleagues (at BNB and elsewhere) had been producing for several years. It has been heralded as a "highly skillful document" and attacked as "a political docu- ment cloaked in the authority of technical economic analysis. ""' Irrespective of the outcome of the debate, few can argue that it has not been the most important document affecting modern Northeastern history. The Furtado report not only plunged national thinking firmly into the "new era" con- cerning policy prescriptions for the Northeast, it also (perhaps more cor- rectly, Furtado himself) forged the institutional framework (SUDENE) and the programs that had, at least, a strong potential for dealing with the Northeastern problem. Significantly, it emphasized that the Northeast had to be treated as an integral part of the greater nation of Brazil, and its deteriorating economy as part of a national, rather than regional, problem. Furtado recognized that the Northeastern economy, dependent on sugar and cotton exports to foreign markets as well as food imports from the 30. Robock, Brazil's Developing Nortbeast, p. 88. 31. Hirschman, Journeys toward Progress, pp. 62-6 3. 32. Grupo de Trabalho para o Desenvolvimento do Nordeste, Uma politica de desenvol- vimento para o Nordeste (Rio de Janeiro: Conselho de Desenvolvimento, 1959). This historic document is still a point of reference for development policy in the Northeast. Indeed, SUDENE'S participation in the survey employed in the present study was based on the work program outlined in that document. 3 3. 1Ifirschman,Journeys towardProgress, p. 7 3; Robock, Brazil's Developing Aortbeast, p. 11 1. 14 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Center-South, had been disadvantaged by overvalued exchange rates and high rates of inflation, a combination that led to a severe deterioration in the interregional terms of trade. This, in turn, led to underinvestment in Northeastern agriculture and a steady capital outflow to the more dynamic regions. His prescription called for a radical transformation of the regional economy, to be achieved through a comprehensive strategy to strengthen the infrastructural, industrial, and agricultural foundations of the North- east. The key elements consisted of industrial development based on the region's singular comparative advantage, cheap and abundant labor; emigration of the region's surplus population to colonies in Maranhao and to unsettled public land in the more humid areas of central Bahia; and a "more rational" use of land in the sugar zone to produce cheap food for the industrial labor force. The Furtado prescription entailed further complexities, of course. For political reasons it did not directly address certain issues, and for lack of data it was not couched in convincing empirical terms. 4 Nevertheless, it exhibited the clearest understanding of the root of the Northeast problem: too many people on too little land depending on an agricultural system that needed to be made "more rational." That the creation of SUDENE resulted directly from the Furtado report and that the majoritv of subsequent programs in the Northeast owe their origin to that report is now history. These programs have comprised remarkably intensive and innovative regional development strategies, the more important of which are reviewed here. COLONIZATION. Efforts to colonize western and northern Brazil, includ- ing the Amazon region, have been in progress since the rubber boom of the 1940s. Although the GTDN's recommendations concerned the largely virgin Maranhao region, national policy was directed more toward the Amazon region, as was manifested most vividly by the Trans-Amazonic highway. Provisions were made for agricultural settlements along this corridor, and pa-t of their justification was the alleviation of supposed population pres- sure in the Northeastern interior. A variety of programs and institutions have been created to promote this westward push. 34. Admission that the Northeast was overpopulated and, consequently, that emigration was needed was unpopular because it drew attention to the inadequacies of the agricultural system. Furtado resolved this problem by redefining the Northeast to include the frontier state of Maranhao. Land reform in the sugar zone, of course, has been politically explosive in the Northeast, and Furtado has been "noncommittal and reserved" (Ilirschman,Journeys toward Progress, p. 78) on the topic, calling instead for "more rational land use" and "reorganization." INTRODUCrION AND SUMMARY 15 Though it is too early to evaluate the colonization efforts, there are three indications that such programs will not be sufficient to solve the Northeast problem. First, emigration has proven to be a highly controversial issue among landowners who have traditionally benefited from the comparative advantage of abundant, cheap labor. Sociopolitical leaders have also re- jected the colonization proposal because it would reduce their constituen- cies. Second, where colonization has in fact taken place, agricultural pro- ductivity in these areas has been exceedingly disappointing. Even in Maranhao where rainfall is sufficient, yields are poorer than in the drought- prone Sertao, and soil erosion usually confines farming to slash-and-burn techniques followed, after two or three years, by cattle ranching. Third, much of the settlement has been spontaneous; the record of official pro- grams has been poor. By 1972, the Brazilian Institute for Agrarian Reform (IBRA), the agency directly responsible for settlement, had managed to establish only 5,000 families on less than 5 00,000 hectares." Cavalcanti and Cavalcanti note that even if official settlement goals are reached the various schemes will annually accommodate a number of families equal only to the net annual increase in the rural Northeast.36 If this judgment is correct, colonization can at best only keep pace with the increase in the rural poor; by no means can it be expected to eliminate poverty. INDUSTRIAL INCENTIVES. Having ruled out conventional development policy options such as the expansion of exports, the Furtado group empha- sized the need for industrialization to develop the Northeast from within by providing employment for the transient Sertao population, establishing a regional entrepreneurial class, and keeping capital in the Northeast." A fourth impulse to internal development would be the creation of an active industrial center that would generate incentives to growth throughout the Northeast through backward linkages. The recommendation of GTDN that had the greatest impact on the North- eastern economy was the creation of the Northeast Development Superin- tendency (SUDENE) in 1959. Shortly after it was established, SUDENE began promoting industrial development in the Northeast through a highly in- 35. William H. Nicholls, "The Brazilian Agricultural Economv: Recent Performance and Policv," in Brazil in the Sixties, ed. Riordan Roett (Nashville, Tenn.: Vanderbilt Universitv Press, 1972), p. 174. In 1971 IBRA was merged with two other agencies into the National Institute of Colonization and Agrarian Reform (INCRA). 36. Cavalcanti and Cavalcanti, "Regional Development Policies in Brazil," p. 35. 37. Katzman, Cities and Frontiers in Brazil, p. 136. 16 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL novative system of fiscal incentives named after the law (34/18) by which it was enacted in 1961. This law enabled Brazilian corporations to reduce their tax liabilities by up to 50 percent by making deposits in a special fund established in the BNB and administered by SUDENE. This fund could then be drawn upon for private investment in the Northeast in projects approved by SUDENE. The 34/18 program resulted in a massive redirection of interregional capital flows, which has brought about an impressive transformation of the re- gion's industrial structure. Debates continue as to its success in meeting its stated goals,38 but it seems clear by now that the program has had very little effect on the rural economy. During the scheme's first eight years, only 0.4 percent of the funds were invested in the poorest state, Piaui, generating employment for only 275 persons.39 Nor has the program had an appreciable impact on the region's underemployment problem: By 1976, no more than 200,000 jobs were to have been created, a number amounting to 5 percent of the urban labor force and only 11 percent of the incremental growth in the labor force since 1960.40 If 34/18 has failed to solve the Northeast problem, it is not only because of its inherent emphasis on subsidizing capital in a situation of excess labor but also because the projects supported have been mainly confined to the coastal urban centers and have had very limited backward linkages to primary production. LAND REFORM. Land reform has been adequately legislated, staffed, and funded but has not enjoyed sustained support. Indeed, there has been substantial official reluctance, along with the traditional opposition of littoral and Sertao landlords, to undertake reform on any appreciable scale despite widespread recognition of the inherent inefficiency of unused 38. See, for example, David Goodman, "Industrial Development in the Brazilian North- east: An Interim Assessment of the Tax Credit Scheme of Article 34/18, " in Roett, Brazil in the Sixties; David Goodman and Roberto Cavalcanti, Incentivos a industrializacao e desenvolvimento do Nordeste (Rio de Janeiro: IPEA/INPES, 1974); and D. E. Reboucas, "Interregional Effects of Economic Policies: Multisectoral General Equilibrium F.stimates for Brazil" (Ph.D. disserta- tion, Harvard University, 1974). Goodman and Cavalcanti argue that the tax incentive scheme concentrated on capital-intensive sectors and failed to generate the expected employment opportunities. Reboucas, however, reinterpreting the evidence through a two-sector general equilibrium model, concludes that in the years 1959-70 the net result of the scheme was to increase industrial employment by about 40 percent over what it would otherwise have been. 39. Goodman, "Industrial Development in the Northeast," p. 260. 40. World Bank, "Rural Development Issues and Options in Northeast Brazil," report no. 665a-BR (Washington, D.C., June 23, 1975; restricted circulation), p. 116. INTRODUCTION AND SUMMARY 17 land's being concentrated on large estates while underemployed labor is abundant.4' In evaluating the effects of these three programs, it is important to stress that, despite Furtado's plans, they were introduced piecemeal, rather than being adopted as part of a comprehensive strategy for the development of the region. By 1961 SUDDENE was emphasizing the lack of such a strategy: The growing populational pressure in the Northeast and the structural incapacity of the region's economy to absorb the surplus population have been the chief cause of some social and political problems of the utmost seriousness. These problems are aggravated by the fact that the North- eastern economy is based substantially on a subsistence agriculture, practiced mostly in zones of poor soils and subjected to periodic droughts. Such problems can be thus summarized: (a) a general climate of dissatisfaction; (b) the upsurge of much resentment toward the coun- try's most developed areas; (c) the rising of rural workers' leagues claim- ing [that] the solution to the problem [is] immediate access to landown- ership; (d) increasing unemployment; (e) diminishing prestige of the Public Power in the greatest populational segments. All these facts, which do constitute a threat to the unity and the internal security of the country, come to a great extent from the inexis- tence of an overall policy for the economic development of the Northeast, as well as from the failure of only partial solutions that have been tried. The lack of an integrated conception of the economic problems that assail the Northeast has led the Public Power to a perfunctory action, involving high and growing expenditures of a merely assistential [sic] character and investments not always coordinated and complemented, with small positive repercussion on the precarious living conditions of the Northeastern populations." This assessment was apparently still valid eleven years later, for Cavalcanti and Cavalcanti stated in 1972 that "it is impossible to identify a prevailing policy in the Northeast, either of agricultural development, or industrial development or emigration."43 Even SUDENE, which in its first decade represented a politically and economically cohesive force in the Northeast, became guilty of the same 41. Comite Interamericano de Desarrollo (CIDA), Land Tenure Conditions and Socioeconomic Development of the Agricultural Sector, Brazil (Washiigton, D.C.: Pan American Union. 1964). 42. SUDENE, "The Brazilian Northeast and Its Institutions for Economic Development" (English version, Recife, 1961), p. 3. 43. Cavalcanti and Cavalcanti, "Regional Development Policies in Brazil," p. 31. 18 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL weaknesses charged against its predecessors. Its early role as an increasingly powerful agency and its determination to transform the traditional socioeconomic infrastructure brought SUDENE under stronger attack by the power elite which upheld that infrastructure.4 Molded by such pressures, SUDENE programs were thus redirected toward tvpically noncontroversial transportation, energy, and water storage projects. A severe drought in 1970 led to the creation of yet another federal program, PROTERRA (Program for Land Redistribution and Stimulation of Agro-Industry in the Northeast), whose financial resources, amounting to about US$1,000 million, by 1976 had grown to US$4,000 million, a figure twice the value of the gross annual production of the Northeast and an investment that attests to the level of official commitment to the Northeast's problems. Though the primary intention was to redistribute land in the Northeastern areas of greatest tension, this program has gradually been forgotten and become a line of credit that has actually made PROTERRA a mechanism of economic concentration. 45 As of the last available tally (1975), 17,000 hectares of land had been redistributed to less than a thousand recipients .46 That none of the above approaches had successfully addressed the rural poverty problem of the Northeast was apparent from the creation in 1974 of a major program designed specifically to raise productivity and rural in- comes, particularly for small- and medium-scale farmers and sharecrop- pers. POLONORDESTE (Development Program for Integrated Areas in the Northeast) is a well-financed and staffed set of individual projects, alreadv undertaken and in various stages of planning. The project areas are selected either because they have intense poverty problems or because they hold promise for rapid development with an appropriate infusion of modern inputs and techniques. For the most part, POLONORDESTE operates within the existing agrarian structure and the policies pertaining to it. Unlike previous programs dedicated to the establishment of new public works and social services, POLONORDESTE'S main objective was to redirect and fortify existing projects to achieve greater efficiency and productivity. Despite POLONORDESTE'S potential for alleviating pockets of poverty and raising the productivity of its target groups, it does not provide "an overall policy for the economic development of the Northeast." The continuing 44. Hirschman, Journeys toward Progress, p. 89. 45. This sentence is translated from a quotation by former minister of agriculture Luis Fernando Cirne Lima, inJornal do Brasil, November 20, 1977; it echoes William Cline's (and others') characterization of the program. 46. World Bank, "Rural Development Issues and Options in Northeast Brazil," p. 31. INTRODUCTION AND SUMMARY 19 plight of the Northeast and the scope of efforts to deal with it are perhaps best illustrated by the scale of recent drought relief works. From January 1976 to January 1977, for example, SUDENE spent about Cr$1,200 million (about US$100 million at that time) out of a fund of Cr$2,500 million to finance temporary work for 279,000 peasants.4' This expenditure was roughly equal to the annual government contribution to POLONORDESTE. In a bad drought year, the number of workers seeking relief employment can reach I million. Yet the creation of PROTERRA, POLONORDESTE, and EMBRAPA (the Brazilian Agricultural Research Company that has recently exhibited substantial concern for the Northeast) indicates that policymakers in Brazil are giving increased attention to the notion that the Northeast problem is rooted in agriculture and cannot be solved easily through roundabout policies and indirect interventions, no matter how well administered and funded. The Approach Followed in This Study The foregoing discussion has hypothesized that the root of the Northeast problem lies in agriculture, but numerous programs and policies have not significantly improved the state of the rural Northeast for decades, perhaps centuries. If the Northeast problem originated in and is perpetuated by the agricultural system, this system demands close scrutiny, unencumbered by political passions and enhanced by quantitative analyses. In short, the solution to the problem demands "fresh" numbers and "fresh" tools. The SUDENE/World Bank survey provides the fresh numbers, and modern econometric and mathematical programming techniques provide the fresh tools. The approach followed in this study is twofold. First, we shall use data collected by a joint SuDENE/World Bank survey of 8,000 farms to describe the salient features of Northeastern agriculture: the size and distribution of the resources; the products and the inputs employed in the various farming systems; incomes and their distribution; and the performance of the various agents. Second, we shall employ a large-scale mathematical programming model of the agricultural sector, which represents farms of different types in different physiographic zones, to investigate the constraints in North- eastern agriculture. The model will then be used as a simulation tool to explore alternative means of relaxing those constraints and to estimate the 47. "O feijao e o sonho," Veja, March 2, 1977, pp. 29-30. 20 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL tE MARANHAo NT l. A S ~~~~~~~eresin !\ ICEARA 1 RIO NDE X - < 00 P| AU} PARArnA PERNAMBUCO I2 I 8AI-IvA 1 r ° 9 0 : 0 SB A R I A GOIAS MINAS &rRAZIL Notrs INTRODUCTION AND SUMMARY 21 benefits that might accrue from various policies or programs to assist the agricultural sector. The remainder of this first chapter is a summary of the study. Summary and Conclusions Climate and ecology in the Northeast are extremely diverse, and even within the smallest of its nine states as many as three different physio- graphic zones can be distinguished. This diversity is reflected in farming systems and in production and marketing patterns and may be the single most important reason the formulation of agricultural policy has been so difficult and disappointing. Most of the analysis in this study divides the Northeast into seven physiographic zones delineated in a SUDENE study." These are shown on the accompanying map and described in the appendix to this chapter. They are, briefly: -The West (zone A), a long strip comprising the westernmost portions of Maranhao, Piaui, and Bahia, of low population density, poor soils, though adequate rainfall, and slash-and-burn subsistence agriculture by often untitled squatters coexisting with extensive cattle ranches. -The Middle-North (zone B), comprising the northern and coastal portions of Maranhao and Piaui; very similar to the West except for its closer proximity to the markets of Teresina and the port of Sao Luis and the predominance of an inferior variety of rain-fed rice, arroz seco. -The Sertao (zone C), the vast drought-prone interior region where cotton is the most important crop and sharecropping on largefazendas, or plantations, is the dominant form of enterprise. -The Semihumid Southeast (zone D), comprising most of the state of Bahia, with sufficient and timely rainfall and a wide variety of farming operations: subsistence farmers producing mainly manioc; large cattle ranches, which are the dominant agricultural activity; and cacao plantations. -The Humid East or Zona da Mata (zone E), the traditional sugarcane area along the coastal portions of Bahia through Rio Grande do Norte. -The Humid Southeast (zone F), a fertile pocket in the state of Bahia where cacao plantations dominate economic activity. 48. M. Lacerda de Melo, "Espacos geograficos e politica espacial: 0 caso do Nordeste," Boletim economico, vol. 5, no. 2 (July 1967/December 1971), pp. 7-139. 22 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL -The Agreste (zone G), a transitional zone between the drought-prone Sertao and the Humid East, characterized by mixed farming on small to medium-size enterprises and only rarely subject to drought. Most observers have concluded that the heart of the Northeast problem lies in the paucity of its resource base. Chapter 2 describes the labor force, land, and capital in the region. Of the 18 million rural Northeasterners, between 5 and 8 million are members of the labor force, depending upon the definition used. The demographic census of 1970 found 5.2 million persons "economically active in agriculture." The agricultural census of the same year found 7.8 million persons "occupied in agriculture," of whom 2.2 million were "producers," in the sense of being a landowner, renter, sharecropper, or "occupier." The SUDENE/World Bank survey undertaken for the present study dis- tinguished different groups in the agricultural labor force according to the tenurial arrangements that affect their access to land: proprietors or land- owners; renters, who are relatively rare in the Northeast and pav a fixed fee in money or product for the use of land; sharecroppers, who pay the landlord a share of their product, usually under an annual verbal contract; squatters, who work public or private land without compensating the owner and often without the owner's knowledge; permanent workers, who are either moradores, resident workers who receive, as part of their com- pensation, a dwelling and a small parcel of land for family production of subsistence crops, or nonmoradores, who are contracted annually but receive no housing or land; and temporary workers, who are employed when labor demand is at its peak and are paid dailv or weeklv at rates slightly more than US$1 a day.49 The survey reported a total of 5.7 million agents in agriculture, and we take the figure of 6 million as a compromise estimate of the agricultural labor force. Of the 6 million, less than 800,000 own land. Another 900,000 are working members of landowners' families; 26,000 are renters; and 1. 14 million are either sharecroppers or permanent workers. Nearly 3 million agricultural workers have no formal or legal access to land and subsist on temporary employment or scratch out a living on landholdings so poor or so 49. Unless otherwise stated, the U.S. dollar figures in this study were converted from cruzeiros at the CR$6.4:US$1 rate prevailing in spring 1974 (the harvesttime in the Northeast, during which the largest number of monetary transactions occur). Because manv observers believe the cruzeiro to have been overvalued by 20-25 percent during that time, the 1974 dollar figures quoted in this study may be unrealistically high. INTRODUtCTON AND SUMMARY 23 remote that they were not registered by any of the censuses, the cadastral survey, or the SUDENE/World Bank survey.50 About 50 percent of the available labor appears to be underemployed. Seasonal fluctuations clearly account for part of this, but the lack of continuous access to productive land seems to be the single greatest cause of unemployment and poverty in the rural Northeast. Most landowners (even if they own only two or three hectares of poor land), sharecroppers, and permanent workers obtain twice or three times as much employment in a given year as do temporary workers. Temporary workers and their depen- dents make up the hard core of poverty in the Northeast. On the average, these workers are employed for only 5(}60 days a year, at wage rates that give them family incomes of only about US$200 a year (1974 dollars). It is not possible to determine the number of workers in a typical landless family or what incomes they may obtain outside of agriculture, but it is obvious that many such families have incomes far below the absolute poverty line and that there may be up to 5 million people in this group. Half of the 80 million hectares of agricultural land in the Northeast is concentrated in properties larger than 500 hectares. These farms number only about 33,000 and account for 4 percent of all landowners and a negligible percentage of all workers. One third of the farms in the North- east have an average size of five hectares, on 1.4 percent of the region's agricultural land. Farms larger than 200 hectares, a size large enough to provide an adequate family income, make up nearlv 70 percent of the agricultural land; their owners number 83,000-10 percent of the landown- ers and little more than 1 percent of all agricultural workers. In essence, 99 percent of the labor force does not have entrepreneurial access to nearly 70 percent of the agricultural land. The survey data make it possible to value the capital embodied in on-farm structures, livestock, and equipment.5' On this basis, the total agricultural capital of the Northeast amounts to about US$5,000 million, which gives a capital/output ratio of about 2. It is, however, concentrated in 50. This estimate of the number of temporary workers should be treated cautiously, as some may have been double-counted and some small farmers or their family workers may be included if, at the time of enumeration, they were working on farms other than their own. (Permanent workers and sharecroppers are not generally allowed to work part-time on other farms.) 51. Though the value of structures per farm appears to be correlated with farm productiv- ity, rigorous analysis of the influence of structures is difficult because it is not possible to distinguish items such as housing from directlv productive assets. 24 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL structures (which include housing) and livestock; factors such as equipment that contribute directly to modern production practices account for only 3 percent of the value of agricultural capital. Farm equipment inventories are extremely small and are particularly limited on small farms. In much of the Northeast, farms of less than 10 hectares have equipment inventories averaging about US$18-only three times the cost of a single hoe. The estimation of capital/output ratios suggests that substantial returns are possible from investment in Northeastern agriculture. Average capital/ output ratios vary little among different zones and size classes of farms, however, despite the fact that land/labor ratios vary dramatically. There appear to be severe distortions in the region's markets for capital items usually considered prerequisites for modern agricultural development. Chapter 3 discusses agricultural production and incomes. Crops account for the bulk of agricultural production in the Northeast,52 and virtually half of crop production is made up of subsistence crops, particularly rice, corn, beans, and manioc (cassava). Cotton and perennial or tree crops such as sugar, cacao, babacu, and carnauba (varieties of palm) are exportable and contribute about 38 percent of the value of crop output; the remainder is composed of fruits and vegetables that are perishable and have high income elasticities and limited markets within the Northeast. Almost all farms have animals of some type, whether cattle, pigs, goats, or chickens, though few can be said to specialize in livestock, and organized dairy production for the market is virtually nonexistent. Beef, mostly poor in quality and produced by fazendas, is by far the most important animal product in the Northeast. That the bulk of Northeastern production is inferior and consumed locally suggests that the product mix itself contributes to the region's stagnation. No product has emerged to take the place of sugar in providing an export base to fuel growth in the rural sector. Indeed, most Northeastern products, including those currently exported, are of a type or quality for which there is a low income elasticity of demand, so that the region has not been able to respond to the growth of markets either in other parts of Brazil or abroad. The gross value of agricultural production in 197 3, a year not affected by drought, amounted to US$2,100 million. This implies that the gross output per hectare of agricultural land was less than US$30 and that the output per 52. Crops accounted for between 66 percent and 75 percent of gross agricultural output in 1973, depending on the source of the estimates: The first percentage is based on survey estimates for both crops and livestock production; the second is based on a combination of Instituto Brasileiro de Geografia e Estatistica (IBGE) estimates for crops with survey estimates for livestock. INTRODUCTION AND SUMMARY 25 member of the labor force was US$400. If the Northeast is not a net importer (there is no evidence that it is) and if one ignores stocks and savings, agricultural consumption valued at farm-gate prices in that year in the region was at most only US$66 a person, since some crops are exported. A breakdown of production costs sheds some light on the reason for such low productivity. Expenditures on purchased inputs of all types (other than hired labor) average less than US$2 a hectare, but even this figure under- states the primitive nature of Northeastern agriculture. In most zones, inputs for livestock account for at least half the average expenditures. The bulk of the expenditures in the Northeast are for fertilizer on sugar and cacao plantations and for insecticides and pesticides, which in some areas are prerequisites for cropping. The "green revolution" has not touched the Northeast: Less than 10 percent of all farms use selected seeds of any variety, and many of these are "selected" because they are available only from specialized sources such as cotton gins. Income from agriculture in the Northeast amounted to about US$2,200 million, or US$120 per rural resident, in 1973. Its distribution was highly skewed. Farmers, owners, and their families accrued 75 percent of this total, although they made up only about 28 percent of the agricultural families. The remaining 2 million agricultural families earned total incomes from agriculture (both in money and in kind) of US$5 30 million, or US$265 per family. Because the average family consists of five persons, per capita incomes from agriculture do not exceed US$50 by much for these 10 million people. Permanent workers and their families probably have in- comes slightly above absolute poverty levels, but the employment and wage rates of temporary workers suggest that they are much worse off. Monetary incomes, however, may have limited meaning in the rural Northeast. The National Study of Family Expenditures (ENDEF) undertaken by the Insti- tuto Brasileiro Geografia e Estatistica (IBGE) in 1974-75 found that nearly half the welfare of rural residents accrued from goods and services pro- duced and consumed outside the marketplace. Farm incomes are highly skewed, depending partly on farm size and partly on location. In most size classes and zones, average farm incomes exceed US$250 a year by at least a factor of 2. For the class of farms of less than ten hectares in the interior regions, however, even the average income does not exceed US$207. About a fourth of the farm income accrues to the 4 percent of farms of more than 500 hectares; it is noteworthy that though farm income generally rises with farm size it does so rather erratically. Because poverty in the rural Northeast is closely linked to lack of entrepreneurial access to sufficient cultivable land, or lack of full-time employment on such land, the performance of farms of different sizes and 26 T HE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL types is examined. The analysis in Chapter 4 of factor use and productivity on farms of different sizes and types strongly suggests that the large farm sector is not using resources with maximum efficiency to produce profits. Survey data reveal that the intensity with which land and labor are used declines sharply and consistently as farm size increases, from the smallest of six size groups (under ten hectares) to the largest (over 500 hectares). The smallest farms employ about half their land in crops; the largest. only one-tenth. About 85 percent of all land on farms larger than 500 hectares is not used for production of crops or of fodder. Farms in this size class control 48 percent of the land area and contribute 31 percent of crop production and 34 percent of livestock production. The smallest farms control 1.4 percent of the land, contribute 8 percent and 7 percent, respectively, of crop and livestock production, and produce up to 100 times as much output per hectare as the largest. The reason for the difference in land use lies almost entirely in the use of labor: Small farms throughout the entire Northeast apply twenty-five to forty-five times as much labor per hectare as the largest. Different size farms perform very differently even though they have similar capital/output ratios, land quality, and crop yields and receive and pay similar prices. Farmers have substantial freedom to choose what they produce and to substitute among most factors of production, and produc- tion possibilities do not seem to differ according to farm size. Nonetheless, large farms tend to employ less labor than they would require to maximize profits, even though there is no shortage of labor available. On these farms the marginal productivity of labor is, in general, twice as great as the going wage rates. Small farms tend to employ more labor than profit maximiza- tion would warrant, probably in part because family members have dif- ficulty obtaining employment elsewhere. Labor is potentially the only factor in the region for which markets, delivery systems, and mobility are sufficient to balance demand and sup- ply. There are strong suggestions of distortions in the markets for land and capital that, even more noticeably than labor, do not appear to be allocated according to rules for profit maximization. The marginal productivity of land on large farms is far below reasonable opportunity costs, whereas for permanent land improvements (structures) and livestock the marginal rates ot return are erratic and generally below those from comparable assets outside the Northeast. For modern inputs and credit, the rates of return are consistently extremely high. Utility maximization, and farmers' expecta- tions about prices and yields, may account for the divergences from optimal behavior. An alternative explanation may be that in the use of inputs other than labor the majority of the Northeastern farmers have very little free- INTRODUCTION AND SUMMARY 27 dom of choice because credit is unavailable and delivery systems are inadequate. Although size is the single most important determinant of farm perfor- mance, farm characteristics vary substantially within the six size classes, and the need to aggregate these classes to a smaller, more manageable number for the programming model requires a more formal farm typology (derived in Appendix C, Chapter 5). Discriminant analysis based on the degree of commercialization (percentage of output marketed) and the per- centage of labor supplied by the farm family, as well as farm size, yields three relatively homogeneous types of farms: family farms, medium-size or transitional farms, and estates. Family farms have an average size of ten to thirty hectares. They typically rely on family members for at least three- fourths of their labor needs, market only about 35-40 percent of their output, and are all highly land constrained. Medium-size farms, averaging fifty to a hundred hectares, are large enough to require hired labor and market a much higher percentage of output, and they have mixed patterns of constraints. Those referred to as estates, with an average size of more than 500 hectares, are generally plantations, ranches, or mixed fazendas.53 Much of their land is not intensively used, and they generally exhibit the poorest performance in terms of output and employment. The second part of this study uses a linear programming model, which is described in Chapter 5. The model was designed to provide a consistent, quantitative framework within which to identify the factors constraining development in the sector and which could then be used to simulate the effects of policy interventions. The model differs from other agricultural sector models in two important ways. First, it is based on survey informa- tion collected specifically for it; in particular, the technology data set, based on nearly 40,000 observations of individual activities, should provide a more satisfactory base than the secondary or informal data sets that have generally been available in the past. Second, because it was realized that the constraints on agricultural production in the Northeast differ among sizes and types of farm, the full sectoral model was built up of individual representations of farm types. In each zone, three or four farm types are distinguished so that any solution to the model yields not only aggregate indexes of production, employment, income, and so forth but also a detailed account of the operations and performance of each farm type included. Because the model contains an accounting framework for em- ployment and income by type of agent (farmers of different types, share- 5 3. The generic term "estates" is used for these large farms for the sake of convenience, but many of them are not richly endowed; nor are their owners necessarily wealthy. 28 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL croppers, permanent workers, and temporary workers), the distributional characteristics of alternative solutions can be made explicit. Solutions against the data base year, 1973, reported in Chapter 6, were undertaken primarily to test how accurately the model captures the impor- tant facets of Northeastern agriculture. The simulated aggregate levels of production and labor use revealed a relatively small but consistent bias toward more production and employment than was recorded by the sur- vey. This is typical of optimization models because they do optimize, whereas the real world obviously includes some elements of nonoptimal behavior, and any mathematical representation cannot include all con- straining elements. The source of the bias, however, lies mostly in the simulated performance of the estates. The model results fully support the hypothesis raised in the cross-sectional analysis that the estates do not behave optimally in their choice of production patterns and their use of resources. These farms, controlling half of the Northeast's agricultural land, would employ nearly twice as many workers as they do at present if thev wished to maximize profits in the neoclassical sense. In sharp contrast, the model reveals that the family farms, all of which have a severe land constraint, are using family labor and working their land far more inten- sively than is optimal from the same profit-maximizing criterion. These results suggest two conclusions: First, strong elements of noneconomic behavior exist in the use of both land and labor on estates; and second, if developmental goals include employment generation, output increases, more intensive use of the resource base, and provision of adequate con- sumption, family farms are the most viable units for achieving those goals. The model is employed to simulate and evaluate a range of interventions undertaken within the existing agrarian structure. The effects are mixed and in general not very significant. Technological progress, for example, is the goal of several massive programs in Brazil supported bv international institutions. If it succeeds in raising crop yields in the Northeast by 25 percent, this will benefit consumers and small farmers but could cause a net reduction in employment and is likely to result in even less intensive land use on the larger farms. The subsidization of employment, simulated up to a rate of 20 percent of permanent laborers' wages, would promote employ- ment and raise production somewhat but would have no appreciable in- fluence either on overall employment rates or on the hiring practices and land use patterns of larger farms. Furthermore, intervention that affects the contractual basis of sharecroppers is likely to lead to their eviction, because landowners are close to being indifferent between sharecropping arrange- ments and other, less labor-intensive modes of operation. The influence of demand is investigated through simulations of a guaran- teed price support program (which does exist in the Northeast but is INTRODUCTION AND SUMMARY 29 ineffective because of institutional weaknesses, the remoteness of most producers, and a lack of storage and transport infrastructure) and of raising the general level of demand by 25 percent through roundabout means. Effective price guarantees would substantially increase the incomes of smaller farms (by 17 percent) but would have negligible effects on gross output and employment. This disappointing result reflects a lack of re- sponse from estates: Even though the model assumes that they optimize, the simulated price supports do not raise gross margins enough to affect the employment and land use practices on these farms. The promotion of demand by 25 percent, through marketing and urban incomes policies, would appreciably raise output (by about 8 percent) and the incomes of nontenured labor (by about 16 percent). This result suggests that market- oriented schemes might be profitably adopted in the Northeast, but the outcome would largely depend on whether the estates adopt economically optimal behavior. When development strategies based on combinations of the above in- terventions were simulated, the best that any of them could accomplish was a 20 percent increase in output, with gains to consumers through a 13 percent decline in the output price index or, alternatively, a 10 percent rise in output with a 14 percent increase in the general level of employment. None of these strategies is Pareto-optimal, and none provides a panacea for the Northeast. FEven if all the instruments were applied simultaneously, with output rising by 20 percent and consumer prices falling by 13 percent, total employment would increase by less than 2 percent. None of the strategies would have a substantial effect on poverty or on employment for the landless. The results force us to consider the fully legislated, adequately funded, but officially unsupported issue of land reform. Chapter 8 describes a simulation in which the land in the estate subsector is redistributed into module farms, a unit proposed in the Brazilian Land Statute of 1964 and defined in detail by the National Institute of Colonization and Agrarian Reform (INCRA). As simulated by the model, the modules are very attractive farming units. They are large enough to provide acceptable family incomes, to employ family labor fully, to satisfy family consumption needs, and also to take advantage of improving markets and advances in production tech- nology. The 37,000 existing estates, averaging 740 hectares each, could be con- verted into 790,000 modules.54 With such a reform there would be moderate 54. Those referred to as estates include all 3 3,000 farms larger than 500 hectares and an additional 4,000 medium-size farms that are similar to the large ones in their use of resources and degree of commercialization. 30 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL gains in production: Output of all commodities, including exportables, would increase by nearly as much as under any of the nonstructural policy interventions. More important, such a redistribution would ensure em- ployment and adequate consumption for more than twice the number of families the agricultural sector is currently accommodating. Of the policv and project interventions described and analyzed by this study, land reform appears to be the only intervention with any hope of alleviating the poverty of the landless rural workers and their families. A land reform may be a prerequisite for solving the problem of the Northeast. Such a course of action, however, would pose an immense political and administrative challenge without giving any guarantee that its results would be sufficient. The resource base is simply not large enough to provide adequate incomes for the 18 million rural Northeasterners. Mea- sures to help increase farm productivity (agricultural research, extension services, and more efficient and flexible marketing channels) need to be supplemented by interventions in other sectors. In particular, programs are needed to develop human resources and promote the emigration of ade- quately educated and trained Northeasterners, and the industrial sub- sidization scheme needs to be restructured to meet the employment goals it originally embraced. This "fresh" view using fresh numbers and fresh tools is not an optimistic one. The Northeast problem demands political action, recommendations for which lie outside the scope and expertise of the authors. Without such action, the analysis reveals little hope for solving the problem in a region where up to 10 million people remain near absolute poverty while agri- cultural land the size of France lies idle or greatly underused. Appendix: Zonal Delineation of the Northeast Although the Northeast region comprises nine states (plus a part of a tenth, Minas Gerais, which was not included in this study), there are such large physiographic, climatic, and agroeconomic variations even within states that state boundaries cannot be used as delimiters. Paraiba, for example, one of the smallest states in the Northeast, includes no less than three distinctly different ecological zones: the humid coastal Mata, the relatively fertile Agreste, and the semiarid, drought-prone Sertao. This ecological diversity causes differences in all aspects of agricultural activity. IBGE, the Brazilian statistical bureau, has defined homogeneous microre- gions in all states for use in presenting the results of the agricultural census. These microregions comprise clusters of municipios (counties) that have INTRODUCTION AND SUMMARY 31 similar ecological characteristics. The 1 14 such microregions defined for the entire Northeast, however, are too numerous for this kind of pre- sentation. We have adopted a zonal delineation proposed in a 1971 SUDENE studv that combined demographic, climatic, and agronomic factors to divide the Northeast into seven physiographic zones.55 With few exceptions, these zones are aggregates of the IBGE microregions and thus allow some compari- sons to be made with census results. The map on p. 20 shows the seven zones superimposed on the nine states. Some of the principal characteristics of each zone are shown in Table 1-1. Zone A: The West Zone A, comprising the western portions of the states of Maranhao, Piaui, and Bahia, stretches from the coast in the north to the border of Minas Gerais in central Brazil; its western border is a frontier area near the Amazon region. Population density averages less than three inhabitants a square kilometer; this is the most sparsely populated zone in the Northeast. Annual precipitation varies between 600 and 2,000 millimeters, evenly distributed over six to eight months. Drought is seldom a problem, but the zone's poor soils and lack of access to markets prohibit intensive agricultural activities. The zone has no paved roads and no market centers, although the ports of Belem and Sao Luis and the Federal District of Brasilia all lie about 200 kilometers from its borders. Three types of agricultural activity predominate. The swidden or slash- and-burn practices of nonpropertied farmers can be found scattered throughout the zone. After clearing the land, subsistence crops of rice, corn, beans, and manioc are planted for two or three years or until the nutrient value of the soil is depleted and the farmers move to another plot. The cleared land left behind is often taken over by cattlemen, usually nomadic, who use land-intensive methods. Purchased inputs for livestock in this zone average only about US$1.50 per animal a year. The third type of activity is extractive, involving wild carnauba and babacu palms. Zone B: The Middle-North The Middle-North zone-the northern parts of Maranhao and Piaui- shares many of the physical characteristics of the West, but because it contains two major cities, the port of Sao Luis and Teresina on the Parnaiba 55. Lacerda de Melo, "Espacos geograficos." Table I- 1. Principal Characteristics of the Physiographic Zones D B Semi- E F A Middle- C humid Humid Humid G Characteristics West North Sertao Southeast East Southeast Agreste Component states Maranhao Maranhao Ceara Bahia Rio Grande Bahia Rio Grande (parts) Piaui Piaui Rio Grande do Norte do Norte Bahia do Norte Paraiba Paraiba Paraiba Pernambuco Pernambuco Pernambuco Alagoas Alagoas Alagoas Sergipe Sergipe Sergipe Bahia Bahia Bahia Major city or port None Sao Luis Fortaleza None Recife Salvador None Teresina Salvador Rainfall (millimeters a year) 600-2,0N)0 600-2,000 400-600 800-1,600 1,000-2,000 1,20-2,000 600-1,000 Susceptibility to drought Negligible Moderate Severe Negligible Negligible Negligible Moderate Principal crops Rice Rice Cotton Coconut Bananas Cacao Cotton Beans Babacu Beans Beans Sugar Beans Manioc Beans Manioc Manioc Coconut Manioc Corn Manioc Corn Corn Beans Corn Corn Manioc Corn Percentage of Northeast's total Agricultural land 18.0 21.3 36.5 7.5 3.2 5.5 8.0 Cropoutput 3.5 6.3 28.1 5.5 4.2 39.8 12.5 Livestock output 9.9 7.5 41.1 10.7 4.5 0.3 25.9 Agricultural employment 11.5 7.2 41.8 8.0 3.8 7.5 20.3 INTRODUCTION AND SUMMARY 33 River, there are better opportunities for marketing agricultural products. As in zone A, slash-and-burn, extensive ranching, and the extraction of babacu and carnauba are important. The Parnaiba River, which stretches throughout the zone, allows the cultivation of irrigated rice, either on small plots adjacent to the river or with water carried to nearby plots surrounded by bunds. Modern irrigation techniques are little used in the Northeast except in the other major river valley, that of the Sao Francisco. Where modern irrigation has been attempted, salinity has usually rendered the soil useless after a few years. The eastern parts of the zone, bordering on Ceara, are more akin to the Sertao. Here scattered patches of cotton are often interplanted with various subsistence crops. These areas of zone B are quite susceptible to drought, unlike areas near the river and the coast. Zone C: The Semiarid Sertao With the possible exception of the coastal sugar zone, the Sertao is the most famous and interesting part of the Northeast from historical, sociolog- ical, and folkloric as well as economic points of view. This vast zone begins about 100 kilometers inland from the east coast and stretches almost to the pre-Amazonic frontier. North to south, its boundaries are the sea and the Center-South state of Minas Gerais. The overriding characteristic of the Sertao is its susceptibility to drought. Over the past century, major droughts have occurred every eight to ten years, causing almost total crop failure, decimation of cattle, and the starvation of thousands. Annual rainfall is usually between 400 and 600 millimeters but is irregularly and unpredictably distributed over four to six months. Under such circumstances, the water hole and adequate stocks of food or money are vital to sustain life. Massive emigration to both the coastal centers and the Center-South takes place. More often than not, the opportunities for migrants there are so limited and the urban slum condi- tions so squalid that the news of "rain in the Sertao" sends them scurrying back. The people, the crop varieties, and the farming practices have evolved to cope with the harsh conditions. The Sertao farmer or sharecropper who is able to maintain a family has been called "a hero and a genius. 6I With little more than a hoe and leftover seeds from the previous year, he produces subsistence crops (corn, beans, and manioc) for his family and a cotton crop to purchase minimum necessities. Over perhaps centuries, he has de- 56. From conversations with a Brazilian agricultural economist (unnamed). 34 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL veloped interplantings of crops that unquestionably increase his chance of survival. 57 Zone D: The Semibumid Southeast The Semihumid Southeast, wholly within the state of Bahia, lies be- tween the Bahian Sertao and the fertile cacao-producing coastal areas. Its soils and climate are similar to those of the Agreste (described below). Annual rainfall varies between 800 and 1,600 millimeters, and its regularity permits the cultivation of cash crops such as cacao, coffee, and tobacco as well as corn, beans, and manioc. Livestock ranching is the dominant activity. Zone E: The Humid East The Humid East, a narrow strip of eastern coast running from Bahia in the south to the northeasternmost part of Rio Grande do Norte in the north, is part of what is commonly called the Zona da Mata (forest zone), although the forest was cut four centuries ago to allow intensive cultivation of sugar. Rainfall is almost always greater than 1,200 millimeters a year, and (Iroughts do not affect the zone. With the major ports of Recife and Salvador, as well as four other state capitals, the access to urban and export markets is excellent. The history of this zone is the history of sugar. The intensive cultivation of sugar over the centuries, usually to the exclusion of all other crops, has depleted the once excellent soils. As a result, sugar as well as many other crops require fertilization. The yields of crops other than sugar are quite good in this zone and better than in most other areas of the Northeast. In addition to the usual Northeastern food crops, coconut is planted along the coast. Zone F: The Humid Southeast The Humid Southeast is the smallest zone in the Northeast and is the southern extension of the Humid East, which it resembles in its rainfall and soils. Its slightly cooler climate allows cacao production, by far the most important activity in the zone. This zone is quite small and has relatively few farms; land and incomes are more evenly distributed than in the other 57. A background study employed mathematical programming models to simulate farmers' behavior in the Sertao. It found that no better risk-reducing farm plans could be devised for their circumstances than those the farmers were already using. INTRODUCTION AND SUMMARY 35 zones; the productivity of the land is the highest in the Northeast; and our survey accordingly covered only one municipio in the zone. Zone G: The Agreste The Agreste is a wide belt bounded on the west by the Sertao, on the east by the Zona da Mata, and on the south by the Humid Southeast. Rainfall between 600 and 1,000 millimeters a year, fairly good soils, and low likelihood of drought permit a wide range of agricultural activities, and mixed farming predominates. Although the zone has no major cities, much of it lies within 200 kilometers of the coast, and thus access to markets is good. This locational advantage and the relatively high productivity of several varieties of planted pasture make livestock a productive enterprise, highly competitive with the wide range of crops that can be grown in this zone. Apart from the humid zones, many observers view the Agreste as having the highest development potential: It suffers neither the locational dis- advantages of the interior and western zones nor the drought susceptibility of the Sertao, and there is evidence that improved crop varieties would thrive, given the right combination of modern inputs. The seven-zone delineation is used for the cross-sectional analysis in Chapters 2 through 4, but it is sometimes necessary to omit specific reference to zone F because of the limited size of the sample for that zone. The seven zones are aggregated to five in the second part of the study, where a programming model is used for the analysis, so as to keep the model to a manageable size. Because the West and Middle-North have similar production patterns and technology, these are combined into the "West," and the cacao-producing zone F is combined with the larger Semihumid zone D. 2 Labor, Land, and Capital Resources Most observers have concluded that the heart of the Northeast problem lies in the paucity of its resource base; consequentlv, manv of the policy interventions have been designed at least in part to augment that base. Before proceeding to any analysis of the region's problems, one must examine the nature, magnitudes, distributions, and uses of the primary factors of production in the rural Northeast-labor, land, and agricultural capital. The Agricultural Labor Force The 1970 demographic census found a total Northeast population of 28. 1 million, of whom 16.4 million, or 58 percent, were rural. The state-by-state totals are given in Table 2-1. Bv 1973, when the survev for the present study was undertaken, the Northeast population had passed 30 million, and its rural population, despite continued emigration, had risen to nearly 18 million. Various definitions of the agricultural labor force are available. The demographic census, which uses the term "economicallv active" in agricul- ture (including a presumably small proportion engaged in mining, hunting, and fishing), found 5.2 million persons, or about one-third of rural North- easterners, so engaged in 1970. The agricultural census of the same vear, which used more disaggregated definitions, found 7.8 million people "occu- pied" in agriculture, as reported in Table 2-2.' The agricultural census 1. The problems of reconciling different Brazilian data sources have plagued other re- searchers. Cavalcanti and Goodman, for example, found "vast discrepancies" between demo- graphic and sectoral censuses and opted for the former's "economicallv active population" as the most reliable labor force data. See David Goodman and Roberto Cavalcanti, Incentivos a industrializacao e desenvolvimento do A'ordeste (Rio de Janeiro: IPEA/INPES, 1974). 36 LABOR, LAND, AND CAPITAL RESOURCES 37 Table 2-1. Total and Rural Population and Number Economically Active in Agriculture, by State, 1970 Percentage of population Economically Total Rural in rural active in State population population areas agriculture, Rio Grande do Norte 1,550,244 812,976 52.4 240,955 Paraiba 2,382,617 1,380,461 57.9 437,937 Pernambuco 5,160,640 2,349,797 45.5 764,719 Ceara 4,361,603 2,581,510 59.2 749,090 Piaui 1,680,573 1,143,961 68.1 346,875 Maranhao 2,992,686 2,240,659 74.9 762,900 Alagoas 1,588,109 956,370 60.2 323,155 Sergipe 900,744 485,392 53.9 161,815 Bahia 7,493,470 4,407,987 58.8 1,437,364 Northeast total (average) 28,110,686 16,359,613 (58.2) 5,224,810 a. Includes mining, hunting, and fishing. Source: Instituto Brasileiro de Geografia e Estatistica, Censo demografico [by state] VIII: Recenseamento geral, 1970, vol. 1, nos. 5-10 (Rio de Janeiro, 1972). Table 2-2. Agricultural Labor Force, by State, 1970 Producers Total Propri- Sbare- Occu- pro- Occupied State etors Renter croppers pants ducers persons Rio Grande do Norte 62,927 17,479 2,400 21,571 104,377 315,377 Paraiba 108,801 33,063 5,569 22,563 169,996 613,176 Pemambuco 202,683 59,138 10,108 59,757 331,686 1,157,039 Ceara 159,135 21,238 27,672 37,910 245,955 1,081,501 Piaui 73,744 50,402 43,853 49,993 217,912 519,931 Maranhao 48,397 147,629 10,493 190,405 396,924 1,196,611 Alagoas 74,270 18,192 1,883 11,051 105,396 440,342 Sergipe 71,700 11,975 406 11,842 95,923 275,751 Bahia 455,017 18,750 7,903 62,384 544,054 2,214,914 Northeast total 1,256,674 377,866 110,287 467,476 2,212,303 7,814,648 Source: Instituto Brasileiro de Geografia e Estatistica, Dados preliminares gerais do censo agropecuario VIII: Recenseamento geral, 1970 (Rio de Janeiro, 1972). .38 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL divides "producers" into the categories of proprietors, renters, sharecrop- pers, and occupants, a better term for whom is "squatters." Together the producers recorded by this census numbered 2.2 million, or 28 percent of the occupied persons. The present study sought new definitions of agents in the agricultural labor force for several reasons. First, the number of proprietors recorded by the agricultural census does not coincide with the number of properties given by a 1972 cadastral survey undertaken by the National Institute of Colonization and Agrarian Reform (INCRA). This cadastral survey was the basis for the sUDENE/World Bank survey undertaken for the present study.2 It found only about 850,000 properties, or 400,000 fewer than the number of proprietors recorded by the census. The SUDENE/World Bank survey had as its basis the farm, defined as a property or collection of properties under the same administration, and found only 800,000 such entrepreneurial units, as some individuals farmed more than one property. Similarly, the SUDENE/World Bank survey found far fewer renters than did the agricultural census, 26,000 versus 378,000, and about 240,000 sharecroppers, compared with 110,000 found by the agricultural census. These discrepancies probably again reflect differences in the classifications used. Most sharecroppers (parceiros) in the Northeast tended to describe themselves as renters (arrendatarios) even though the rent they pay is a portion of the harvest, whereas the SUDENE/World Bank survey distin- guished such agents on the basis of the form of rental contract. The number of occupants or squatters given bv the agricultural census is (lifficult to verify from the suDENE/Vorld Bank survey. Because the survey was based on agricultural properties as recorded by INCRA, it took account of squatters only when the owner of the property was aware that they were occupying it. A supplementary subsample was taken of squatters on public land in Maranhao, but it is certain that the information obtained was by no rneans complete. The authors' observations indicate that a very large population of migratory slash-and-burn agriculturalists in the Northeast is not adequately captured by any census or survey, including the SUDENE/ World Bank survey. The implication from the agricultural census is that one-fifth of the producers neither hold a title nor have any contract, written or unwritten, to the land they are working. A more functional breakdown of the agricultural labor force, employed by the sUDENE/World Bank survey and this analysis, is as follows: 2. Preliminary data from the INCRA cadastral survey were made available to StIDENE in raw form in 1973 to provide the sampling base for the SUDENEAVorld Bank survey. LABOR, LAND, AND CAPITAL RESOURCES 39 -Owner-operator: the entrepreneur who owns part or all of the farm and resides thereon. -Administrator: a nonowner who performs the entrepreneurial function. -Permanent worker: a permanently contracted worker who may or may not perform entrepreneurial or managerial tasks. A morador is a permanent worker who receives, as part of his compensation, a dwell- ing on the farm and a plot for family cultivation of subsistence crops; a nonmorador is contracted annually but receives no housing or land. -Sharecropper: an agent who pays a previously agreed-upon share of the produce in exchange for the use of a plot of land; he may or may not receive compensation for work performed off the sharecropped plot. -Family worker: a member of the entrepreneur's or sharecropper's im- mediate family who is economically active on the farm. -Temporary worker: a worker contracted and paid on a short-term (daily, weekly, or monthly) basis, usually to assist in planting and harvesting crops. -Renter: an agent who pays a fixed fee in money or produce for the use of a plot of land. -Squatter: an occupant of private or public land who makes no com- pensation to the owner, often using the land without the owner's knowledge. The number of these agents estimated by the SuDENE/World Bank survey are reported in Table 2-3 for the different physiographic zones of the Northeast. The Northeast total of 5.7 million in 1973 is nearer to the demographic census total of 5.2 million economically active persons than it is to the 7.8 million occupied persons found in the agricultural census. For the rest of the analysis we shall take 6 million as the approximate size of the agricultural labor force. The average farm household in the Northeast has five persons (adults and children), of whom 2.2 man-equivalents are members of the labor force. (Casual observation suggests that members of the extended family, includ- ing parents, brothers, and more distant relatives, often depend on a single farm even if it is very small.) Table 2-3 shows that of the 6 million in the labor force less than 800,000, or 13 percent, own the land they operate; this percentage rises to 30 when their family workers are included. Nearly half the labor force, or 2.9 million workers, rely on temporary employment with no continuous access to land.3 3. See Chapter 1, note 50. Table 2-3. Agricultural Labor Force, by Zone, 1973 Owner- Hired Family Share- Permanent Temporary Zone Zone operators' administrators Renters workers croppers workers" workers total West 88,213 1,271 621 85,905 7,159 36,668 375,833 595,670 Middle-North 72,533 3,204 4,215 57,560 56,045 84,825 387,773 666,155 Sertao 321,420 3,312 3,179 438,388 172,108 159,119 1,149,551 2,247,077 Semihumid Southeast 55,991 0 131 55,991 1,680 38,634 169,653 322,080 Ilumid East 29,771 1,130 1,510 23,176 927 46,970 109,390 212,874 Humid Southeast 57,381 0 0 29,838 0 384,453 215,179 686,851 Agreste 139,639 3,654 16,276 237,866 2,860 144,726 465,702 1,010,729 Northeast total 764,948 12,571 25,932 928,724 240,785 895,395 2,873,081 5,741,436 a. Includes squatters on public lands. b. Includes both moradores and nonmoradores. Source: SUDENF/WOrld Bank survey (extrapolated results). LABOR, LAND, AND CAPITAL RESOURCES 41 Table 2-4 reports the actual use of labor from the various categories of workers by zone. The Northeast total of 1.77 million man-years of labor use implies un- and underemployment of up to 70 percent. Because the SUDENE/World Bank farm survey did not capture employment in nonagri- cultural sectors (nor does any other data source permit estimates of such employment for rural residents), nothing conclusive can be said of un- and underemployment rates. Casual observation suggests that nonfarm em- ployment opportunities are quite limited, a view supported by expenditure survey data in the appendix to Chapter 3 that indicate 93 percent of rural households depend mainly on agriculture. That there is substantial under- employment in agriculture is beyond question. It is partly due to the seasonal nature of cropping in the Northeast and is strongly associated with the degree of access to land. Farmers and their family workers, as well as permanently employed workers and sharecroppers, work two to three times as many days a year as do the temporary workers. Temporary workers average only sixty days of agricultural employment annually; thus, if not productively employed in other sectors, they are clearly the most impoverished group. Nonetheless, all types of agents in all zones appear to be underemployed. Although Table 24 includes only labor directly related to cropping, stock raising, and farm maintenance, there may be up to 4 million man-years of labor available for productive employment in the rural Northeast. This reservoir represents an untapped resource of potentially immense dimen- sions. Land: Its Distribution and Quality At the risk of stating the obvious, it must be emphasized that land- ownership alone does not guarantee an adequate income. Many factors determine whether a property is, or can be, economically viable: its size, soil quality, and access to markets; the attitudes, skills, and motivation of the proprietor; and the climate (in particular, sufficient and timely rainfall). Tables 2-5 through 2-7, summarized in Table 2-8, reveal that farm sizes vary dramatically throughout the Northeast but that in all zones land is concentrated in large farms. About 33,000 farms (4 percent) contain 49 percent of the agricultural land and are larger than 500 hectares; 32 percent of all farms have an average size of five hectares, on 1.4 percent of all agricultural land. Gini coefficients of concentration, reported in Table 2-8, range from 0.75 to 0.89 for the various zones and average 0.83 for the entire Table 2-4. Intensity of Agricultural Employment, by Zone Average days worked a year Man-years Total worked man-years Family Permanent Share- Temporary Total per per average worked Zone workers workers croppers workers farm farm per zone West 152 183 113 46 494 1.94 173,599 Middle-North 138 118 69 43 368 2.15 162,835 Sertao 153 171 119 53 496 2.31 754,751 Semihumid Southeast 162 222 100 64 548 2.25 125,979 Humid East 130 180 100 68 468 2.51 77,562 Humid Southeast 109 234 - 162 505 7.81 96,591 Agreste 126 211 150 59 546 2.63 376,861 Northeast total 1,768,178 - Not applicable. Note: The typical working month in the Northeast is twenty-five days; the typical working year, 300 days. Source: SUDENE/World Bank survey. Table 2-5. Estimated Size Distribution of Farms, by Zone (number of farms) Farn size (hectares) Zone Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500 + total West 20,218 30,708 12,290 10,534 8,460 7,274 89,484 zt Middle-North 10,791 19,666 11,584 16,217 10,198 7,281 75,737 Sertao 94,174 122,521 48,081 30,412 20,594 10,950 326,732 Semihumid Southeast 9,286 27,034 9,621 4,565 2,895 2,590 55,991 Humid East 15,166 10,714 2,321 1,048 956 696 30,901 Humid Southeast 14,637 22,610 7,056 7,870 3,460 1,748 57,381 Agreste 84,665 39,334 8,188 5,394 3,358 2,354 143,293 Size class total 248,937 272,587 99,141 76,040 49,921 32,893 779,519 Source: SUDENE/World Bank survey. 44 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 2-6. Average Size of Sampled Farms, by Farm Size and Zone (hectares) Farm size (hectares) Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500 + West 3.70 25.45 71.94 138.90 313.22 1,178.03 Middle-North 3.69 26.57 68.80 138.93 317.20 1,395.89 Sertao 4.88 27.51 72.37 143.27 288.11 1,059.15 Semihumid Southeast 4.75 24.20 71.84 138.42 282.47 1,210.60 Hlumid Fast 3.71 26.11 72.82 143.62 283.51 2,303.59 Hlumid Southeast 5.60 27.09 70.63 142.29 294.60 620.00 Agreste 4.01 26.52 73.36 143.40 299.12 1,135.33 Source: SUDENE/World Bank survey, based on INCRA cadastral survey. Northeast., Concentration is highest in the Zona da Mlata where, histori- cally, the usinas dominated all economic activity and were sufficiently wealthy to absorb small farms. The cacao zone appears to have the most even land distribution, but another study argues that here the land distribu- tion is much more concentrated than it appears because of the prevalence of multiple-property holdings.5 Property size, of course, is only one dimension of the potential value of a farm as a productive asset. If small farms had a higher percentage of good soils, better access to water, or more capital invested in their land (for example, through clearing and leveling, prior fertilization to build up soil quality, irrigation and drainage, or stocks of tree crops), the distribution of productive potential would be less skewed. Bv the same token, if the land concentrated in large holdings were composed of poorer soils and located in areas far removed from market opportunities, the skewness in the distribu- 4. An anonymous reviewer has pointed out the likelihood that the INCRA cadastral survey of properties was biased toward a more equal land distribution because large landowners may have attempted to disguise the size of their holdings by subdividing their properties among family members and friends. This observation is consistent with popular accounts of one family's owning the western half of Pernambuco and two families' controlling the Paraiba sugar zone. See Joseph A. Page, The Revolution that Never WVas: Nortbeast Brazil, 1955-1964 (New York: Grossman Publishers, 1972), pp. 3-4, 224. 5. Gervasio Castro de Rezende, "Plantation Systems, Land Tenure, and Labor Supplv: An Hlistorical Analvsis of the Brazilian Case with a Contemporary Study of the Cacao Regions of Bahia, Brazil" (Ph.D. dissertation, University of Wisconsin, Madison, 1976), pp. 248-51. Table 2-7. 'Total Area in Farms, by Farm Size and Zone (thousands of hectares) Farm size (kectares) Zone Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500 + total West 74.8 781.5 884.1 1,463.2 2,649.8 8,569.0 14,422.4 Middle-North 39.8 522.5 797.0 2,253.0 3,234.8 10,163.5 17,010.6 Sertao 460.5 3,371.8 3,475.8 4,354.4 5,933.3 11,597.7 29,193.5 Semihumid Southeast 44.1 654.2 691.2 631.9 817.8 3,135.5 5,975.0 Humid East 56.3 279.7 169.0 150.5 271.0 1,603.3 2,529.8 Humid Southeast 82.0 612.5 498.4 1,119.8 1,019.3 1,083.8 4,415.8 Agreste 339.5 1,043.1 600.7 773.5 1,004.4 2,672.6 6,433.8 Total 1,097.0 7,265.6 7,116.2 10,746.3 14,930.4 38,825.4 79,980.9 Source: sUDENE/World Bank survey, based on INCRA cadastral survey. Table 2-8. Distribution of Farms and Farm Area, by Size and Zone (percent) Middle- Semibumid Hlumid Humid Farm West North Sertao Southeast East Southeast Agreste Northeast size -_-- (hectares) Farms Area Farms Area Farms Area Farms Area Farms Area Farms Area Farms Area Farms Area 0-9.9 22.6 0.6 14.2 0.2 29.0 1.6 16.6 0.7 49.1 2.2 25.5 1.9 59.1 5.3 32.0 1.4 10-49.9 34.3 5.4 26.0 3.1 37.7 11.6 48.3 10.9 34.7 11.1 39.4 13.9 27.4 16.2 35.1 9.1 50-99.9 13.7 6.1 15.3 4.7 14.2 11.4 17.2 11.6 7.5 6.7 12.4 11.3 5.7 9.4 12.5 8.7 100-199.9 11.8 10.1 21.4 13.2 9.4 15.0 8.1 10.6 3.4 5.9 13.7 25.3 3.8 12.0 9.8 13.5 200-499.9 9.5 18.4 13.5 19.0 6.3 20.4 5.2 13.7 3.1 10.7 6.0 23.1 2.4 15.6 6.4 18.7 500 + 8.1 59.4 9.6 59.8 3.4 40.0 4.6 52.5 2.2 63.4 3.0 24.5 1.6 41.5 4.2 48.6 Gini coefficient 0.82 0.84 0.80 0.81 0.89 0.75 0.85 0.83 Source: SUDENE/Worid Bank survey, based on 1972 INCRA cadastral survey. LABOR, LAND, AND CAPITAL RESOURCES 47 tion of land would only compensate for these factors.6 Furthermore, ques- tions of performance differentials, taken up in Chapter 4, would have no meaning. Therefore, we take up the issues of land value, location, and quality at this juncture. We employ several different approaches because several dimensions of productive potential are involved, and some of the data are subjective and possibly prone to bias. If such data were completely reliable, land values would provide the most meaningful comparators because they would reflect differences in locational advantage, soil qualities, and capital embodied in the land. In the rural Northeast, however, land sales and rental markets are probably the least perfect of all markets, and because land taxes-notwithstanding the insignificance of their rates-depend on land value, the available land value data must be cautiously analyzed. Three sources (Getulio Vargas Founda- tion, INCRA, and the sUDENE/World Bank survev) of land value data are available, but only two will be used here. (The most up-to-date INCRA land valuations were not available; however, this source is probably the most bias-prone because tax rates and land reform plans depend on them.) In a related World Bank studv of land distribution by state (as opposed to distribution by zone, as in this study), Gini and Pareto coefficients of landownership concentration were calculated for both physical area alone and physical area adjusted by a land value index based on reported sales prices and rental values.' For seven of the nine states, the distribution adjusted for land value was even more skewed than the distribution of physical area. For Sergipe, one of the smallest states, the skewness was the same before and after adjustment for land value. Only for Piaui, the poorest and least productive state, did the adjustment improve the distribution of land. The SUDENE/World Bank survey asked farmers to give a subjective estimate of the value of their land, not including the replacement value of the structures, as a proxy for its productive potential. Table 2-9 reports the average land values per hectare, by farm size class and zone, based on the valuations by farmers. These averages vary wildly throughout the North- east, ranging from a low of Cr$77 (US$12) per hectare for the largest farms 6. Schuh, for example, suggests that "large farms may be large of necessity because the soil productivity is low" and says that "many of the large land holdings are relatively isolated from the mainstream of economic life, and are being held for purely speculative purposes." G. Edward Schuh, TbeAgricultural Development of Brazil (New York: Praeger, 1970), pp. 415-16. 7. World Bank, "Rural Development Issues and Options in Northeast Brazil," report no. 665a-BR (Washington, D.C., June 1975; restricted circulation), pp. 8-9. The index emploved was constructed by the Getulio Vargas Foundation and was based on reported sales prices and rental rates. Table 2-9. Average Land Values, by Farm Size and 7one (cruzeiros per hectare) Farn size (hectares) Zone Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500 + average' West 1,348 192 218 196 227 93 142 Middle-North 821 189 142 113 113 77 97 Sertao 785 359 330 723 205 216 329 4- Semihumid Southeast 900 1,532 955 928 786 950 989 Ilumid East 2,273 1,282 784 886 1,037 2,317 1,878 Humid Southeast 2,993 6,168 9,428 8,057 8,098 10,824 8,544 Agreste 1,453 1,016 1,039 947 972 653 875 Size class average' 1,266 1,054 1,064 1,318 785 625 838 Analysis of variance from comparisons of degrees of freedom: df1 = 5, df2 = 36, F = 0.05 class means J 05(5, 36)= 2.5 a. Weighted by total area of farms in each size class. Source: SUDENE/World Bank survey. LABOR, LAND, AND CAPITAL RESOURCES 49 in the Middle-North to a high of Cr$ 10,824 (US$1,600) per hectare for the cacao estates in the Humid Southeast. Such variation may be partly due to the unreliability of subjective (and possibly bias-prone) valuations, but it undoubtedly reflects the diversity of the ecological base within the rural Northeast. In the three westernmost zones and the Agreste, there appears to be a significant downward trend in land value with respect to farm size. For the Northeast as a whole, such a trend does not appear until farm size exceeds 200 hectares. This fact raises the suspicion that owners of the largest farms may indeed have attempted to conceal the value of their farms (a suspicion supported by the location and land quality tests that follow). Land values, however, are quite stable throughout the size groups in the Semihumid Southeast, show a strong tendency to increase with farm size in the Humid Southeast, and exhibit a U-shaped pattern in the Humid East. A possible explanation for this pattern is that in the West and Middle-North larger farms may have lower-valued land because many of them have absorbed the smaller slash-and-burn farms for grazing purposes after the soils have been depleted. Similarly, the larger farms in the Sertao, and to a lesser extent the Agreste, are predominantly pasture, possibly because their soils are less fertile. The U-shaped pattern in the I-lumid East may partly reflect the proximity of the largest and smallest farms to urban centers. The largest farms, which are invariably sugar plantations, have relatively high levels of investment, and many small farms have very fertile land used to produce vegetables for the coastal urban markets. In the Humid Southeast, the higher land values of large farms may reflect the denser cacao plantings on these farms, whose more ready access to credit makes it easier for them than for small farms to meet the investment costs of cacao planting. Although such interfarm variations in average land value are important in some parts of the Northeast, they are not significant for the Northeast as a whole. The F-test on size-group mean values reveals that there are no statistically significant differences in these means (at the 95 percent confi- dence level). This result is due partially to the wide variation within size groups (Table 2-9) but mostly to the overwhelming weight of the variation across zones. The analysis-of-variance test implies that the interzone varia- tion is about twenty times as great as the intrazone variation.' The average values for the zones decline markedly as the distance from the coast increases, reflecting not only the locational advantages of the 8. In layman's terms, if the value per hectare were to be estimated for a farm picked at random in the Northeast, knowledge of the zone in which the farm was located would be far more useful than knowledge of the farm's size. 50 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Humid East and Humid Southeast over the West and IMiddle-North but also the investment in perennial crops in the former zones. In the Humid Southeast, with its large plantings of cacao trees, land values are markedly higher in all farm size classes. Climate and the probability of drought also appear to influence land values heavily. Large parts of the Sertao are no less favorably situated than the Semihumid Southeast or the Agreste, but land values there are roughly one-third of those in the latter two zones, which are less prone to drought. Regardless of these latter factors, it is clear that the SUDENE/World Bank survey did capture the locational differences in land value, and these differences are far more marked between zones than among farm size groups within a given zone. The location hypothesis can be addressed at the more micro level because the suDENE/World Bank survey tabulated the distance of the sampled farms from the nearest municipio seat. In the rural Northeast, the municipio seat is usually the largest nearby town and constitutes the market for the farms as well as the transportation hub for goods not consumed locally. Table 2-10 reports the average distances in kilometers from the sampled farms to the municipio seats, by zone and farm size group. As was the case with land value, no clear pattern emerges. Farms in all size groups and all zones are quite widely dispersed. In most zones, the medium-size farms are, on the average, farthest from the municipio seat, rather than the largest farms, as might have been expected. Also, the smallest farms appear to have a locational advantage in zones E and D, but this advantage is very small. When the smallest size class (0-10 hectares) was compared with the largest class (more than 500 hectares), it was found that a difference-of-means test was significant at the 95 percent confidence level only in zones A, B, and D. For the Northeast as a whole, there is no difference in the mean distances for the two extreme size groups, again at the 95 percent confidence level. The analysis-of-variance test reported in Table 2-10 also reveals no statisti- cally significant locational differences across all size groups within given zones. Variations in land quality were investigated through a study of the fertility and topography of a special subsample of about 2,200 farms scattered in the seven zones. These two dimensions of land quality are particularly important in Northeast Brazil because the chemical composi- tion of the soil varies greatly and because many farms are narrow strips of land stretching from a river valley to the top of an altiplano. Farmers were asked to classify their land as high, medium, or low yield. Their classifica- tions were cross-checked by the interviewer (generally an agronomist) with LABOR, LAND, AND CAPITAL RESOURCES 51 Table 2- 10. Average Distance from Sampled Farms to Municipio Seat (kilometers; standard deviations in parentheses) Farm size (hectares) Zone 0-10 10-50 5(-100 100-200 200-500 500 + West 49.1 44.5 62.6 54.1 56.6 60.9 (51.9) (53.3) (100.5) (53.4) (78.3) (55.0) Middle-North 24.1 22.3 40.2 34.8 38.6 29.6 (30.8) (23.1) (54.6) (55.1) (53.6) (18.1) Sertao 22.5 21.5 22.8 22.8 24.2 22.8 (54.3) (19.0) (22.8) (17.3) (33.6) (16.5) Semihumid Southeast 31.0 37.1 42.1 36.1 34.4 34.9 (29.3) (23.4) (21.2) (22.7) (21.5) (23.8) Humid East 10.0 18.8 31.0 17.2 18.2 11.0 (10.4) (14.2) (122.4) (11.8) (18.5) (8.1) Hlumid Southeast 5.5 9.0 13.1 13.1 15.5 13.0 (5.0) (7.0) (10.4) (5.9) (8.2) (0.) Agreste 16.1 24.2 21.1 21.1 13.4 14.4 (65.8) (83.1) (57.3) (36.3) (11.7) (14.1) Analysis of variance from comparisons of degrees of freedom: df=5, df236,F =0.04 group means F005(5, 36) = 2.5 Source: SUDENE/World Bank survey. further inquiries on the yields of the three types of land crop by crop.9 Farmers then classified the topography of their land into three groups- lowland, hillside, and highland. Table 2-11 reports the results of statistical tests on the land quality subsample. The first test involves a log-linear regression of the quantity of the most productive land qualities against total farm size. '0 Apart from the Humid Southeast (cacao region), none of the elasticities is significantly different from one at the 95 percent confidence level. In other words, the proportion of a farm's land that is composed of either high- or medium- 9. Some additional control is provided by the data on yields. In fact, yields varied less within groups of homogeneous soil quality than they did between groups of different soil qualities. 10. The remaining land qualities were omitted from the tests because they are generally unsuitable for crop cultivation. 52 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 2-11. 7ests for Independence between Land Quality and Farm Size, by Zone Log-linear relation between quantity of land Chi-square (x2) test for e quality andfarm sizea independence between proportion of land Zone Elasticity R2 by quality andfarm sizeb West High-vield land 1.027 0.991 9.73 Medium-yield land 0.975 0.998 1.53 'liddle-North Hligh-yield land 0.954 0.998 4.86 Medium-yield land 1.217 0.587 0.69 Sertao High-yield land 0.928 0.998 6.32 Medium-yield land 1.051 0.999 2.32 Semihumid Southeast I ligh-yield land 1.002 0.996 4.23 Medium-yield land 1.036 0.994 4.44 Humid East Hligh-yield land 1.074 0.998 7.84 Medium-yield land 0.957 0.994 6.68 l Iumid Southeast Hligh-yield land 0.817 0.892 41.47 Medium-vield land 1.467 0.939 16.86 Agreste High-yield land 0.973 0.998 1.70 Medium-yield land 1.031 0.997 3.09 a. An elasticitv greater (less) than one implies the proportion of land of a given quality expands faster (slower) than the size of farms in a given zone. None of the elasticities in the table are significantlv different from one, at the 95 percent confidence level, except for those pertaining to the Humid Southeast. b. The x2 is computed from unweighted, grouped data to test the hypothesis that the percentage of land of different qualitv is not significantly related to farm size. The 5 percent critical value of the X2 for five degrees of freedom is 11.07. Source: SUDENE/OVrld Bank survey. yielding soils is invariant with the size of the farm. In the Humid Southeast, larger farms appear to have lower proportions of high-yielding land but much larger proportions of medium-yielding land. This does not, however, adversely affect the value of these farms; as shown in Table 2-9, the average land value per hectare has a marked tendency to increase as farm size increases in this zone. The chi-square test reported in the far-right column of Table 2-11 supports the regression tests. Again, only in the Humid Southeast are the LABOR, LAND, AND CAPITAL RESOURCES 53 Table 2-12. Agricultural Capital, by Zone (thousands of cruzeiros) Total capital Zone Structures Animals Equipment Cruzeiros Percent West 995 1,649 79 2,723 8 Middle-North 930 1,212 133 2,275 7 Sertao 5,472 4,996 479 10,947 32 Semihumid Southeast 2,019 2,470 77 4,566 13 Humid Fast 966 497 76 1,559 5 Humid Southeast 5,596 194 85 5,875 17 Agreste 2,983 2,995 192 6,170 18 Northeast total 18,961 14,013 1,143 34,117 100 Note: By type of capital, percentages are: structures, 56 percent; animals, 41 percent; equipment, 3 percent. Source: SUDENE/World Bank survey. proportions of high- and medium-yielding land not independent of farm size. From these investigations into the value and quality of agricultural land and the distance from the sampled farms to the nearest markets, several conclusions can be made that will have important implications for later analysis. First, the skewness in the distribution of land throughout the entire Northeast is not mitigated bv the consideration of land value. Although there is a clear trend for average land value per hectare to decline as farm size increases in the Agreste and the interior zones, this is not true for the rest of the Northeast comprising the humid and semihumid zones of highest productive potential and best locational advantage. Second, for the Northeast as a whole, the zone in which a farm is located has far greater importance in determining its land value than does the size of the farm. Third, within sampled municipios, the smallest farms do not have a statisti- cally significant locational advantage over the larger farms. And, fourth, there is no evidence that larger farms outside the Semihumid Southeast have smaller percentages of high- and medium-yielding soils than do small and medium-size farms. Agricultural Capital Other forms of capital such as structures, livestock, and equipment are much easier to value than those embodied in land. Table 2-12 reports the 54 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 2-13. Average Agricultural Capitalper Farm, by Zone (cruzeiros) Zone Structures Animals Equipment Total West 11,123 18,432 886 30,441 Middle-North 12,292 15,999 1,759 30,050 Sertao 16,748 15,290 1,474 33,512 Semihumid Southeast 36,067 44,120 1,367 81,554 Humid East 24,765 12,730 2,469 39,964 Humid Southeast 97,519 3,376 1,480 102,375 Agreste 20,818 20,904 1,340 43,062 Northeast average' 24,324 17,976 1,466 43,766 a. Weighted by estimated farm distribution. Source: SUJDENEIXVorld Bank survey. total values of these items by zone, and Table 2-13 reports the average value per farm of the same items. Structures account for over half the value of capital items recorded by the survey. They include the houses of the proprietor and those of any perma- nent resident workers, storage buildings, livestock pens, and water storage and distribution facilities. Though the value of structures per farm appears to be correlated with productivity (being lowest in the poorest zones, the West and Middle-North, and highest in the more humid eastern zones), rigorous analysis of the influence of structures is difficult because it is not possible to distinguish items such as housing from directly productive assets. Animals account for about 40 percent of the value of the three capital items discussed here. Virtually all farms have cows, pigs, sheep, goats, chickens, and turkeys. Animals are treated as a capital item in the sense that farms invest in them to obtain either a flow of services (for example, draft power) or a flow of production (meat, dairy products, eggs, hides). The use of animals for draft power is relatively rare in the Northeast, but cattle are highly significant in generating income. Other animals are kept largely for home consumption. There is also widespread evidence that many animals are maintained as a store of value, particularly in the interior where finan- cial markets are primitive, but also perhaps as a hedge against runaway inflation. The authors' observations suggest that farmers rely heavily on the occasional marketing of a goat or a flock of hens for cash between harvests. Only 3 percent of the agricultural capital considered here is in equip- ment. This category covers virtually any nonfixed item directly or indi- LABOR, LAND, AND CAPITAL RESOURCES 55 rectly related to production or marketing, ranging from simple hoes" to tractors and implements, water pumps, scales, carts, jeeps, and trucks. The values of the equipment inventories reported in Tables 2-12 and 2-13 are extremely low. They range from Cr$886 per farm (US$138 in 1973 dollars) in the West to Cr$2,469 (US$386) in the sugar zone; in the other zones they are closely clustered about the Northeast mean figure of Cr$1,466. The distribution of equipment among farm size groups is shown in Table 2-14. Comparison of this table with the size distribution of farms in Table 2-8 gives another indication of the paucity of resources with which small farms have to work. In the Sertao, for example, 29 percent of the farms possess only 2 percent of the equipment; in the Agreste, 59 percent of the farms possess 6 percent of the equipment. If those in the relatively wealthy zones, the Humid Southeast and the Humid East, are omitted, farms under 10 hectares have an average equipment inventory of Cr$ 115 (US$18 in 1973 dollars). Inasmuch as a single hoe costs about one-third of this amount, the level of technology on these 249,000 farms is obviously very low. Casual observation suggests that, apart from the land distribution, the key factors constraining agricultural progress in the Northeast are the paucity of equipment and of structures directly related to production (particularly storage facilities and water storage and distribution facilities). This hypothesis may be crudely tested by estimating incremental capital/ output ratios (ICOR). Table 2-15 reports log-linear regressions of capital (defined as housing, other structures, and equipment) on gross output. The high elasticities observed in all zones but the Semihumid Southeast imply that substantial returns are possible from investment in Northeastern agriculture: The elasticity of annual gross output with respect to housing, other structures, and equipment is always greater than 0.4, and these types of capital have very long lives. Its importance to Northeast production is further substantiated in Table 2-16, which reports average capital/output ratios (ACOR). These are remarkably stable across zones and farm size groups, though land/labor ratios vary widely. (As shown in Chapter 4, these ratios vary by as much as a factor of 45 over size groups.) Clearly, farms in different size groups and in different zones freely substitute along the land/labor isoquant, but the relative constancy of the ACOR and their closeness to the ICOR imply rigidities in the markets for housing, other structures, and equipment. 11. The typical hoe in the Northeast is a combination hoe-ax-pick; it is heavy, durable, and employed for a variety of uses. For more than a few Northeastern agriculturalists, it is the only equipment available. I'able 2-14. Distribution of Equipment, by Farm Size and Zone (cruzeiros; percentages in parentheses) Farm size (hectares) Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500 + Total West 3,494 10,685 17,118 4,949 21,521 21,543 79,310 (4) (13) (22) (6) (27) (27) (100) Middle-North 1,231 17,461 26,318 19,168 25,896 43,175 133,249 (1) (13) (20) (14) (19) (32) (100) Sertao 9,047 83,60() 73,909 74,431 71,294 166,489 478,770 (2) (17) (15) (16) (15) (35) (100) Semihumid Southeast 540 24,628 12,920 7,837 6,085 24,528 76,538 (1) (32) (17) (10) (8) (32) (100) Hlumid East 20,767 12,263 6,907 7,117 20,462 8,782 76,298 (27) (16) (9) (9) (27) (16) (100) hlumid Southeast n.a. 5,810 9,551 25,612 42,336 1,588 84,897 n.a. (7) (11) (30) (50) (2) (100) Agreste 10,777 31,160 15,900 49,369 53,477 31,310 191,993 (6) (16) (8) (26) (28) (16) (100) n.a. Not available. LABOR, LAND, AND CAPITAL RESOURCES 57 Table 2-15. Regression Results: Incremental Capital/Output Ratios (IcoR) Elasticity of output with respect to Zone Constant value of capital ICOR R2 West 67.93 0.513 1.95 0.796 Middle-North -314.27 0.902 1.11 0.928 Sertao 268.29 0.525 1.90 0.997 Semihumid Southeast 471.37 0.274 3.65 0.962 Humid East 1,244.30 0.474 2.11 0.920 Agreste 126.50 0.403 2.48 0.997 Note: Table is based on grouped data. Dependent variable: annual gross output. Indepen- dent variable: capital = value of housing, other structures, and equipment. Table 2-16. CapitallOutput Ratios: Value of Housing, Other Structures, and Equipment Average capitalloutput ratios (ACOR) Semi- Farm humid Humid size Middle- Soutb- Humid Soutb- (hectares) West North Sertao east East east Agreste 0-9.9 1.81 1.72 1.85 1.81 1.96 0.44 1.26 10-49.9 1.95 1.40 1.62 2.62 1.84 0.85 1.64 50-99.9 1.67 1.80 1.63 1.47 1.52 1.14 1.41 100-199.9 1.87 1.87 1.73 3.37 1.62 0.74 2.28 200-499.9 2.61 1.68 1.66 3.50 1.32 0.81 2.23 500 + 1.52 1.11 1.95 3.51 2.21 0.54 2.30 ACOR, all farms 1.89 1.46 1.73 2.80 1.80 0.77 1.81 ICOR, all farms 1.95 1.11 1.90 3.65 2.11 0.59 2.48 3 Outputs, Inputs, and Incomes Before we analyze the way the principal factors-labor, land, and capital- are employed on different types of farms, it is necessary to discuss the ends of agricultural production: outputs and incomes. Because markets are primitive in much of the Northeast and many farmers and sharecroppers produce largely for subsistence, output and income are closely intercon- nected. Agricultural Products The size of the Northeast region and the diversity of its farming systems mean that no fewer than sixteen crops must be considered to obtain a representative picture of the region's crop output. The crops currently produced are described briefly below. Short-Cycle Crops Five principal short-cycle (annual) crops are cultivated in the Northeast: berbaceo cotton, rice, corn, manioc, and two varieties of beans, de arranca and de corda. Cotton is the principal cash crop in the Sertao and Agreste regions; the other four crops are grown throughout the Northeast and make up a large proportion of the typical diet. Rice is grown on both irrigated and rain-fed land, though the scarcitv of irrigation works largely confines irrigated rice production to the river valleys in the western regions. About 80 percent of the corn produced is a common maize varietv; several hybrid varieties make up the remainder. Virtually all farms, save the sugar and cacao plantations, produce at least two annual food crops, both for their own consumption and for the market. Manioc is grown very commonlv, partly because storage facilities on most farms are limited and manioc roots can be left in the ground for up to two years. 58 OUTPUTS, INPUTS, AND INCOMES 59 Long-Cycle Crops Agricultural resources in the Zona da Mata have been almost entirely devoted to sugar production since the sixteenth century. Sugar remains the largest source of foreign exchange for the rural Northeast, followed by cacao, which is produced mainly in the central region of Bahia (the Humid Southeast zone in this study). Whereas most of the agricultural technology in the Northeast is primitive, sugar and cacao production techniques rely to some extent on mechanized equipment and chemical fertilizers. Of the farms sampled by the SUDENE/World Bank survey, at least 30 percent of those in the Humid East and Humid Southeast used chemical fertilizers, whereas only 1 percent of those in the interior (West, Middle-North, and Sertao) did so. Cotton is the dominant long-cycle crop in the interior zones. The most common variety is the long-fiber moco, which is highly drought-resistant and has a life cycle of five to eight years. A hybrid between moco and the annual herbaceo is verdao, which has a life cycle of about three years. Although it is inferior to moco, verdao shares some of moco's drought- resistant qualities and, because it is more bushlike, leaves the land easier to clear. Tropical fruits, particularly bananas, oranges, and coconut, are also widely cultivated in the Northeast. Two types of long-cycle fibers, sisal and tucum, were once relatively important in the Northeast, but their production has declined rapidly in the face of competition from synthetic substitutes and instability in interna- tional markets. Extractive Activities Apart from forestry, which is not considered in this study, extractive activities from babacu and carnauba, two types of indigenous palm, are prevalent in the West and Middle-North. The babacu produces nuts, which are gathered in the slack agricultural season for a variety of uses and for their oil. The carnauba is of economic importance to the frontier areas and to the Sertao because of the high-quality wax derived from its leaves. Livestock Stock raising of some type is observed in all zones and on almost all farms, though few, except in the western zones, can be said to specialize in livestock. This study will not consider the small animals kept on most 60 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL farms; it is difficult to attribute particular inputs to their production, and they account for a negligible proportion of marketed output. Roughly 18 million head of cattle are maintained in the Northeast. Dairy products are best considered as by-products, except in a few farms near major market centers. In the interior, because of poor transport and a dearth of refrigera- tion facilities, organized dairy production for the market is almost nonexistent. By far the most important output from stock raising, therefore, is beef. The singular advantage of beef cattle appears to be that they can transport themselves (often 200 miles or more) to market and, unlike crops, their sales can be timed to yield a steadier flow of income. The quality of beef produced, however, is extremely poor. Little or no effort is made to preserve breeds, and the resulting mixture of bloodlines, combined with primitive technology, yields very low carcass weights. Agricultural Output Official Brazilian figures estimate the gross output from crops in the Northeast in 1973 at Cr$11,281 million, or US$1,763 million.' The crop yields reported by the sL-DENE/WIorld Bank survey suggest that this figure may be biased upward, even when allowance is made for the possibility that it takes less than full account of production consumed on the farm by either people or animals or of payments in kind to workers. Nevertheless, it is the best available starting point to analyze agricultural productivity in the Northeast. Data on the consumable output from livestock are not available from official sources,3 but the survey suggests that meat and dairy pro- duction for 1973 amounted to Cr$2,316 million or US$362 million, giving a total of Cr$13,597 million or US$2,125 million for all agricultural production. For the rural population of 18 million in 1973, gross output from agricul- ture thus amounted to US$125 per capita. Aside from the fact that farm 1. In every interior market that we have observed, all perishable products (meat and dairy) are sold on the same day they are brought to market. 2. Instituto Brasileiro de Geografia e Estatistica (IBGE), Anuario estatistico do Brasil (Rio de Janeiro, 1975). 3. IBGE reports animal population but not production. Production estimates from other sources, principally the Ministry of Agriculture, have been termed "highly suspect . . . unreliable and often inconsistent." William H. Nicholls, "The Brazilian Agricultural Economy: Recent Performance and Policy," in Brazil in the Sixties, ed. Riordan Roett (Nash- ville, Tenn.: Vanderbilt University Press, 1972), pp. 162-63. OUTPUTS, INPUTS, AND !NCOMES 61 profits and labor remuneration make up the bulk of value added in agricul- ture, the virtual absence of industry and the visibly meager tertiary activi- ties in the rural Northeast suggest that rural per capita incomes may not be much higher than this figure. If the balance of trade between the Northeast and the rest of Brazil and the world is zero,4 the gross output figures imply that per capita annual consumption of agricultural products in the North- east is at most US$66, if stocks and savings are ignored. The data that were given in Tables 2-3 and 2-4 imply that the output per capita of persons in the agricultural labor force is US$400. The output for each man-year of full-time employment is US$1,202. For all agricultural land-that is, land contained in properties registered by the INCRA cadastral survey-gross output is US$29 per hectare (1973 dollars). For land actually cultivated and harvested, gross output is US$183 per hectare.' As noted above, it is possible that the estimate of aggregate crop output may be too high, and hence that these vield figures may be overstated.6 Although the survey results are generally quite close to census figures for comparable aggregate items and for prices and cropped areas, there are some substantial differences for crop yields. Table 3-1 reports IBGE's estimates of yields with those from the survey for both good- and medium-quality cropland, for the principal crops that were reported in comparable units.7 The survey and the census are remarkably close for crops (cotton and sugar) that require processing and for which formal control over production statistics is thus possible. They differ dramatically for crops, particularly beans and manioc, that are primarily subsistence crops with the bulk of production never reaching a formal market or processing center. This 4. The Northeast exports sugar, cocoa, cotton, babacu, sisal, and carnauba, but informa- tion on its food imports is not available. 5. This may be broadly compared with a world average grain yield of about US$350 per hectare. 6. Gross crop output may have been overestimated if the procedures used in the census (on which the estimate is based) double- or triple-count the output of crops from land that is interplanted. A hectare interplanted with manioc and beans, for example, may have been recorded as one full hectare of manioc and one full hectare of beans and the size of the harvest then deduced on this basis. IBGE does not report the procedures it uses in such cases, but the yield data shortly to be discussed suggest that some overcounting does occur. Survey estimates (based on the INCRA land registration) indicate that 13.8 million hectares were harvested in 1973, nearly 2 million more than the 11.6 million reported by the Bank of the Northeast from IBGE sources. 7. Whereas I1GE reports yield data in kilograms per hectare, the survey employed the most common units of measurement used by the farmer. Most fruits and some vegetables were reported in hundreds of fruits, which cannot be strictly compared with the weight measures. 62 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 3-1. Average Yields of Principal Crops, 1973: Alternative Estimates (kilograms per hectare) sLDJ, Sertao 0.78 0.67 0.90 0.73 0.78 0.67 Semihumid Southeast 0.96 0.56 0.82 1.00 0.78 0.66 Elumid Fast 0.99 0.47 0.46 0.58 (1.56 0.35 Agreste 0.88 0.76 0.92 0.99 0.25 0.69 Note: Capital refers to the valie of housing, other structures, and equipment; labor is measured in man-days. a. The llumid Southeast is excluded because the test revealed no statistically significant relation betwccn farm size and the variables considered. Source: SUDENE/World Bank survey, regression estimates on grouped data. FARM SIZES, TYPES, AND PERFORMANCE 87 Chapter 2), suggests that large-scale farmers face severely limited capital markets or that they are not interested in using their land with maximum efficiency. Table 4-4 gives additional details of trends in capital and labor use as farm size increases. Though the aggregate value of land rises roughly in propor- tion to area, the value of all other capital items, including livestock-an important part of the capital stock on larger farms-shows a rather mixed pattern of response and usually rises less than proportionally with area. That the various indicators of farm performance decline as farm size increases is to be expected; that they decline so dramatically and so consis- tently, over both zones and variables, requires explanation. Before delving more deeply into the typology of farms in the Northeast and into questions of profit-maximizing behavior, we will briefly consider several possible hypotheses. The first and most obvious may be termed the ecological/locational hypothesis. In this view, the larger farms are operated less intensely, simply because they have poorer soil and are located in more remote areas with tenuous links to markets. This hypothesis is undoubtedly true for some individual farms, but the evidence presented thus far strongly indi- cates that it is not true in the aggregate. In Chapter 2 the various statistical tests showed no systematic deterioration in locational advantage or land quality over farm sizes in all zones. The variation in land value per hectare appears to be too erratic to draw any firm conclusions except to say that the interzone variation is far greater than the intrazone. Nonetheless, the relations in Table 4-2 show that these variations have a negligible effect on productivity. Furthermore, the variation in the elasticities over zones in Table 4-2 lend little support to the locational component of this hypothesis. Apart from the cacao zone, the fastest rate of decline in productivity with respect to farm size occurs in the East (elasticity of -0.971), which is closest to both the Northeast's urban markets and export channels. If anything, the data show that the farther inland a zone is located, the slower is the decline in productivity with respect to farm size (Agreste, -0.643; Sertao, -0.635; Middle-North, -0.600) until the trend reverses in the most remote zone (West, -0.890), as might be expected. A second explanation for lower productivity of large farms concerns implicit taxation of agriculture through distorted or chronically overvalued exchange rates. At least one student traces the origin of the Northeast problem to the mid-nineteenth-century system of exchange rates, which favored export earnings of coffee, produced mostly in southern Brazil, while virtually pricing out of the market the Northeast's export crops, sugar 88 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL and cotton.3 Most observers feel that the exchange rate has been overvalued throughout the postwar period and that this has been a conscious govern- ment policy designed to assist import substitution in the industrial sectors. The effect has clearly been to tax agriculture to the extent that the over- valuation has resulted in lower prices paid to farmers for exported crops. This explanation has some validity, but it is questionable whether all the performance differentials can be attributed to it. First, the "tax" affects all agricultural producers except the smallest, subsistence-oriented farms. Why, then, do farms in the range of 100 to 200 hectares produce on the average twice as much per hectare as do farms larger than 500 hectares when they face the same "tax" rate? Second, less than 25 percent of Brazilian agricultural production is exported. Whv has the estate sector not adapted to the large and growing domestic market to effectively avoid the "tax"? A second explanation related to macroeconomic policies is inflation, which has often exceeded 50 percent a year in the postwar period. Inflation adversely affects agricultural productivity in at least two ways.4 First, the ex post results of production decisions are often quite different from the ex ante expectations because inflation rates are difficult to predict and different products are affected to different extents. This serves to diminish incentives to utilize resources at their full capacity and to invest in new agricultural capital. Second, inflation increases the value of land as a speculative asset relative to its value as a factor of production. When inflation leads to disorganized capital markets, land is one of the few assets available as a hedge against losses in purchasing power. As was the case with the ex- change rate "tax," inflation is likely to affect small, subsistence-oriented producers to a far lesser extent than larger, potentially commercial opera- tioins. Finally, the social unrest and periodic calls for land reform throughout much of this century have also contributed to uncertainty and to lower levels of productivity than would have occurred in a more stable political environment. The possibility of a land reform undoubtedly lessens the incentive of large landowners to engage in labor contracts with sharecrop- pers and permanent residents (moradores) because these agents would presumably be the beneficiaries of a reform. Such disincentives, therefore, make less labor-intensive activities such as stock raising more attractive and lead to a heavier reliance on temporary labor. 3. Nathaniel 1l. Leff, "Economic Development and Regional Inequalitv: Origins of the Brazilian Case," Quarterly Journal of Economics, vol. 86, no. 2 (May 1972), pp. 243-62. 4. G. Edward Schuh, The Agricultural Development ofBrazil (New York: Praeger, 1970), p. 346. FARM SIZES, TYPES, AND PERFORMANCE 89 Apart from the ecological/locational explanation, all the above- mentioned factors probably do have an adverse impact on the operations and productivity of the large-farm sector. It is questionable, however, whether they can explain, even if taken together, the sheer size of the performance differentials. When labor-to-land ratios vary by a factor of twenty or more, and output per hectare by a factor of 100 or more, the full answer must be sought in the functioning of the factor markets and in the behavioral responses of farms of different sizes and types to productive opportunities. Market Dependency: Family and Nonfamily Farms Extensive experimentation convinced us that there were no significant differences in the production functions of farms of different size classes. To further investigate the reasons for the differences in patterns of factor use on farms of different sizes, it is necessary to look at the organization of production. Theoretical studies, supported by most of the empirical litera- ture on Brazilian agriculture,, suggest a sharp dualism between a subsis- tence sector and a market-oriented sector, each of which has stylized characteristics: Subsistence sector -Labor is mainly supplied by the family. -Output is mainly consumed on the farm so that the marketable surplus is relatively small and not increasing. -There is little or no technical change. -The returns to capital and labor are low so that there is little incentive to save, to invest, or to emplov labor beyond the activities strictly required for subsistence. Market-oriented sector -Labor is mainly supplied by hired workers. -There is a high reliance on the market both to dispose of the marketable surplus and to procure inputs. 5. These tests are reported in SUDENE, A economnia agricola do Nordeste: Diagnostico parcial e perspectivas (Recife, 1976). 6. See the discussion on technological dualism in Brazilian agriculture in Ruy Miller Paiva, "Modernizacao e dualismo tecnologico na agricultura," Pesquisa eplanejamento economico, vol. 1, no. 2 (December 1971), pp. 171-234; W. 11. Nicholls, "Paiva e o dualismo tecnologico na agricultura: Utm comentario," in Pesquisa e planejamento economico, vol. 3, no. I (March 1973), pp. 15-50; G. E. Schuh, "Modernizacao e dualismo tecnologico na agricultura: Alguns comentarios," Pesquisa eplanejamento economico, vol. 3, no. I (March 1973), pp. 17-94. 90 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Figure 4-1. A Farm Classification Scheme Based on Market Dependency Percentage of labor hired 0 50 10() Commercial farm HI III c .7Vonfamily 50 Family Farm f -50 C' farm i l ~~~~~production farm 100 5 0 0 Percentage of familv labor Source: C. Nlakajima, "Subsistence and Commercial Farms: Some Theoretical Mo1dels of Subjective F-quilibrium," in Subsistence Agrzulture and Fconomic Development, ed. C. R. Whar- ton, Jr. (Chicago: Aldine, 1970). -Technical change, as reflected in the adoption of new techniques, is rapid. -The returns to capital and labor are substantial, and the supply of these factors is responsive to movements in their relative prices. Nakajima has suggested that all farms can be classified according to two continuous variables relating to the proportion of family labor in total labor use, and the proportion of production consumed on the farm.7 Figure 4-1 7. Chihiro Nakajima, "Subsistence and Commercial Farms: Some Theoretical Models of Subjective Equilibrium, " in Subsistence Agriculture and Economic Development, ed. C. R. Whar- ton, Jr. (Chicago: Aldine, 1970). FARM SIZES, TYPES, AND PERFORMANCE 91 Figure 4-2. Preliminary Classification of Sampled Northeastern Farms Percentage of labor hired 0 50 10( C G22 F4 1 F6 ._ SO * /J | 44 G3 BS ~~F3 50 14 16 D64. G *B6 e t *B~~~~~~~I D1 1 o ~~~~~~D2 D4 166 GI C2500 A2 *.4A * 12 I C ~~~~cl * A3B3 51) ~~~Al *BI 1)3 =5 DI 100 50 0 Percentage of family labor Note: Farms are denoted by zone (A, West, through G, Agreste) and by size class (I 0-9.9 hectares; 6 = 500+ hectares). represents this two-dimensional continuum. Two of the four quadrants in the figure (I and III) indicate, respectively, family subsistence farms and commercial enterprises that hire labor. Farms in the other two quadrants, II and IV, are more difficult to characterize, and their classification requires additional criteria. Figure 4-2 shows the mean values of the two Nakajima variables for the different farm size classes and zones of the Northeast. It shows, first, that the two elements of market dependency (proportion of produce sold and proportion of labor hired) are highly correlated. Exceptions include a few farms of small and medium size in the Zona da Mata, Semihumid South- east, and Agreste, which rely on family labor but market at least half their 92 TIIE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL produce, and all the farms over fiftv hectares in the West, which rely on hired labor but market less than half their output. Second, the more commercial farms tend to be large, whereas the subsistence farms tend to be small. Third, agriculture is noticeably more market-oriented closer to the coast. All farms in the Zona da Mata and the cacao zone are commercial: Except for those under ten hectares in the sugar zone, at least two-thirds of their labor is hired, and they market at least two-thirds of their output. As the distance from the coast increases, it is only progressively larger farms that are commercial. Fourth, except in the sugar and cacao zones, farms of medium size (50-200 hectares) are neither distinctly commercial nor dis- tinctly oriented to subsistence. More variables will be necessary if we are to characterize their activities. More systematic tests of the efficiency of family subsistence and non- family commercial farms are needed to establish whether the differences in the performance of farms of different sizes are associated with differences in the production functions of family and nonfamily farms; whether different factor intensities are caused by imperfections in the factor markets; and whether there is evidence of non-profit-maximizing behavior. For these tests, farms are classified into two subsamples according to the proportion of labor that is hired. The first subsample comprises all the farms that emplov more familv labor than wage labor (designated family farms), and the second comprises all the others (nonfamily farms). To test for differences in production functions, for each subsample and each zone we fitted a simple Cobb-Douglas production function using ordinary least squares,8 according to the following model: log Qi=a(, 0+ 8 aj log Xi+ ui, where Qi = total output for the ith farm (cruzeiros) Xi I = total value of farm land (cruzeiros) Yi2 = labor in man-years Xi3 = quantity of insecticide used (kilograms) Xi, = quantity of fertilizer used (kilograms) Xis - value of credit received in the last two years (cruzeiros) Xi6 = value of structures and permanent land improvements (cruzeiros) Xi7 = value of equipment (cruzeiros) S. The use of this type of function is justified in Scandizzo and Barbosa, "Substituicao e produtividade." FARM SIZES, TYPES, AND PERFORMANCE 93 Xi8 = value of livestock (cruzeiros) ui = error term.9 Table 4-5 reports the regression results. With few exceptions the estimated equations are similar for family and nonfamily farms. A simultaneous test for differences among the regression parameters (the Chow test) also yielded negative results for all zones except the Agreste.'° Furthermore, there are no significant differences in marginal rates of return to land and capital between the two types of farms. There do appear to be significant differences in the labor markets facing the two types of farms, as is shown by the comparison of marginal produc- tivities and wage rates reported in Table 4-6.1" If their objectives were simply to maximize profits, the family farms, where workers' marginal products are much smaller than their wages, would employ fewer workers than they actually do, and the nonfamily farms would employ more. If one interprets the marginal products of family laborers as their reservation wages, there appears to be a considerable incentive for them to seek employment on large farms, where wages are much higher. There are obviously many possible reasons the excess labor on small farms is not absorbed by the large farms through downward movements in the wage rate. People may prefer to work on their own farms, and there may be institutional barriers, such as legislated minimum wages or restric- tions to labor movement. But the widespread unemployment and the large seasonal variations in agricultural labor requirements suggest that another possible reason lies in the family workers' limited opportunities for finding outside employment. Table 4-7 shows the results of a test of this hypoth- esis. It compares the ratio of marginal products of labor on family farms to the wage rates on nonfamily farms with estimates of the probability of family workers' finding temporary employment off their own farms. 12 In all 9. Variables Xi, and Xi5 through Xj8 are measured as stocks and hence are representative of the productive capacity of the farm rather than of its input levels. For all these variables, however, a flow measure was either impossible by definition (as for structures or permanent land improvements) or impractical (as for livestock or equipment). 10. See G. S. Chow, "Tests of Equality between Sets of Coefficients in Two Linear Regressions," Econometrica, vol. 28, no. 3 (uly 1960), pp. 591-605. In all cases, except for the Sertao, the results of this test do not permit us to reject the hypothesis that the two sets of coefficients are equal at the 5 percent significance level. 11. For the theoretical basis of this type of comparison, see, for example, E. 0. Heady and J. L. Dillon, Agricultural Production Functions (Ames: Iowa State University Press, 1960). 12. SeeJ. R. Harris and M. P. Todaro, "Migration, Unemployment, and Development: A Two-Sector Analysis," American Economic Review, vol. 60, no. I (March 1960), pp. 126-42. 94 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 4-5. Cobb-Douglazs Production Functions, by Type of Farm and Zone (standard errors in parentheses) West Middle-North Sertao Non- NVon- Nl'on- Variable family Family family Family family Family Constant 4.873' 4.626' 4.620' 3.973 4.039a 3.610 (0.364) (0.292) (0.300) (0.363) (0.219) (0.166) Land value 0.096a -0.0018 0.1673 0.166' 0.I IV 0.065a (0.042) (0.041) (0.037) (0.043) (0.012) (0.033) Labor 0. 770a 0.566a 0.451' 0. 409 0.433a 0. 507a (0.081) (0.093) (0.066) (0.079) (0.039) (0.041) Pesticides 0.061 0.075 __d __d 0.082' 0.080 (0.063) (0.155) __ d __d (0.028) (0.033) Fertilizer __d _d _d _d _d _d _d _d _d _d _d _d Credit 0.0035 0.031 0.014 0.037 0.028a 0.048a (0.020) (0.03) (0.020) (0.030) (0.011) (0.013) Structures -0.0059 0.061 0.084b -0.0007 0.152- 0.127' (0.0036) (0.038) (0.034) (0.037) (0.029) (0.025) Equipment 0.071a 0.148' 0.034 0.077' 0.073 0.125' (0.024) (0.036) (0.022) (0.035) (0.015) (0.014) Livestock 0.064' 0.095a 0.040 0. 130a 0.077' 0.129a (0.27) (0.033) (0.026) (0.030) (0.016) (0.016) Coefficient of scale (I1d) 1.060 0.974 0.692 0.560 0.956 1.080 R2 0.609 0.428 0.369 0.355 0.553 0.535 Degrees of freedom 237 266 378 250 888 984 Note: Family farms are defined as having 50 percent or more of total farm employment supplied by family labor. The Humid Southeast is excluded because there were too fewv family farms in the sample. a. Significant at the less than 0.1 percent level. cases the two measures are indeed very close and, when sample variability is considered, not significantlv different from each other at any reasonable degree of confidence. Farm Sizes and Types: A Multiple Classification Though the distinction between family and nonfamily farms may be adequate to illustrate the mechanisms at work in the allocation of labor, other dimensions of economic behavior that vary with farm size may be FARM SIZES, TYPES, AND PERFORMANCE 95 Semibumid Southeast Humid East Agreste N.ortheast iVon- iVon- ivon- Nvon- family Family family Family family Family family Family 3.505" 1.758' 6.239" 4.039' 5.716' 4.420" 4.795" 4.073' (0.888) (0.651) (0.421) (0.541) (0.322) (0.258) (0.125) (0.115) 0.441a 0.367' 0.011 0.055 0.083a 0.061" 0.105' 0.087' (0.111) (0.064) (0.035) (0.057) (0.027) (0.027) (0.013) (0.013) 0.426a 0.492" 0.708' 0.437" 0.784' 0.497' 0.582" 0.510a (0.124) (0.173) (0.068) (0.141) (0.075) (0.078) (0.026) (0.031) 0.130" 0. 144" _d _d 0.078b 0.182" 0.060 0.043" (0.042) (0.077) __d _d (0.033) (0.047) (0.016) (0.024) __d _d 0.047' 0.091" _d __d 0.043a 0.062' d _d (0.023) (0.055) d _d (0.012) (0.019) 0.025 -0.110" 0.020 0.062 0.025b 0.089, 0.015 0.047' (0.022) (0.118) (0.025) (0.062) (0.016) (0.021) (0.007) (0.010) -0.071 0.055 -0.0012 0.0009 -0.005 0.034 0.090' 0.056' (0.139) (0.081) (0.049) (0.0055) (0.04) (0.034) (0.016) (0.015) 0.047 0.032 0.050" 0.063 0.043 0.021 0.057" 0.095" (0.032) (0.055) (0.028) (0.062) (0.024) (0.028) (0.0028) (0.011) -0.042 0.060 0.059" 0.124b 0.0012 0.088" 0.039 0.107a (0.047) (0.048) (0.029) (0.043) (0.026) (0.023) (0.010) (0.011) 0.957 1.050 0.894 0.833 1.009 0.972 0.991 1.007 0.509 0.446 0.646 0.364 0.489 0.436 0.528 0.435 132 132 184 109 327 361 2,225 2,146 b. Significant at the 0.1-1 percent level. c. Significant at the 1-10 percent level. d. Too few observations to estimate the coefficient. Source: SUDENE/Worid Bank survey. overlooked. To try to capture these, multiple comparison methods are used to estimate Cobb-Douglas production functions, wvith dummy variables representing: (1) five size classes, or strata, of farms (the first five used in the tabulations by farm size, excluding farms of 500 hectares and over); (2) the family or commercial nature of the enterprise (defined according to the proportion of hired labor in its total labor use, as specified in the previous section); (3) the presence of sharecropping; (4) the use of irrigation and of mechanical or animal traction equipment; and (5) the seven physiographic zones. These variables are used to test the hypothesis that farms of different sizes and modes of organization and in different zones may operate on 96 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 4-6. Comparison of Marginal Products of Labor with Wage Rates, by Type of Fann and Zone (geometric averages) Nonfamily farms Family farms t- Degrees of t- Degrees of Zone Al/W statistic freedom A1/W statistic freedom West 1.11 -2.33 345 0.55 -6.75 175 Middle-North 1.04 0.74 481 0.52 -8.89 183 Sertao 1.18 4.76 978 0.75 -6.72 633 Semihumid Southeast 0.97 -0.39 160 0.48 -4.66 86 Hlumid East 1.45 5.30 216 0.50 -4.87 71 H lumid Southeast 1.08 1.73 41 1.18 2.09 41 Agreste 1.46 6.98 378 0.54 - 7.92 218 Note: The hypothesis tested is that the marginal product of labor (.U) is significantly different from the wage rate (W4), or that AM1W is significantlv different from one. The M41W variable is assumed to be (asymptotically) lognormally distributed. more-or less-technologically advanced production functions. The na- ture of the test is such that only intercept levels, which indicate the level of technological efficiency abstracting from differences in the amounts of resources used, can be compared. As noted earlier, however, no significant differences in the production function coefficients could be detected by comparing the performance of farms of different size groups. The results of one series of tests are reported in Table 4-8, analogous to Table 4-5, which shows the effects of the scalar and dummy variables on output.'3 As can be seen from the second part of 'I'able 4-8, farm size generally has a large and significant effect on efficiency at a high confidence level, and, except in the Humid East, this relation is apt to be stronger as farm size increases. This implies that for a given quantity of resources used, smaller farms appear to obtain less output than larger farms (and family farms obtain less than nonfamily farms). For four zones out of seven, however, classification by size yields coefficients that imply substantial diseconomies of scale. This is a somewhat surprising result, given that the test (Table 4-5) of the effects of farm size alone-that is, with no account taken of other effects that may be associated with size-showed that the 13. Because the dummy variable is equal to one when the farm belongs to the corresponding stratum and is zero otherwise, and because the dummies correspond to the first five strata, a positive effect of farm size will show up as a negative coefficient. When a farm belongs to the siKth (highest) stratum, all size dummies are zero, and farm production will just equal the positive intercept. FARM SIZES, TYPES, AND PERFORMANCE 97 Table 4-7. Comparison of Marginal Products of Labor on Family Farms with the Probability of Finding Alternative Employment Labor marginal products Probability of on family farmslWage rate alternative t- Zone on nonfamily farms employmenta statistic West 0.52 0.51 0.01 Middle-North 0.55 0.46 0.31 Sertao 0.57 0.51 0.15 Semihumid Southeast 0.30 0.54 0.81 liumid East 0.39 0.43 0.11 Humid Southeast 0.34 0.42 1.10 Agreste 0.51 0.48 0.08 a. The probability of finding temporary off-farm employment is estimated as the ratio of average yearly employment per family worker outside his own farm, in man-days, to the standard availability for employment of 300 man-days a year. returns to scale were constant. Our interpretation is that economies of scale as generally measured in agriculture tend to capture the combined effects of several variables associated with the size of the operation. These may include technical efficiency, the proportion of labor supplied by the family, or the degree of decentralization of farm management, all of which are variables whose effect is at least partly captured by the dummy shifters of the equations in Table 4-8. Once these variables are taken into account, the coefficient of scale expresses the economies (or diseconomies) associated with residual factors such as managerial ability and the organization and supervision of farm operations. Some Conclusions The production function analysis leads to some conclusions about returns to scale, efficiency in factor allocation, and profit-maximizing be- havior on Northeastern farms. Returns to Scale There is some mixed evidence of constant or decreasing returns to scale. Although small farms are much less efficient than large farms at the same level of input use, when allowance is made for differences in input use there appear to be substantial diseconomies of scale. Once appropriate allowance is made for input use and other aspects of farm structure, however, small Table 4-8. Scalar and Dummy Variable Analysis of Effects on Gross Output of Farm Size, Sharecropping, and Three 7 jpes of Equipment (standard errors in parentheses; dependent variable = gross output) Semi- humid Humid Middle- South- Humid South- North- Variable West North Sertao east Fast east Agreste east Scalar variable Land value 0.045 0.073b 0.067' 0.354' 0.(09' 0.026 0.060' 0.096' (0.032) (0.030) (0.015) (0.062) (0.034) (0.062) (0.020) (0.009) Labor 0.679' 0.518' 0.477' 0.511' 0.623' 0.505 ) 0.674' 0.603' (0.062) (0.050) (0.028) (0.102) (0.075) (0.207) (0.056) (0.020) Insecticides 0.087 -0.164 0.060b 0.138' _d _-d 0.112' 0.047' (0.060) (0.158) (0.021) (0.039) __d __d (0.027) (0.013) Fertilizer __d _d _d __d a).0 __d 0.063' _d _a __a __d (0.022) _d _-a (0.010) Credit, 0.017 0.019 0.036' -0.014 0.038 __d 0.061' 0.021' (0.017) (0.015) (0.008) (0.021) (0.024) _a (0.013) (0.005) Structures 0.028 0.031 0.088' 0.062 0.026 0.174 0.011 0.063' (0.027) (0.024) ((.019) (0.046) (0.038) (0.173) (0.028) (0.011) Equipment 0.106 0.078' 0.055' 0.043 0.052' 0.024 0,036 (0.052- (0.031) (0.025) (0.015) (0.046) (0.040) (0.044) (0.027) (0.010) Livestock 0.072 0.070 0.085' 0.026 0.069b -0.080 0.045' 0.071' (0.021) (0.019) (0.011) (0.033) (0.025) (0.098) (0.019) (0.007) Coefficient of scale (Y.Bi) 1.034 0.625 0.868 1.120 0.866 0.699 0.999 1.016 t test (t[lBi - 1]) (0.012) (-6.015) (-2.134) (0.812) (-2.174) (-1.715) (0.001) (0.131) Constant 4.917' 5.616' 4.766' 2.500' 5.454' 8.106' 5.511' 4.984' (0.282) (0.293) (0.210) (0.801) (0.576) (1.608) (0.374) (0.116) Dummy variable Stratum I -0.316 - 1.122' -0.896' - 0.807' -0.309 - 3.061c -0.792l -0.834' (0-9.9 hectares) (0.212) (0.235) (0.148) (0.490) (0.414) (1.276) (0.286) (0.085) Stratum 2 -0.232 -0.962' -0.551' -0.025 0.688 -2.104c -0.499b -0.521a (10-49.9 hectares) (0.153) (0.163) (0.119) (0.343) (0.366) (1.079) (0.286) (0.085) Stratum 3 -0.241 -0.796' -0.422' -0.023 0.310 - 1.581 - 0. 519c -0.435' (50-99.9 hectares) (0.160) (0.157) (0.114) (0.309) (0.373) (1.105) (0.242) (0.067) Stratum 4 -0.090 -0.688' -0.350' -0.493' 0.560 - 1.251 - 0.205 -0.330a (100-199.9 hectares) (0.154) (0.138) (0.108) (0.028) (0.357) (1.049) (0.230) (0.063) Stratum 5 -0.102 -0.546' -0.167' -0.104 0.545 - 1.286 -0.067 -0.228' (200-499.9 hectares) (0.130) (0.119) (0.107) (0.258) (0.366) (1.123) (0.238) (0.061) Proportion of farm - 0.353a -0.430' - 0.340' - 0. 703' -0.810 -0.0001 - 0.612' - 0.487a labor supplied by (0.095) (0.093) (0.049) (0.166) (0.148) (0.379) (0.097) (0.035) familyf 0.114 1.011 0.666' 0.791c 0.848 - 0.835' 0.761a Sharecroppers (0.213) (0.123) (0.061) (0.347) (0.590) - (0.198) (0.052) Equipment NO Irrigation -0.722 -0.247 0.136 -0.820 0.080 - 0.233 -0.048 (0.306) (0.634) (0.196) (1.138) (1.186) - (0.348) (0.164) Mechanical - 0.056 - 0.152 0.073 0.358 0.051 0.263 - 0.028 0.084 (0.184) (0.172) (0.081) (0.261) (0.233) (0.214) (0.135) (0.058) Animal traction 0.031 -0.164 0.233 -0.460 0.167 0.111 -0.093 -0.041 (0.130) (0.159) (0.063) (0.212) (0.220) (0.561) (0.113) (0.046) R 2 0.560 0.514 0.623 0.640 0.645 0.869 0.614 0.589 Degrees of freedom 502 627 1,871 263 292 32 687 4,366 - Not applicable. the effect of several "residual" or unspecified variables related to manage- a. Significant at the less than 0. 1 percent level. rial ability and the ability to obtain financial resources. It is very difficult, b. Significant at the 0.1-1 percent level. however, to separate credit from other forms of capital. c. Significant at the 1-10 percent level. f. In alternative specification without dummy variables for size. d. Too few observations to estimate the coefficient. Source: SUDENE/World Bank survey. e. The credit variable has to be interpreted as a shifter that absorbs 100 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL farms appear to have lower levels of production than those they are apparently capable of achieving. This may be because small-scale farmers have restricted access to inputs other than labor; rely on family workers who have unequal skills; or have more difficulty than large-scale farmers in marketing their output. Small family farms do maximize expected income, if their use of labor is indeed regulated bv the wages and chances of obtaining employment on the large farms. The small family farms'absorp- tion of labor is also socially efficient: Although they use too much labor for optimum productive efficiency and its marginal product is low, the "extra" output they produce partially compensates for the underproduction on large farms. Thev thus help offset the effects on resource use of the skewed land distribution, the lack of well-developed land markets, and the only partially productive efficiency of the large farms. Efficiency in Factor Allocation For the great majority of Northeastern farms, land and labor are essen- tially the only two factors available. Labor is the onlv factor for which markets, delivery systems, and mobility are sufficient to balance demand and supply. Large (nonfamily) farms presumably have greater freedom to choose both land and labor intensities so as to maximize profits if they wish to do so, whereas familv farms are more likely to be constrained bv lack of financial resources, poor access to credit, and small land endowments. Even so, large farmers tend to employ less labor than they would require to maximize profits, even after proper account is taken of their relative abun- dance of land. Conversely, small farmers tend to employ more labor than profit maximization would warrant, probablv in part because family mem- bers have difficulty in obtaining enough employment elsewhere. Profit-Maximizing Behavior There are strong suggestions that factors other than labor are not being allocated according to profit-maximization rules. The marginal productiv- ity of land on large farms is far below reasonable opportunity costs, and for permanent land improvements (structures) and livestock the marginal rates of return are erratic and generally below those from comparable assets outside the Northeast. Modern inputs and credit are consistently associated with higher output, even for small, homogeneous groups of users. Utility maximization, and farmers' expectations about prices and yields, may account for the divergences from optimal behavior. But an alternative explanation may be that, in the use of factors other than labor, market FARM SIZES, TYPES, AND PERFORMANCE 101 conditions are such that the majority of Northeastern farmers have little freedom of choice. The problems, development options, and policy choices facing agricul- ture in the Northeast are diverse and complex and demand that all available tools of analysis be brought to bear. The analysis presented thus far has permitted the description of statistical patterns and identified some hypotheses; some of the questions raised by the descriptive statistics have been addressed in a more systematic way by the econometric analysis in this chapter. One of the advantages of the econometric methods used is that they permit hypotheses to be tested with a minimum of assumptions about underlying behavioral and technical causes. Often, however, statistical analysis cannot go far enough in explaining the relations it reveals and indeed may simply reveal new hypotheses to be tested. For our purposes in this study, it has two other important drawbacks. First, when a large number of variables need to be considered simultaneously, the treatment of the data needs to be highly aggregative, and important aspects of farmers' behavior-for example, their responsiveness to price changes or their aver- sion to risk-are ignored as a consequence. Second, and even more limit- ing, is the fact that the statistical tables and tests can only provide a view of an existing situation: They do not permit us to analyze how that situation would change if its underlying causes were changed. The programming model to which we now turn is designed to complement and extend the statistical analysis, to reach a deeper understanding of the workings of the agricultural economy and the constraints on its productivity and expansion. 5 The Structure of the Programming Model T'he value of large quantitative models, which can bring many variables and constraints into focus simultaneouslv, has been demonstrated in many agricultural policy and project applications. Such models cannot adequate- ly treat nonquantifiable social, political, and institutional influences on an economy. But otherwise their scope is limited onlv bv the data at hand, the skills of the modelers in translating what they observe into a mathematical framework, and the algorithmic facilities available for deriving solutions. The modeling exercise in this study was undertaken for two main reasons. First, a properly specified model can go beyond the "positive" or explanatory function: If it can adequately simulate the main characteristics of the economy and can identify the factors constraining growth and income, it can then be used in the normative sense to estimate the changes that would be induced by policy interventions or alterations in economic structure. We constructed the model with a view to simulating the effects of alternative actions, previously considered or hypothesized, that have a hope of alleviating the Northeast problem. The second reason was a desire to use an internally consistent quantitative framework of analysis, particularly in view of the inconsistency of existing data sources on the Northeast but also to evaluate the effects of policy interventions. Evaluating proposed in- terventions in a partial context, rather than tracing their effects through the economy as a whole, can give highly misleading results. The direct effects of a new policy may be desirable but may be lessened or nullified by its indirect effects, which are more difficult to predict and observe. The model selected for this endeavor follows that popularized by the Duloy and Norton CHAC model for Mexico and applied to several other developing economies for the analysis of agricultural policies and identifica- 102 STRUCTURE OF THE PROGRAMMING MODEL 103 tion and analysis of projects.I These models are linear programming mod- els, though they can incorporate many nonlinearities. In the sense that their solution procedure depends upon a linear (or linearized) function to be optimized, they are optimizing models. The objective is usually specified, however, so as to yield a competitive equilibrium; for example, maximizing the sum of producers' and consumers' surplus will (approximately) equate market prices with the marginal costs of production. These models typical- ly contain detailed representations of the production technologies being used in the different areas or the productive units they consider, based on Leontief assumptions. Thus, for each of the different areas it considers, such a model will yield a production pattern that conforms to competitive equilibrium criteria-that is, a pattern in which product prices are equated to marginal costs, assuming that producers maximize profits. The model will also yield aggregate welfare indexes for producers, consumers, and government, as well as tabulations of employment, land use, fertilizer demands, and such items ordinarily of interest to agricultural planners and policymakers. Two considerations led us to design the Northeast model rather dif- ferently than such accepted models as CHAC. First, the diversity and com- plexity of agriculture in the Northeast pointed to the need for a model composed of individual models for different types of farms in different physiographic zones. Second, because the descriptive analysis suggested that many of the economic, social, and political problems of the Northeast are associated with the distribution of resources, the model contains an accounting framework for employment and income by type of agent: farmers of different types, sharecroppers, permanent workers, and tempo- rary hired workers. Hence the solutions may be analyzed both in terms of their efficiency-as measured by aggregate output, income and employ- ment, and the utilization levels of the basic factors of production-and in terms of their distributional characteristics-the patterns of employment and income over types of farms and types of agents. Because readers will have different degrees of interest in the model structure, the first section of this chapter gives a general description in- tended for the layman. The following sections are more detailed, and purely technical material, including the algebraic statement of the model, is contained in the appendixes to the chapter. 1. John H. Duloy and Roger D. Norton, "CHAC: A Programming Model of Mexican Agriculture," in Multi-Level Planning: Case Studies in Mexico, ed. Louis M. Goreux and Alan S. Manne (Amsterdam: North-Holland Publishing Company, 1973). 104 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL An Overview of the Model The individual farm models that are the microeconomic foundations of the sector model view the farm as a decisionmaking entity, endowed with resources and facing production possibilities and opportunities to augment its resources-for example, through hiring labor or borrowing on credit. In general, the farmer maximizes profits subject to the constraints on his production and a penalty on the riskiness of the farm plan chosen. It is assumed that all farms face a common core of activities and con- straints. Augmenting this core are features that distinguish farm models of different types in different zones: differences in resource endowments, constraints, production, and the supply of farm inputs; and different input parameters, such as the elasticities of demand for their particular products. The typology of Northeastern farms used in the model is derived in Appendix A to this chapter. A brief description follows. Family farms have an average size of 10-30 hectares, depending on the zone, and supply more than three-fourths of their own labor. Though many are subsistence farms (the majority in the West, Middle-North, Sertao, and Agreste), a third of them are commercially oriented, marketing 80-90 percent of their output. Family farms make up just under a third of all farms in the Northeast. Estates, which average more than 500 hectares in most zones, are tvpicallv livestock fazendas in the West and Middle-North; fazendas with sharecrop- pers, livestock, and a feudalistic structure in the Sertao; and sugar or cacao plantations in the East or Southeast, relying on hired labor. The generic term "estates" is used for these large farms for the sake of convenience, but it should be borne in mind that many of them are not richly endowed; nor are their owners necessarily wealthy. As measured by the application of labor to land and by the proportion of land under crops, much of the land on estates is not intensively used. Estates account for about 6 percent of all Northeastern farms. Medium-sizefarms have an average of 50-100 hectares, depending on the zone, and market more than half their output. Other characteristics vary from zone to zone. In the coastal areas they are mainly modern farms relying on hired labor and producing almost exclusively for the market. Medium-size farms make up roughly two-thirds of the farms in the North- east. STRUCTURE OF THE PROGRAMMING MODEL 105 The Core of the Farm Models Table 5-1 presents the core matrix of each farm model, which comprises crop production and selling activities; the total use of labor and the hire of temporary or permanent labor; constraints on the availability of land of different types, working capital, and family labor; and balances on produc- tion and product distribution, and labor sources and use. The objective function is, in the core matrix, the maximization of the difference between the returns from product sales and the costs of labor use and other inputs. The crops that can be produced and the possible ways of producing them obviously differ according to the zone; the derivation of production coef- ficients is described in Appendix C to this chapter. In all cases, a production activity is defined as the cultivation of one hectare of land, with correspond- ing requirements of labor by month and working capital (short-term finan- cial capital). The outputs of the production activities, in physical terms, are registered in the output balance rows, which are linked to the product- selling activities. Labor requirements are linked to labor summation activi- ties, which register the total labor requirements of the farm plan. These are linked to the labor balances from which the labor demands are supplied. The labor balance rows are constrained by the availability of family labor, on the assumption that family labor will be used before outside labor is hired. The main constraints on the farm models are thus on land, family labor, and working capital. Land as represented in the models is of three types, good, medium, and poor, classified according to farmers' opinions of the value of their land, supplemented by estimates of its productivity. In general, "good" land is flat and near a water source or in a humid, low-lying area; "medium" land is hilly or more arid; and "poor" land is useful only for natural pasture or occasionally certain tree crops. The average farm family is assumed to have five members, of whom 2.2 (man-equivalents) are workers. The family labor constraint can be relaxed by hiring temporary workers (on a monthly basis) or permanent workers (on an annual basis). All workers are assumed to be equally productive. Their wages differ, however; family workers have a reservation wage rate about half that of temporary workers. This was the rate for family workers estimated with production function analysis, and it corresponds quite closely to the expected wage in the Elarris-Todaro sense (namely, the Table 5-1. Core Matrix of Farm Models Activity block Temporary Permanent Constraint Crop Product Total labor labor block production selling labor bire bire Objective function - - - - - + + + +. - - - - - - - Good land 1 1 1 'LG Medium land LM Poor land L, + + - Labor + =0 requirements + + + + - + + -1 Output + balances + + - Working + + + + __ WC capital Labor . Family balances labor Note: A minus sign indicates a negative number, and a plus sign indicates a positive number, in Tables 5-I through 5-5. STRUCTURE OF THE PROGRAMMING MODEL 107 current wage rate multiplied by the probability of obtaining a job as a temporary worker),2 as was seen in Chapter 4. Permanent workers, con- tracted on an annual basis, are cheaper per month than temporary for the same reasons but will be contracted only if the farm plan can use them fully over the course of the agricultural year. (Sharecropper labor is treated differently in the model and is discussed below.) The availability of short-term financial capital, that is, the previous year's savings plus any available short-term credit, is very difficult to capture adequately in linear programming farm models. To do so would require detailed disaggregation of the production year, knowledge of the farms' budgetary constraints in different seasons and of their intertemporal utility functions, as well as assumptions about the operations of short-term credit markets, their lending criteria, and so forth. Such details are beyond the scope of this study, but because experience in other areas suggests that agricultural credit plays an important role in the adoption of modern techniques and the full employment of land and labor resources, we include a working capital constraint in each farm model. As a proxy for the short-term capital available, the model registers, on an annual basis, all the money costs of the activities undertaken by the farm, as reported by the survey. These costs include the wage payments to workers; the costs of seeds, fertilizer, and other inputs purchased for the cultivation of crops; and the costs of rations and veterinary services for maintaining a livestock herd. These expenditures are registered in the working capital row and are constrained by the average amount actually spent during the survev year by the farm type being represented (an amount reported separateiy in the survey). This treatment does not explain farm expenditures; actual expenditures may depend on the perceived profitability of the enterprises farms could undertake or on individual preferences and attitudes toward risk, as well as on the amount saved from the previous years' operations or the restrictions of the credit market. Nonetheless, with such a formulation the model solutions can broadly indicate whether easier access to short-term credit would have beneficial effects and what interest rates farmers would, on the average, be willing to pay for it. It should be borne in mind that the use of cash outlays as a proxy for the available short-term capital can give the impression that the entire amount spent within a year is available to the farmer at any given point in the year-obviously an oversimplification. Hence, the average expenditure level that is used as a constraint in the 2. J. R. Harris and M. P. Todaro, "MNigration, Unemployment, and Development: A Two-Sector Analysis," American Economic Review, vol. 60, no. I (March 1970), pp. 126-42. 108 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL model must be considered as an upper bound to the short-term funds effectively available to the farm. Moreover, because the sources of such funds are not differentiated in the model, any result that shows working capital is scarce will hold even more conclusively for the manv real situa- tions in which only a part of farm expenditures can be financed bv a single, uniformly priced, outside credit source. Two further aspects of crop production, not mentioned in Table 5-1, are reflected in the linear programming vectors. The first is the interplanting of two or more crops in the same field, which is more common than single- crop planting in the Northeast, partly for agronomic reasons and partly to reduce risks.' 'I'he model treats such a "consortium" much like a single crop-that is, joint production of the individual crops concerned-and uses the production coefficients estimated for those crops as though they were being grown separately. (The model's treatment of cropping activities is described in Appendix C.) Second, many crops grown in the Northeast have a life cycle longer than one year, the time coverage of the model. The method adopted for this study divides these crops into two categories: medium-cycle crops (three to seven years) for which decisions in a one-year, comparative static model are relevant; and long-cycle crops, some of which have lives of 100 years or more, for which decisions to expand or contract the stock of trees have no effect on production in a one-period model. For the medium-cycle crops, tree cotton, sugar, and bananas, the stream of inputs and outputs is dis- counted and averaged, and where consortiated planting is possible in the first one or two years, rotation constraints are added. Cotton moco, for example, is a five-vear crop that is interplanted with corn and beans in its first year. The rotation constraint requires that for every hectare of moco grown in consortium, as represented in the farm model, there must be four other hectares wholly of moco being grown on that farm. In this manner the model for a single year represents a steady state while capturing, to some extent, the investment nature of the medium-cycle crops. It is beyond the scope of a one-period model to capture investment decisions on long-cycle tree crops such as cacao, coconut, oranges, and the oil palms babacu and carnauba. For these, the production activities in the model represent harvesting from the existing stock of trees. Because the size of the harvest is constrained by the actual stocks of trees, the decision is simply to harvest them fully or not. The shadow prices on these stock constraints are of interest, however; they can indicate the net return, for the 3. The spacing requirements of first- or second-year tree cotton, for example, are such that annual food crops can be grown in between the young trees. STRUCrTURE OF THE PROGRAMMING MODEL 109 year of the solution, from increasing the stock and may be compared, if data become available, with the costs of establishing new plantations. Additional constraints on production activities not shown in Table 5-1 are proxies for agronomic characteristics that are not explicitly captured in the model. Rice, for example, can be grown only near rivers, whereas cotton herbaceo requires humidity and low-lying land, and the amounts of land on which these two crops can be grown are constrained by the estimates from the survey. Characteristics of Individual Farm Models The characteristics that distinguish the farm models and augment the core matrix common to all of them are discussed below in turn. Farm Family Consumption Most farms in the Northeast, except the large commercial farms and those situated near market centers, produce food for their own consump- tion. Corn, beans, manioc, and rice are the main crops produced for this purpose, though the combination varies according to the zone. Within zones, the survey data show that the market value of the bundle consumed varies little among different types of farms. To model on-farm consump- tion, for those farm types in which it is applicable, estimates are made for the average farm in each zone of the amount and value of consumption of the three principal self-produced foods. The farm is assumed to change its consumption mix according to movements in the relative prices of the three crops. (If corn prices rise, for example, it will be more profitable to sell corn and to reserve more rice for family consumption, as a substitute.) Three alternative consumption bundles are then derived for the farm, each domi- nated by one of the three main food crops in that zone and each with the market value already estimated for consumption on that farm. The three crops are substituted for one another, across bundles, at rates equal to the observed price ratios.4 The part of the farm model matrix relevant to family consumption is shown in Table 5-2. The crux of this treatment is that the farms that are observed to supply their own food are required to do so in the model solutions, but they may vary the composition of the consumption 4. For a fuller description of this method, seeJohn H. Duloy and Roger D. Norton, "Prices and Incomes in Linear Programming Models," American Journal of Agricultural Economics, vol. 57 (November 1975), pp. 595-98. 110 THE AGRICUILTURAL ECONOMY OF NORTHEAST BRAZIL Table 5-2. Family Consumption Tableau Activity block Constraint Crop Product block production sales Consumption Objective function - _ _ _ + + + + + + Output + balances + + . _ _ =0 + Consumption I I I C constraint bundle according to the relative prices of its components. Family consump- tion activities are included in the models for all the farms in the West, all farms except estates in the Agreste and Sertao, and family farms in the East and Southeast. Risk Rainfall varies widely from year to year throughout most of the North- east, the coastal areas excepted, rendering crop production quite risky. Inadequate or untimely rainfall may destroy certain crops but have little effect on others. Furthermore, different crops may respond in opposite ways to climatic variations: Little or late rainfall may ruin a bean crop but induce high yields of some varieties of cotton. A wide body of literature recognizes that farmers are responsive to such variations and generally adopt farm plans that are less than optimally efficient in order to reduce the risk of failure, The present study uses techniques developed by Hazell for farm models and by Hazell and Scan- dizzo for sector models that capture the influence of risk in linear program- ming applications.' As Table 5-3 shows, these techniques involve the inclusion in the model of a time series of actual deviations from mean S. P. B. R. Hazell, "Farm Planning under Uncertainty," American Journal of Agricultural Economics, vol. 53 (February 1971), pp. 53-62; and P. B. R. Hazell and P. L. Scandizzo, "Competitive Demand Structures under Risk in Linear Programming Models," American .Journal of Agricultural Economics, vol. 56 (May 1974), pp. 235-44. STRUCTURE OF THE PROGRAMMING MODEL il Table 5-3. Risk Tableau Activity block Sum of Constraint Total negative Risk block production deviations penalty Objective function -f (0) + - +- 1 Revenue deviations + + ± + + Sum of negative 1. I -1 =0 deviations revenues per hectare, summing the negative deviations, and charging a risk penalty to the objective function. This penalty varies directly with the riskiness of the combination of enterprises selected and thus provides a measure of the tradeoff between the expected profitability of the farm plan and the risk associated with it. Because different types of farms in different regions presumably have different attitudes toward risk taking, the penalty is scaled by a factor phi (tk): The lower the phi, the greater the weight of expected profit in the objective function; and the higher the phi, the more risk-averse the simulated farm plan. Phi is also taken to be larger, the smaller the farm: Subsistence farmers iP drought-prone areas need to ensure survival, whereas large farms are undoubtedly more willing and able to gamble for higher expected profits. In the present study, phi varies from zero, for large farms near the humid coastal areas, to a value of 2.5 in the interior drought-prone areas. Risk considerations apply in all the farm models, though in the models where phi is zero they have no effect. Livestock The importance of cattle production and the proportion of agricultural land used for native and cultivated pasture (about 75 percent) do not allow us to ignore this subsector. In some places, cattle production competes with crop production for resources, particularly medium-quality land and work- ing capital. The treatment of livestock in the model is designed to capture 112 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL this competition as simply as possible.6 It ignores small animals, even in the few instances of specialization in small-animal production (scattered chicken and pork enterprises near towns), and deals only with work ani- mals, beef, and dairy cattle. Because the survey reveals that the great majority of cattle enterprises do not specialize in different aspects of production such as breeding, dairying, or beef production, the model does not include differences in herd composi- tion, either among zones or among farms of different tvpes. It thus refers to the maintenance of an "average" animal. Inputs, by contrast, vary substan- tially, both among zones and among farms of different types. Those other than labor may be classified and represented as follows: -Items that are necessary no matter what the feeding technology, such as veterinary services, vaccinations, and mineral supplements (particu- larly salt). For this class of inputs, the total average annual cost per animal is simply charged to the objective function. -Native pasture, which is the poor-quality land in the model, requires no labor for its maintenance and is the basis for the lowest level of feeding technology. -Planted pasture is land of either high or medium quality cultivated for grass. The model derives labor coefficients for planted pasture as it does for crop production, but the "output" of pasture enters directly into the livestock-feeding activities, as described below. -Purchased rations are available in many different types, the nutritional content of which is to a large extent reflected in their prices, and those at a given price are close substitutes for one another. In the model the various types are denoted by their prices, in cruzeiros. Because the amounts of native pasture, planted pasture, and rations per animal vary widely with and across zones, we assume they can be substi- tuted freely for one another. Cobb-Douglas production functions were used to construct three alternative livestock maintenance technologies for each zone. These alternatives, shown in schematic form in Table 5-4, are in turn intensive in planted pasture, natural pasture, and purchased rations, and each represents a point on the three-dimensional Cobb-Douglas unit isoquant. The row FA T serves as a convex combination constraint permit- ting combinations of the three technologies. The labor rows register the total labor requirements, by month, for the technology employed, includ- 6. For a more detailed livestock model, see Wilfred V. Candler, "A Demonstration Model for the Dairy Industry of La Laguna," in The Book of cHAc: Programming Studies for AMexican .Agriculture, ed. R. D. Norton and L. Solis M. (Baltimore, Md.: Johns Ilopkins University Press, forthcoming). STRUCTURE OF THE PROGRAMMING MODEL 113 Table 5-4. Livestock Tableau Activity block Livestock Constraint feeding block alternatives BEEF RA T Objective function + -1 Working +1+ + capital + + + Planted pasture + + + Natural pasture + + + = 0 Rations + + + FAT 1 =0 Labor + ing the cultivation and maintenance of planted pasture, if used. The activity BEEF registers the total number of animals maintained. It has an objective function entry equal to the average annual revenue per animal derived from sales of meat and dairy products, net of the technology- invariant costs (veterinary services, vaccinations, and mineral supple- ments). RAT charges the costs of the rations to the objective function and the working capital constraint. This somewhat simple structure permits the size of the livestock herd to be determined endogenously, gives some flexibility as to production tech- nology, and captures the competition between livestock and crops for land, labor, and working capital. Livestock activities are included in the models for all farms in the West and for medium-size farms and estates in all other zones, except the cacao plantations in the Southeast. Sharecropping Sharecropping is an important form of agrarian contract in the North- east; it is concentrated in the cotton-growing regions of the Sertao and to some extent in the rice-growing areas of the West. Debate continues in the literature over the efficiency of resource allocation, impediments to innova- tion, and myriad other issues surrounding share contracts. These issues TIable 5-5. Sharecropping Tableau Activity block Sharecropper Sharecropper Number Constraint Sharecropper product family of block production shares consumption sharecroppers Good land + + Medium land + + Plot size ........I -p =0 Sharecropper + + -b labor . balances + + b Landowner s+. product balances s + Sharecropper (l -)+ _ 1 _ _ = 0 product balances (I -s)+ .I _ _ Sharecropper consumption 1.1 -- I I =0 level Note: The parameters are: p average plot size; b = average size family in terms of labor available; and s = share of product accruing to the landowner. STRUCTURE OF THE PROGRAMMING MODEL 115 cannot be ignored in a rural development study. Their complexity, and the lack of previous linear programming studies addressing them, forced us to undertake a background study of the economics of sharecropping in the Northeast before this form of contract could be incorporated in a framework that represents the farm in its entirety.7 That study focused on a single typical sharecropped plot in the Sertao, in a partial equilibrium frame not concerned with other aspects of farm behavior. It examined the nature of the share contract and the way changes in the contract would influence incomes of landowner and sharecropper, gross outputs, and efficiency of resource use. One of its principal features was the recognition that landowners and sharecroppers have different, and presumably opposing, objectives that must be reconciled in the share contract. The landowner is presumably concerned only with his money income from the sharecropped plot, whereas the sharecropper is interested in his own money income, his family's consumption, and his leisure. The principal conclusions of the study were that the observed pattern of crop shares reflected agreement on the farm plan between landowner and share- cropper and was Pareto-optimal, and that this condition held only for a small range about the actual shares. Any change in favor of the sharecrop- per might encourage the landowner to choose other methods, such as cultivating the land himself with hired labor or using it as pasture. At a minimum, imposing a farm plan on a tenant would require enforcement costs; it might also induce the sharecropper to seek alternative employment. The treatment of sharecropping in the present model has the following elements: The shares accruing to the landowner are those actually observed; the plot size, which is largely determined by the size of the sharecropper's family and its ability to supply labor, is fixed at the average observed size; and the farm plan adopted must allow the sharecropper to supply his family's food at the average on-farm consumption level of the zone in question. Though these assumptions are somewhat restrictive, they permit a fa- zenda model in which the number of sharecroppers and their levels of income are determined endogenously. They also permit an explicit choice between sharecropping and cultivation by the landlord with hired labor, and a choice of production pattern, including that between crops and livestock. Table 5-5 illustrates the treatment of sharecropping. Although this is not made explicit in the tableau, sharecroppers compete with owner- 7. Gary P. Kutcher and Pasquale L. Scandizzo, "A Partial Analysis of Sharetenancy Relationships in Northeast Brazil,"Journal of Development Economics, vol. 3 (December 1976), pp. 343-54. 116 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL administered production for both good and medium-qualitv land and com- pete with livestock production for medium-quality land. Sharecropping, however, requires no drawings on the landowner's working capital, unlike other production strategies. We also assume that risks are shared between the landlord and the sharecroppers, so that an indirect form of cost sharing (through the sharing of the landlord's risk penalty) is introduced. Share- cropping activities are permitted in the models for all medium farms and estates in the West and the Sertao. Linkages among the Farm Models The system of computer programs that generates, solves, and reports solutions can handle a single farm model, a collection of farm models such as those representing a zone, or all the farm models in the region. The choice of the level (farm, zone, or region) on which to base particular aspects of the analysis depends on the linkages among the farm models or the extent to which they compete for the same resources. The economic relations among farms in the Northeast are in general quite limited, so that farms can be treated as independent self-sustaining units,' Most of the agricultural resources in the Northeast are endowed at the farm level, and for other inputs there is little competition among farms: For modern inputs, demand is very limited; and for labor, unemployment and underemployment are so extensive in the Northeast that the supplv available for wage employment can be assumed to be perfectly elastic. Labor Markets The treatment of labor markets thus focuses on the demand for labor generated by the production activities. As was described above, the farm models reveal the demand in any given month for hired labor after fam- ily labor has been exhausted. Farms can choose to employ temporary (monthly), permanent (annual), or, in certain zones, sharecropper labor from the common pool available in the zone in question. This pool is defined as the number of rural residents economically active in the sector, 8. Small-scale farmers do offer their excess labor to large farms, but given the extent of unemployment this form of interaction is not relevant for the design of the present model. For an analysis of small/large farm interactions, see Gervasio Castro de Rezende, 'Plantation Systems, Land Tenure, and Labor Supply: An Historical Analysis of the Brazilian Case with a Contemporary Study of the Cacao Regions of Bahia, Brazil" (Ph.D. dissertation, University of Wvisconsin, Mladison, 1976). STRUCTURE OF THE PROGRAMMING MODEL 117 adjusted to man-equivalents, less the estimated number of owner-operators and their family man-equivalent dependents. Unless otherwise noted, these labor pools are never exhausted, and considerations such as interzonal migration are not relevant.9 Product Markets In the base case, the main linkage among farm models is effected through the product markets for food crops. Product markets in the Northeast have many complications, only some of which can be adequately handled in this type of model. First, given the isolation of most producers and a tendency toward "safety first" behavior, much of what is marketed is a residual. A large proportion of the food crops is consumed by farmers and their families, sharecropper families, and other farm workers who may be paid in kind. Second, the market outlets are structurally and spatiallv diverse. Of the food crop production not consumed locally in the interior, part is distributed through middlemen to market centers in the Northeast and, for some crops at some times, other market centers in Brazil. Some products, particularly meat and dairy products, cotton, and sugar, are processed locally, whereas others such as babacu, carnauba, and cacao are exported to the Center-South for processing. Third, supplies and hence prices of perishable food crops fluctuate seasonally. Unpublished SUDENE data reveal that the market prices of beans and corn, for example, have varied by as much as a factor of three during some twelve-month periods. Though the model distinguishes on-farm from marketed consumption and disaggregates the demand for food crops by zone, its structure and the lack of data, particularly on trade flows, do not permit a complete consid- eration of such complications. There are taken to be two distinct product markets in the Northeast: that for export crops, of which the bulk of the output is sent abroad or out of the Northeast to other parts of Brazil; and that for other crops, which are almost entirely consumed within the North- east. Demand for livestock products is assumed to be infinitely elastic. The main export crops depicted in the model are cacao, cotton, babacu, carnauba, cashews, castor beans, and sisal. For all except cotton, supply is subject to the capital stock constraints on long-cycle crops described earlier, and demand is assumed to be infinitely elastic. The supply of cotton is 9. A certain amount of seasonal interzonal migration of temporary agricultural workers is observed, particularly from the interior to the cacao and sugar areas during harvest months. The available evidence suggests, however, that excess labor nevertheless remains in all zones and probably in all months. 118 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL subject to two other constraints: at the farm level, on the availability of suitable land and the rotation constraints described earlier; and at the sectoral level, on the ginning capacity of the region. '° Export demand for cotton products is assumed to be infinitely elastic, whereas domestic de- mand for cotton, as for food crops, depends on price. The principal food crops, rice, corn, beans, and manioc, are short-cycle crops whose supply is subject to risk penalties but not, in general, to additional constraints. The exceptions are rice in the West, which requires low-lying, moist land, and instances where these food crops are grown in consortium with cotton and hence face rotation constraints. Sugar, coco- nuts, and oranges, though occasionallv exported, are almost entirely con- sumed within the Northeast. Two of these, being tree crops, are subject to the stock constraints described above. Sugar production is constrained by processing capacity, but since there is a good deal of excess capacity at the production levels recorded by the survey and entered in the model, this constraint does not become binding. The prices of all these food crops, and of bananas, are endogenous to the model. It is assumed that the marketing and transport infrastructure in the Northeast is so limited that the model need not take account of interzonal trade in crops that are not exported. For all the crops for which price is endogenous to the model, the markets in different zones are thus assumed to be completely separate from one another. This assumption may not be entirely valid but was dictated by the lack of data on interzonal transfers and marketing and transport capacities and costs. A test of the full model permitting interzonal trade at transport costs derived from unpublished SUDENE data revealed that this assumption is not critical. Only about 4 percent of all crop production was traded, and this was chiefly rice from the West. The Treatment of Demand The treatment of price-endogenous demand is based directly on that of Duloy and Norton, though modified where appropriate by the risk consid- erations of Hazell and Scandizzo." Its main elements are illustrated in Figure 5-1, a purely static picture of a single crop equilibrium. Though the 10. In the absence of detailed information on ginning capacity, this constraint was not activated. 11. Duloy and Norton, "Prices and Incomes," pp. 591-600; and Ilazell and Scandizzo, "Competitive Demand Structures under Risk," pp. 235-44. STRUCTURE OF THE PROGRAMMING MODEL 119 Figure 5-1. Demand, Output, and Price Relations: Model Structure Price D \~~~~~~~~~~~ PO - w1s S -~~/ l D QuantitV picture will vary with the different demand structures and supply con- straints that characterize different crops, it is representative of the main Northeastern food crops. For each of these, an exogenous linear demand curve DD is specified, taking the quantity (Q°) and price (PO) from the survey data and estimating the price elasticity of demand so as to derive the slope. 2 The supply curve SS is endogenous and is determined by the costs of production (including opportunity costs). It slopes upward because at higher levels of output the resource costs are greater-for example, hired labor must be used, which is more expensive than family labor, and lower-quality land must be brought into production. 13 (For the special case 12. The estimation of the demand functions and price elasticities is described in Appendix C to this chapter. 13. This class of supply functions is analyzed for intercept, intersection, and elasticity in Gary P. Kutcher, "Agricultural Planning at the Regional Level: A Programming Study of Mexico's Pacific Northwest" (Ph.D. dissertation, University of Maryland, 1972). 120 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL of long-cycle crops, production of which is bounded from above by the level of stocks, the supply curve is of course vertical.) The intersection of DD and SS is obtained by maximizing the sum of the area that is below the demand curve and above the supplv curve. This maximization process mav be used to measure consumers' and producers' surpluses, differentiated by the price line, or may be viewed simplv as a device to ensure a competitive solution in which producers are assumed to equate marginal costs with prices for the commodity in question. The solution obtained in this way does not ensure the full use of resources; (lepending on the position of the demand curve, the equilibrium quantities produced could well result in idle land, labor, or other resources. This treatment of demand is not without its limitations: The surpluses in the maximand can be interpreted in various ways, and agricultural price elasticities are notoriously difficult to estimate with confidence. Moreover, it needs to be assumed that markets clear instantaneouslv and that all farmers maximize profits, in the sense that thev choose production patterns that equate market prices with the marginal costs of production. This last assumption in particular is open to question, since the cross-sectional analysis has strongly suggested that many farmers are not profit maximizers in this sense. The characteristics of these limitations will be investigated further in Chapter 6 to assess their importance for the conclusions to be drawn from the model. It can, however, be said at this juncture that the price-endogenous treatment appears to approximate quite closelv the equilibrium conditions of Northeastern agriculture. The approach permits more realistic assumptions than do the obvious, simpler alternatives: Using a fixed-price objective function, one would need to assume that demand for all products is infinitely elastic; and a cost-minimization approach would eliminate the possibilities for expanding output, which is one of the key variables in experiments to increase employment and consumption. Appendix A: Aggregation of Farm Types Although the analysis in Chapter 4 gives a general idea of the diversity of farming systems in the Northeast, the distinctions it draws between differ- ent types of farms are insufficiently detailed to form the basis for a program- ming model of this highly diverse sector. This appendix describes the procedures used to classifv farms into homogeneous groups, which are represented by the farm-level models described in Chapter 5. A combina- tion of linear discriminant analysis and a multivariate sorting procedure is used to group the sample farms into mutually exclusive classes. The STRUCTURE OF THE PROGRAMMING MODEL 121 number and average size of farms in these classes are then estimated for each zone. The appendix concludes with a short note on linear discriminant functions. The Approach Taken Although necessary and sufficient conditions for the unbiased aggrega- tion of programming models have recently been demonstrated by Paris and Rausser, Day's earlier homogeneity conditions maintain their appeal be- cause of their simplicity and ease of application. 14 These conditions are: that the matrix of input-output coefficients be the same for all the farms to be aggregated (technological homogeneity); that the coefficients of the objec- tive function of each farm differ from each other by the same factor of proportionality (pecuniary proportionality); and that the vector of binding constraints for each farm be proportional to the vector for the aggregate of farms (institutional proportionality). As several studies have suggested, the most restrictive and yet the simplest way of interpreting Day's conditions is as a quest for homogeneity among the farms to be aggregated. "I If we assume that the farm population can be divided into groups made up of members identical with each other except for a random component, not only will the aggregation conditions be respected "on average" but also the aggregates can be interpreted as average farms or representative farm types. We also assume that Day's first condition is met; that is, that all farms face the same production possibility set. As noted in Chapter 4, production function analysis revealed no difference in the elasticities of substitution between capital and labor for farms of different types over all zones, and, as Chapter 2 reported, capital/output ratios are similar on farms of all sizes. This evidence is supported by the fact that yields of the major crops do not differ with farm size (Chapter 3). Given that land quality also appears to be homogeneous across farm size (Chapter 2), it appears that the input-output 14. Quirino Paris and G. C. Rausser, "Sufficient Conditions for Aggregation of Linear Programming Models," American Journal of Agricultural Economics (November 1973), pp. 659-66; and Richard ES. Day, "On Aggregating Linear Programming Models of Production," Journal of Farm Economics, vol. 45 (November 1963), pp. 797-813. 15. See, for example, T. A. Miller, "Sufficient Conditions for Exact Aggregation in Linear Programming Models," Agricultural Economics Research, vol. 18 (1966), pp. 52-57; and A. E. Buckwell and P. B. R. Hazell, "Implications for Aggregation Bias for the Construction of Static and Dynamic Linear Programming Supply Models,"Journal of Agricultural Economics, vol. 23, no. 2 (1972), pp. 119-34. 122 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL coefficients for land and capital do not vary for farms of different sizes and types. In Chapter 4, following Nakajima, we drew a distinction between family and nonfamily farms according to the percentage of the labor used on the farm that is supplied by the farm family and the percentage of the farm's output that is marketed. 16 To obtain a more detailed and useful typology that satisfies Day's second and third conditions, the two Nakajima variables are now considered in conjunction with other aspects of farm organization and behavior: the degree of decentralization of farm management, which will be used to measure pecuniary proportionality; land value; and patterns of resource use, which will indicate institutional proportionalitv. The procedure has the following stages: 1. Construct an initial classification of groups of homogeneous farms, determine the number of groups, and tentatively assign to them the observations of farms with extreme characteristics. 2. Compute a set of discriminant functions and critical values to predict the membership of the groups. 3. If necessary, reclassify those observations with extreme characteris- tics in accordance with the criteria given by (2); 4. Classify the residual farms with intermediate characteristics, accord- ing to the criteria given by (2). An Initial Farm Typology The organizational characteristics of large farms are fundamental to their objectives. In the Northeast the plantation is typical of the sugar zone, w,here crop specialization and the traditions remaining from slavery require the landlord to be heavily involved in the administration, with a pyramidal structure to control hired labor. Fazendas, which are found mainly in the livestock areas of the Southeast and in the Sertao, are more diverse and often highly decentralized through permanent laborers (moradores) and sharecroppers. Although the centralized plantation is market-oriented and can be assumed to maximize its profits without regard to subsistence consumption, production on the decentralized fazenda, with its elements of both family farming and capitalistic enterprise, is likely to reflect a much more complex mixture of commercial and subsistence objectives. 16. Chihiro Nakajima, "Subsistence and Commercial Farms: Some Theoretical Models of Subjective Equilibrium," in Subsistence Agriculture and Economic Development, ed. C. R. Whar- ton, Jr. (Chicago: Aldine, 1970). STRUCTURE OF THE PROGRAMMING MODEL 123 The presence of resident sharecroppers, as reported in the SUDENE/World Bank survey, is taken to indicate that a large farm shares some of the organizational and behavioral characteristics of small farms. An initial classification of farms can now be built tup to take account of the proportion of family labor, the proportion of output marketed, and the type of institu- tion regulating the relations among ownership, entrepreneurship, and labor. Traditional farms -Family farms: all farms where the family supplies more than 50 percent of the labor used -Fazendas with sharecroppers: farms of at least 100 hectares, marketing less than 40 percent of their output -Fazendas without sharecroppers -Plantations: farms marketing more than 60 percent of their output, relying mainly on permanent, nonsharecropping workers (moradores) and with more than 40 percent of their land in nonintensive use Transitionalfarms: farms with intermediate characteristics (40-70 percent of output marketed; 30-70 percent of labor supplied by the family), without openly declared sharecropping arrangements, mostly in the Sertao, the Semihumid Southeast, and the Agreste Modern Farms -Nonfamily farms: capitalistic enterprises based on wage labor, market- ing 70 percent or more of output -Family farms: farms marketing 70 percent or more of output, sup- plying over 70 percent of their own labor. These tentative categories are established on the basis of the mean values, for each size class and zone, of the variables considered. Farms are assigned to categories according to their individlual values of these variables. Patterns of Resource Use This initial classification leaves a sizable number of farms that fit none of its categories. To classify this residual group and to test whether the classification criteria are reasonable, new variables are now introduced: the proportion of farmland under crops; labor use (in man-years) per hectare; and the values per hectare of land, livestock, and farm structures and equipment. Patterns of resource use are much more difficult to capture in any ad hoc classification than are market orientation or organizational structure. As Buckwell and Hazell show, however, Day's requirement of institutional Table 5-6. Man-years of Labor per Hectare, by Farm Size and 7one Farm size (hectares) Zone Zone 0-9.9 10-49.9 50-99.9 100-199.9 200-499.9 500+ average West 0.45 0.08 0.03 0.02 0.01 0.01 0.14 Middle-North 0.23 0.08 0.03 0.01 0.01 0.01 0.06 Sertao (0.23 0.08 0.04 0.02 0.02 0.01 0.11 Semihumid Southeast 0.23 0.08 0.04 0.02 0.01 0.01 (.(09 Ilumid Fast 0.44 0.09 0.05 0.03 0.04 0.01 0.25 Ilumid Southeast 0.25 0.16 0.09 0.10 0.07 0.10 0.16 Agreste 0.51 0.10 0.05 0.04 0.03 0.(2 0.33 STRUCTURE OF THE PROGRAMMING MODEL 125 proportionality can be expressed geometrically as the requirement that all farms in the same group lie along a ray through the origin in the resource space. Table 5-6, which gives average ratios of land application to land, clearly shows some scope for aggregation based on institutional proportion- alitv: Most average farms of the same size class appear to lie on the same ray through the origin. When the analysis is extended to resources other than land and labor, and also to observations within the size classes of farms, visual inspection is no longer possible, and a multivariate technique becomes necessary. In a way similar to that suggested by Buckwell and Hazell, we use a clustering algorithm based on a combination of discriminant analysis and centroid sorting. In the following stages of the analysis, observations are shifted from one group to another so as to maximize a relative measure of similarity within the groups: A = variation between groups/variation within groups. This measure, known as Wilks's lambda, is a multivariate statistic that summarizes the proximity of the individual observations in a group to the mean for the group for any number of classificatory variables. At each stage of the sorting process a value of A is thus computed, which together with a set of linear equations (discriminant functions) permits the assignment of the observations that could not previously be classified and indicates the group to which each observation is most likely to belong. The value of A is then varied by allowing observations to change groups in accordance with their predicted membership. After this, a new set of discriminant functions is computed to yield a new set of assignments to groups. The predicted membership is again compared with actual membership and, if A appears to increase, a new shift of observations is allowed. The relocation of observa- tions is stopped when the value of X does not increase further. At this point, the coincidence of the actual membership of groups with their predicted membership will be near its maximum. Results of the Classification Procedure This procedure yields a complete classification of the sample farms. (It also shows that despite some variation among the zones 80 percent or more of the sample farms with extreme characteristics had been classified cor- rectly on the basis of the degree of market orientation and organizational structure.) The majority of the farms with intermediate characteristics are categorized as plantations and transitional farms. Table 5-7 presents the 17. Buckwell and Hazell, "Implications," pp. 124-27. Table 5-7. Northeast Farm Characteristics, by Farm 'Iype (sample means) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Plan- Transi- family Family Characteristic farm croppers croppers tation tional farm farm Farm size (hectares) 107.26 742.06 441.135 550.48 200.27 212.02 106.73 Farm value (cruzeiros) 5,031.60 28,978.55 26,610.19 51,853.01 9,990.79 43,721.04 17,507.21 Total labor' 0.01 0.01 0.01 0.02 0.06 0.08 0.11 Value of structures' 77.30 25.22 19.01 42.07 59.61 94.08 81.85 Value of livestock"' 44.7(0 17.01 17.37 37.99 46.43 36.40 47.28 Value of equipment"' 1.77 1.79 2.57 2.53 3.97 10.89 2.08 Percentage of output marketed 14.22 20.42 8.95 83.47 63.38 91.50 84.67 Family labor/total labor 89.15 13.41 10.28 27.14 55.04 10.20 91.12 Cropland/total land 16.92 20.24 5.82 6.69 21.59 50.41 23.82 a. Man-years per hectare. b. Cruzeiros per hectare. Source: SUDENE[Worid Bank survey. STRUCTURE OF THE PROGRAMMING MODEL 127 mean values of key variables for the different farm types for the Northeast as a whole. These mean values highlight the differences among the farm types. Family farms, whether traditional or modern, appear to have some similar physical and economic traits, with similar average sizes and similar values of equipment, livestock, and structures per hectare. Modern family farms, however, apply ten times as much labor per hectare as traditional family farms, have a higher proportion of their land under crops, market a far larger share of production, and have higher farm values. The average characteristics of fazendas appear to differ little whether or not they have sharecroppers, except in the proportion of land under crops and the per- centage of output marketed. It is possible that a significant number of landlords did not declare their sharecropping arrangements when inter- viewed for the survey. Plantations differ sharply from other large estates. They market the bulk of their output, they have only a small proportion of their land under crops and relatively high values of livestock and of struc- tures per hectare, and at least a fourth of the labor they use is supplied by the family. Transitional and modern farms are relatively small and labor- intensive, with one-fifth to one-half their land under crops, and generally have higher values of structures and equipment than traditional farms. Tables 5-8 to 5-14 show the characteristics of the various farm types zone by zone and the number of observations of each type in the sample. They show that family farms and the two types of estates, fazendas and planta- tions, have the most distinct characteristics and that the prevalent type of estate varies according to the region. 'Ihey also show that for farms of the same type there are very few large differences between the West and the Middle-North. The Farm Typology for the Programming Model For the final classification of farms used in the programming model, only three farm types and five zones are thus distinguished. The three farm types are: family farms, both traditional and modern; estates, combining, according to the zone, fazendas with and without sharecroppers, planta- tions and family estates; and medium-sizefarms, a residual group, dominated in the coastal regions by modern farms and in the interior by transitional farms. The seven zones used for the cross-sectional analysis are reduced to five by subsuming the Middle-North into the West and combining zones D and F into one zone, the Southeast. Table 5-15 gives estimates for the Northeast of the number and average size of the groups of farms for the five zones used in the model. As the notes Table 5-8. West: Farm Ch'aracteristics, by Farm T'ype (sample weighted means; standard errors in parentheses) 'fraditional Modern Fazenda Fazenzda with withoat Non- Family share- share- Family Itransi- fpmily Family Characteristic farm croppers croppers estate tional farm farm 7otal Total labor' 0.088 000K6 0.006 0.0W5 0.069 t).061 0.129 0.063 (0.022) (0.002) ((000I) (0.002) (0.014) (0,018) (0.034) (0.(08) Value of structures0 0.276 0.838 0.656 0.112 2.994 13.546 0.905 1.886 (0.055) (0.834) (0.464) (0.052) (2.047) (7,854) (0.366) (0.844) Qt Valueof livestockb 48.0)6 29.761 11.574 28.918 76.067 40,146 146.139 58.769 (14.364) (16.852) (3.452) (15.881) (25.746) (25.967) (57.756) (11.817) Value of equipment' 35.990 13.170 6.592 8.716 38.431 27.993 87.190 33.916 (6.892) (6.720) (1.468) (2.507) (6.511) (10.640) (25.434) (3.846) Percentage of 13.234 5.488 6.589 79.829 61.767 91.994 86.082 41.390 output marketed (1.243) (3.439) (1.177) (3.759) (1.052) (2.008) (1.356) (1.434) Famil,v labor/total labor 0.880 0.120 0.087 0.253 0.561 0.103 0.901 0.544 (0,014) (0.056) (0.014) (0.067) (0.022) (0.023) (0.017) (0.017) Croplanditotal land 0.085 0.069 0.029 0.029 0.129 0.488 0.171 0.109 (0.014) (0.057) (0.009) (0.O08) (0.013) (0A)76) (0.035) (0.008) Numbetofcases 134 8 107 17 209 17 45 537 a. Man-years per hectare. b. Cruzeiros per hectare. Tablc 5-9. Middle-North: Farm Characteristics, by Farm Type (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Nvon- Family share- share- Family Transi- family Family Characteristic farm croppers croppers estate tional farm farm Total Total labor' 0.076 0.004 0.007 0.012 0.038 0.061 0.027 0.035 (0.018) (0.w00) (0.001) (0.003) (0.008) (0.015) (0.008) (0.004) Value of structuresb 0.542 0.326 0.720 0.815 3.280 7.548 0.576 2.027 (0.214) (0.205) (0.213) (0.455) (1.236) (2.685) (0.344) (0.493) Value of livestock' 29.085 5.018 7.824 21.852 26.668 23.017 20.391 21.270 (7.261) (1.519) (1.331) (4.534) (5.709) (7.186) (7.296) (2.574) Valueofequipmentb 38.735 5.957 6.906 12.790 21.711 52.352 33.207 22.754 (9.010) (1.175) (1.057) (2.905) (3.893) (23.885) (12.676) (2.744) Percentage of 15.572 17.003 8.848 80.159 61.919 87.573 86.010 46.911 output marketed (1.347) (2.404) (1.247) (1.704) (1.042) (1.620) (1.767) (1.334) Family laborttotal labor 0.857 0.120 0.098 0.358 0.481 0.132 0.932 0.450 (0.016) (0.020) (0.014) (0.043) (0.020) (0.014) (0.015) (0.015) Cropland/total land 0.177 0.218 0.048 0.045 0.156 0.578 0.143 0.157 (0.025) (0.042) (0.010) (0.006) (0.016) (0.045) (0.033) (0.010) Number of cases 100 34 110 58 220 37 41 600 a. Man-years per hectare. b. Cruzeiros per hectare. Table 5-10. Sertao: Farm Characteristics, by Farm Type (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Family Transi- family Family Characteristic farm croppers croppers estate tional farm farm Total Total labor' 0.078 0.014 0.014 0.013 0.054 0.055 0.080 0.057 (0.007) (0.002) (0.001) (0.001) (0.003) (0.006) (0.009) (0.002) Valuc of structuresb 1.409 1.997 1.449 1.857 3.230 8.832 2.403 2.996 (0.290) (0.527) (0.528) (0.515) (0.468) (3.326) (0.922) (0.360) Value of livestock" 40.230 21.238 21.685 26.408 36.589 40.077 35.496 35.672 (4.387) (2.821) (2.966) (3.634) (1.928) (5.737) (4.444) (1.458) Value of equipmentb 53.362 35.971 21.333 21.965 49.285 61.335 44.656 47.351 (4.832) (5.718) (2.548) (2.930) (4.169) (10.769) (3.6(X)) (2.416) Percentage of 15.905 24.905 11.881 79.868 61.903 87.660 83.115 54.321 output marketed (0.786) (1.521) (1.538) (1.444) (0.507) (0.801) (0.615) (0.696) Family labor/total labor 0.900 0.147 0.126 0.308 0.558 0.119 0.909 0.588 (0.008) (0.019) (0.016) (0.040) (0.011) (0.008) (0.007) (0.008) Cropland/total land 0.180 0.216 0.073 0.071 0.241 0.442 0.256 0.230 (0.008) (0.019) (0.016) (0.040) (0.011) (0.008) (0.007) (0.008) Ntumber of cases 326 59 95 62 826 125 218 1,711 a. Man-years per hectare. b. Cruzeiros per hectare. Table 5-11. Semibumid Southeast: Farm Characteristics, ey Farm Type (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Planta- Transi- family Family Characteristic farm croppers croppers tion tional farm farm Total Total labora 0.075 0.004 0.008 0.013 0.047 0.084 0.055 0.052 (0.021) (0.003) (0.001) (0.003) (0.006) (0.012) (0.014) (0.006) Value of structuresb 6.981 0.137 0.438 0.771 1.542 5.921 0.665 2.924 (6.447) (0.135) (0.288) (0.380) (0.681) (2.184) (0.245) (1.649) Value of livestockb 49.234 15.010 36.166 70.690 45.708 48.133 45.013 47.197 (11.116) (4.678) (12.957) (16.737) (6.638) (16.480) (9.944) (4.445) Value of equipment"' 141.558 66.202 27.151 47.095 65.173 107.980 76.650 84.971 (57.220) (45.993) (4.716) (9.984) (9.130) (15.346) (18.758) (15.263) Percentage of 8.620 29.700 6.214 88.044 69.324 95.906 88.383 54.486 output marketed (1.533) (2.600) (2.488) (2.380) (2.076) (1.803) (1.323) (2.439) Family labor/total labor 0.905 0.141 0.138 0.254 0.626 0.101 0.893 0.630 (0.020) (0.045) (0.031) (0.076) (0.030) (0.020) (0.016) (0.023) Cropland/total land 0.081 0.151 0.010 0.039 0.141 0.403 0.135 0.125 (0.012) (0.129) (0.004) (0.010) (0.017) (0.032) (0.022) (0.(10) Number of cases 60 2 22 16 78 18 42 238 a. Man-years per hectare. b. Cruzeiros per hectare. Table 5-12. Humid East: Farm Characteristics, by Farm 'Iype (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Planta- Transi- family Family Characteristic farm croppers croppers tion tional farm farm 7otal Total labor' 0.186 - 0.037 0.025 0.134 0.081 0(.18 0.108 (0.037) (0.014) (0.003) (0.028) (0.012) (0.003) (0.012) Value of structuresb 1.883 - 3.638 2.627 14.698 16.602 1.086 9.569 (1.268) (1.769) (0.843) (5.960) (5.908) (0.419) (2.439) Value of livestock' 76.023 - 24.227 44.052 54.100 37.491 26.714 46.036 (29.302) (9.394) (9.582) (17.882) (6.961) (8.785) (7.368) Value of equipment' 323.706 - 35.918 47.668 185.651 111.691 69.949 144.832 (135.333) (11.517) (12.608) (54.393) (35.759) (20.160) (26.370) Percentage of 13.803 - 11.879 88.300 68.324 93.479 87.721 69.734 output marketed (2.404) (4.138) (2.131) (2.148) (1.070) (1.569) (1.9(9) Family labor/total labor 0.909 - 0.021 0.192 0.422 0.106 0.928 0.444 (0.025) ((.009) (0.053) (0.032) (0.012) (0.015) (0.024) Cropland/total land 0.299 0.204 0.112 0(.323 0.503 0.249 0.321 (0.054) (0.074) (0.009) (0.029) (0.030) (0.041) (0.017) Number of cases 33 0 14 29 91 62 43 272 -Not applicable. a. Man-years per hectare. b. Cruzeiros per hectare. Table 5-13. Humid Southeast: Farm Cbaracteristics, by Farm Type (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Planta- Transi- family Family Characteristic farm croppers croppers tion tional farm farm Total Total labor' - 0.103 0.116 0.134 0.117 (0.069) (0.011) (0.035) (0.010) Value of structuresh - - - - - 2.451 3.886 0.112 3.496 (2.424) (1.089) (0.067) (0.962) Value of livestockb - - - - 11.272 8.982 4.881 8.725 (5.354) (1.940) (4.079) (1.731) Value of equipmentb - - - - 270.332 234.135 90.303 223.202 (203.011) (27.105) (26.258) (25.223) Percentage of - - - - 100.000 99.325 99.775 99.393 output marketed (0) (0.204) (0.225) (0.180) Family labor/total labor - - - - 0.541 0.031 0.969 0.134 (0.155) (0.008) (0.019) (0.042) Cropland/total land - - - - 0.772 0.722 0.651 0.718 (0.028) (0.017) (0.094) (0.017) Number of cases 0 0 0 0 2 40 4 46 -Not applicable. a. Man-years per hectare. b. Cruzeiros per hectare. Table 5-14. Agreste: Farm Characteristics, by Farm Type (sample weighted means; standard errors in parentheses) Traditional Modern Fazenda Fazenda with without Non- Family share- share- Planta- Transi- family Family Characteristic farm croppers croppers tion tional farm farm Total Total labor' 0.226 0.020 0.028 0.034 0.102 0.117 0.241 0.140 ((.052) (0.005) (0.008) (0.004) (0.011) (0.021) (0.051) (0.014) Value of structures" 2.672 9.902 18.711 5.924 4.769 15.598 3.730 6.322 (0.693) (8.911) (9.892) (1.346) (0.958) (2.841) (1.036) (0.775) Valueoflivestockb 53.664 27.114 39.425 92.420 71.264 49.605 51.350 61.865 (6.650) (14.621) (7.004) (11.097) (8.538) (6.290) (8.287) (3.882) Value of equipmentb 114.002 31.106 78.545 95.669 98.600 91.623 196.627 113.313 (23.197) (16.853) (26.709) (16.335) (11.884) (9.145) (41.737) (9.365) Percentage of 12.809 12.533 9.062 87.808 67,226 92.865 83.302 58.827 output marketed (1.136) (6.343) (2.392) (1.504) (1.186) (0.869) (1.002) (1.341) Family labor/total labor 0.898 0.104 0.115 0.199 0.606 0.095 0.909 0.585 (0.013) (0.042) ((0.024) (0.031) (0.019) (0.008) (0.010) ((1.015) Cropland/total land 0.225 (0.182 0.088 0.081 0,240 0.485 0.296 (0.250 (0.023) (0.057) (0.031) (0.007) (0.018) (0.026) ((0.032) (0.011) Number of cases 142 6 34 61 230 75 92 640 a. Man-years per hectare. b. Cruzeiros per hectare. STRUCtURE OF THE PROGRAMMING MODEL 135 to the table show, the organization and product specialization of estates vary according to the zone. The estimates of the number of farms of each type are obtained by extrapolating the sample results to the population, taking account of the farm size distribution: The estimated percentage of farms of each type in each size class is multiplied by the estimated number of farms in that size class (shown in Table 2-5). Farm sizes are derived in the same way, so that the areas quoted are averages weighted by the number of farms in each size class. Last, to maintain a close correspondence between farm size and the classification, a small number of farms exhibiting the characteristics of family farms, but larger than 200 hectares, were shifted into the category of medium-size farms. A Note on Linear Discriminant Functions To illustrate the use of linear discriminant functions,"8 we consider for simplicity the classification of a sample of two variables (say, XI and X2) into two groups. The procedure used in discriminant analysis is to search for a linear combination of the two variables such that the distributions of the two groups display the least overlapping. The linear discriminant function can be indicated as: (5.1) yit = BI Xilt + 62 Xi2l, i = 1, 2 t = 1, 2, . . ., n, where i denotes the group and t the number of observations within each group. Unlike the case in regression analysis, the variable Yi is endogenous, resulting from the linear combination of the observed variables X[, and Xi2t. The discriminant function in equation (5. 1) can be generalized to the case of j variables: (5.2) Yi = 81 Xil + B2 Xi2 + * + BO Xij + .. i= 1, 2 t 1, 2, . . .,ni where Xit denotes the value of thejth variable for the tth observation in the ith group. The problem of discriminant analysis is to obtain an optimum discrimi- nant plane-or hyperplane in the case of equation (5. 2)-according to some prespecified criterion. A popular method consists of maximizing the ratio: variation within groups variation between groups 18. For a more detailed treatment, see B. W. Boich and C. J. Huang, Multivariate Statistical Metbodsfor Business and Economics (Englewood Cliffs, N.J.: Prentice-Hall International, 1974). 136 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 5-15. Northeast: Estimate of NAumber and Size of Farms, by Type Family Medium Zone farms farms Estates West (zones A, B) Number 37,001 113,551 14,569' Average size (hectares) 18.43 108.72 794.57 Sertao (zone C) Number 62,599 252,912 11,221" Average size (hectares) 16.92 81.21 676.43 East (zone E) Number 11,964 17,623 1,314' Average size (hectares) 9.42 60.06 1,086.50 Southeast (zones D, F) Number 26,824 76,574 4,766d 5,208' Average size (hectares) 26.82 47.72 763.47 441.27 Agreste (zone G) Number 69,580 68,545 5,168' Average size (hectares) 8.97 60.06 634.56 a. Mainly livestock fazendas. b. Mainly fazendas with sharecroppers. c. Mainlv sugar plantations. d. Mainlv livestock fazendas. e. Mainiv cacao plantations. f. Mainly diversified fazendas. For the case of two groups and two variables, the above ratio is given by (5.4) i I PI'(XI-X2)(X,--X2)'P n + n2 - 2 ~ ' where 3 is the vector of coefficients to be estimated; A, = vector of average sample values for the ith group; and S* = variance-covariance matrix for the sample data. The maximization in equation (5.4) requires a normalization rule, gener- ally given by (5.5) c = ' (X1 - X2)ID2 = 1, where D2, known as Mahalanobis's D2 or generalized distance, is equal to (XI - X2) 'S* I (XI - X2). Under the constraint (5.5), the first-order conditions for the maximiza- tion of A yield the result: STRUCTURE OF THE PROGRAMMING MODEL 137 (5.6) 1 = S* ' (Xi - A2). Once the discriminant function parameters are estimated from equation (5.6), it is necessary to determine a critical value in order to classify new observations (or to determine how well the old ones have been classified) into groups I and 2. If the costs of misclassification are equal for the two groups, and there is an equal probability that an observation is drawn from either of them, the critical value of Y* is given by (5.7) I = 2( - X2)' The discrimination rules can then be expressed as follows: Given a new observation Y1, not included in the sample, classify it into group I if Yjt Y*; classify it into group 2 if Yj < Y*. Given an observation Y,, of the sample used to construct the classification, consider it well classified if Y, Ž P* and Yi in group I or Y, < Y* and Yi, in group 2; consider it misclassified otherwise. Appendix B: Algebraic Statement of the Programming Model Indexes z Physiographic zone (A = West C = Sertao D = Southeast E = East G = Agreste) s Farm type (I = family farm 2 = medium-size farm 3 = livestock fazenda if z A 3 = sharecropped fazenda if z = C 3 = diversified fazenda if z = D or G 3 = sugar plantation if z = E 4 = cacao plantation if z = D only) i Commodities (04 = cotton moco 05 = cotton herbaceo 138 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL 06 = cotton verdao 11 = rice 13 = bananas 14 = sweet potatoes 16 = babacu 19 = cacao 21 = cashews 22 = sugarcane 25 = carnauba 30 = coconuts 34 = beans de arranca 35 = beans de corda 43 = oranges 48 = manioc 51 = corn 54 = castor beans 69 = sisal) m Month (1 =January, . . ., 12 = December) k Production activity (= 01, 02, . . ., 80) t Year ( = 60, 61, .. ,70) e Demand segment (quantity axis) (= 01, 02, . . ., 20) q Land quality (G = good cropland M = medium cropland or cultivated pasture P = poor cropland or native pasture) j Livestock feeding alternative (1 = cultivated pasture-intensive 2 = native pasture-intensive 3 = rations pasture-intensive) n Family consumption alternative, explained in Chapter 5 (= 1, 2, 3) Activities (variables) Dzie Demand for commodity i, segment e of demand function in zone z (measurement units, SU in thousands, typically metric tons) Z>zs Negative deviation counter for risk penalty, yeary, farm type s (in thousands of cruzeiros) CPkzsq Proprietor-managed cropping activity k, on land quality q (in thousands of hectares) STRUCTURE OF THE PROGRAMMING MODEL 139 CSkzq? Sharecropped cropping activity (in thousands of hectares) Pi. Marketed production of proprietors' crop output (in thousands of SU) Sjzs Marketed production of sharecroppers' crop output (in thousands of SU) TPzsm Total labor use, month m (in thousands of man-months) LFzfm Family labor use (in thousands of man-months) LTzsm Temporary labor hire (in thousands of man-months) LP, Permanent labor hire (in thousands of man-years) NSzs Number of sharecropping plots (in thousands of plots) EPzsm Proprietor consumption, bundle m (in thousands of bundles) ESz,m Sharecropper consumption (in thousands of bundles) LSFzsj Livestock feeding alternative j (in thousands of bundles) BEEF,z Number of cattle maintained (in thousands of cattle) RA TZ Rations demanded (in thousands of cruzeiros) VETz, Veterinary services demanded (in thousands of cruzeiros) TGMv. Good-quality land used as medium-quality land (in thousands of hectares) TMP7s Medium-quality used as poor-quality land (in thousands of hect- ares) Parameters/data dzie Area under demand function in segment f p, Risk penalty Czk Cost of purchased inputs for cropping activities r1. Price of export crop i v, Value of average consumption bundle ez Value of average annual livestock product sales cv Average annual cost of veterinary services per animal wF Family reservation wage rate per month w T Temporary labor wage rate per month wPz Permanent labor wage rate per year ykzq Yield (kilograms per hectare) of ith product from kth activity Cxi Share of product i accruing to proprietor from sharecropping activities 140 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL u Marketing margin retained by proprietor from sharecroppers' product sales fizn Quantity of ith product in consumption alternative n gk Land requirements per unit level of cropping activity (= 1, except for long-cycle crops [= 0], and manioc [ = 1.5, as average cycle is one and a half years]) Vfczkm Labor requirements (man-days per hectare), cropping activity k, zone z, month m &wJjm Labor requirements, livestock feeding activity j(includes mainte- nance of cultivated and native pasture and herd supervision) X1j Medium-quality land requirements, livestock feeding technologyj x2j Poor-quality land requirements, livestock feeding technology] Oz Sharecropping plot size (hectares per family) azie Quantity sold, product i, segment (. Objective function (OBJ) I Y, X dzie Dzie + Y. : Y. r,5 Pj, + X I vz 5EP, z i f z is Ss n [area under demand + [revenue from + [value of on-farm functions] export crops] consumption] + 2 Y. ez BEEFzs - E X X Pz, ZTZ z sz t + [revenue from livestock - [risk penaltv] product sales] - C S > I Czk (CPzks + CSzksq) - E RA 7w - cv Y. I BEEFz, z k s zq zk Y z I - [purchased input costs - [cost of livestock - [cost of veterinarv of cropping activities] rations] services] -wJ>; Y .I LFZSm -X 7 L7Sm -E; Yw LPZ -[family reservation - [temporary labor - [permanent wages] costs] labor costs] Commodity balances Proprietors (BPZE) Y Ykzq CPkzsq + Oti Ykzq CSkq k Yk q [yield from pro- + [proprietors' direct share prietors' crop- of sharecroppers' cropping ping activities] activities] STRUCTURE OF THE PROGRAMMING MODEL 141 + U Sizs - Pizs + [marketing margin - [product sales]'" from sharecroppers' sales] -E fizn EPZ,, =0 for all i, z, s. - [on-farm consumption] Sharecroppers (BSzis) (1 -otd) I I ykzq CSzsq - Si,, k q [sharecroppers' retained -[sharecroppers' production] "sales"] -Efjizn ESszn = for all i, z, s. -[sharecroppers' consumption] Demand balances (price-endogenous crops) (Qz) - azie D0e + . Pzis 20 for all z, i. e S [quantity sold] + [quantity supplied] Demand convex-combination constraints (CCd) I Dzie ' I for all i, z. Consumption levels (EPZ,, ESZS) . EPzs, = A'. for all z, s, [number of = [number of farm bundles] families]20 I ESzn = NS =O for all z, s. [number of = [number of bundles] sharecroppers] 19. Or transfers to market demand functions for price-endogenous crops. 20. If on-farm consumption is applicable. 142 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Deviation constraints (for risk) (DEVtZs) I Oksq CPkzq + X Okzsq CSkzsq k k [revenue deviations of + [adjusted2' revenue proprietors' cropping deviations of activities] sharecropping activities] + Ztz7 2 0 for all t, z, s. + [negative deviation counters] Land constraints Good cropland (HG,5) Y gk (CPkG + CSk25G) + TGM, [good land used for crops] + [good land used as medium land] HG= for all z, s. .c [net (of long-cycle stocks) good land available] Medium cropland (HM,5) I gk (CPk. M + CSk2G) + E xi. LSFzj [medium land used for crops] + [cultivated pasture use] -TGMz. + TMP. c HM, for all z, s. - [good + [medium c [net medium land land used land used available] as medium] as poor land] Poor cropland (HP.) z X2j LSFZ7j - TMP25 [native pasture use] - [medium land used as poor] - ¼/4 1 (CPk*zq + CSk*zq) - HP, - [residual pasture available in ' [poor land fourth year of cotton moco cycle]22 available] 21. To reflect proprietors' component of risk (0* < 0 since si < 1). 22. k* implies the activity contains crop i = 04 (cotton moco). STRUCrURE OF THE PROGRAMMING MODEL 143 Labor Constraints Proprietors' labor balances (RPm) Y Wzk, CPZkf + X wzLm LSF.., q k [labor required for + [labor required for cropping activities livestock maintenance (man-days)] (man-days)] -25TPz,m =Oforallz,s, m. - [total labor required (man-months)] Proprietors' labor constraints (RCz,,) 7Pzsm -LFzsm -LTz5m - LP <0 for all z, s, m. [total labor - [family labor - [temporary - [permanent required] supply] hire] hire] Note: LFz is bounded at farm family labor availability. Sharecroppers' labor balances (RS,,,,) t E W'zkm CSzksq - 55NS5 ' 0. k q [labor required in - [labor supplied by sharecropping each sharecropper's activities] family per month] Regional labor constraints (LABzm) z (LTzsm + LP. + 2.2NS,,) c LABzm for all z, m. [temporary + [permanent + [share- '[landless workers] workers] croppers] labor available] Sharecroppers' plot size (Oz) x CSzkq -OzNS= O for all z, s. [total area - [plot size sharecropped] times number of sharecroppers] Livestock rations balance (RAMTz) Y x3j LSFzj - RATz. = for all z, s. J 144 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL [rations required - [total rations] in feeding alter- natives] Livestock maintenance balance (FAT7.) E LSFj - BEEFs =O for all z, s. [number of - [number of feeding animals alternatives] maintained] Working capital (WC,.) E X C., (CPzkq + CSzksq) + RATk + cv BEEFZS k q [crop purchased inputs] + [livestock purchased inputs] + w z 3 LTzsm + wzp LP , WC,,, for all z, s. m + [temporary + [permanent ' [annual labor wage labor wage working costs] costs] capital availability] Bounds23 LF,sm 'LFzm for all z, s, [farm family labor availability] CPzks,q ' CPzq for all z, s, q, k e Kl (where KL is the subset of cropping [long-cycle crop stocks] activities comprising long-cycle crops). 23. Bounds are convenient ways of constraining individual activities without recourse to additional rows in the problem. STRUCrURE OF THE PROGRAMMING MODEL 145 Appendix C: Estimation and Testing of Parameters Little theoretical guidance is available on the problem of incorporating statistical information from cross-sectional data into a mathematical pro- gramming model. Although there is a considerable body of literature on the stochastic programming aspects of the problem, no significant theory has been developed that explains the properties of a model's solution as a function of the statistical characteristics of its parameters.24 The matrix of production coefficients used in the model is, with several exceptions described below, composed of means computed from the survey data.2" These have a number of possible sources of error. Despite the large size of the sample, the number of observations of the production relations for any individual crop is quite small, because crops are so frequently interplanted. The frequency of observations varies widely among crops and consortiums, and the coefficients estimated for those with only a small number of observations may be significantly biased. Moreover, although the sample points are scattered over large areas, it is possible that the observations of particular crops were spatially concentrated and thus give a distorted impression of their average distribution and also, perhaps, of their average production relations, because local weather, soil quality, and other ecological factors may significantly affect yields, labor requirements, and so forth. Finally, some of the survey data are better than others, because of differences among the survey teams and the degree of understanding of the informants. In view of these problems, the reliability and consistency of the numeri- cal values of the parameters estimated were analyzed before they were incorporated in the model. The analysis had the following parts: -calculation of the distribution statistics of the basic input-output relations -where the same crop is produced in more than one activity, that is, alone or interplanted with other crops, or both, tests for significant differences in yields 24. As Hadley puts it, "when some of the technological coefficients are made random variables, the problem becomes nonlinear and assumes a form for which, at present, there does not exist any general technique for finding an optimal solution." See George Hadley, Non- Linear and Dynamic Programming (Reading, Mass.: Addison-Wesley Publishing Co., 1964). 25. The full technology data set has been published by SUDENE, as Coeficientes tecnicos do Nordeste (Recife, 1976). 146 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL -comparison of the empirical with estimated theoretical distributions of the parameters for different activities (generally normal) and rejection of outliers. Three of the principal parameters in the model required additional informa- tion and background analysis: -the yield (output in kilograms per hectare) coefficients for the various cropping activities -the risk-aversion factors of different types of farms -the demand activity coefficients that are the basis for the determination of price and quantity of nontraded crops. Later, tests were also made of the robustness of the model by varying the values of its parameters. Some of these tests are described in Chapter 6. The StratiJied Yield Estimator The model treats each combination of land (of good or medium quality) and crop (or consortium of interplanted crops) as a separate activity. Each farm model thus has pairs of cropping activities whose nonzero values are identical (in magnitude and by rows) except insofar as they draw from different resources (good or medium-quality land) and their yield coef- ficients are different. (It is assumed that the variable inputs, labor and purchased inputs, are identical.) The yields from both sets are carried into a single set of commodity balances. The yield coefficients are derived from two mutually independent sam- ples. The first of these has a wide coverage but gives yields irrespective of land quality; and the other, smaller and subject to more measurement error, gives yields stratified by good and medium land quality. To incorporate the greater accuracy of the results from the first (y) and the greater detail of those from the second (yi), the estimates (y*) are computed by minimizing z (Yi yi) + E 2 (y- y)] il Si )i=l S where yi= the stratified mean yields from the second sample si = the standard error of yi y = the mean yield (unstratified) from the first sample s = the standard error of y STRUCTURE OF THE PROGRAMMING MODEL 147 wi = a, / on a;, a weight with respect to y ai = the land area devoted to the activity generating yi n = the number of land types. This equation maps the information contained in the two samples (yi andy) into our estimates and takes into account their sampling variance (s; and s, respectively). 26 After a little algebra, the first-order condition yields Yi = gyi + (1 - g)y, whereg = s2/(5 + wD2si2), i = 1, 2, n. In practice, only two land types (high- and medium-yield land) are used: The yield coefficients for each are determined according to the above equation. Treatment of Risk The treatment of risk is based on that described by Hazell and Scandizzo.27 The risk penalty charged to the objective function in each farm model depends upon the riskiness of the farm plan chosen, as estimated from past downward deviations from average yields, scaled by a factor phi, which indicates the ability or willingness of the farm type in question to bear risk. Revenue Deviations A time series exists for 1960-69 of the revenues generated by single crops, state by state.28 These data are mapped into a regional series using a rough weighting scheme. They are then made applicable to crops grown in consortiums, which is more usual than plantings of single crops, using the following procedure. A weight (wi) corresponding to the contribution of the crop in question to the total revenue of the activity is computed by 26. The underlying statistical model can be expressed as: y = , + Ei, E(e) = 0, Var(e,) = a-,, Cov(E,,ej) = 0 (i = 1, 2, ..., n, j = 1, 2, ..., n) y = 5w, + e, E(e) = 0, Var(e) = (2, Cov(e,e,) = 0 (i = 1, 2, ..., n). 27. Hazell and Scandizzo, "Competitive Demand Structures under Risk," pp. 235-44. 28. Unpublished SUDENE data. 148 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL (5.8) Yi pi where yi = the yield of the ith crop in the activity andps = the base-year price of crop i. A new revenue series is then generated by (5.9) rt = wi ri, where rt = the new revenue series (on an activity basis); and ri, = the single-crop series. Two mean revenues are then generated by (5.10) F*= r/T and Y. = yipi. t i Then the adjusted revenue deviation series is generated by r* r** (5.11) d,t= - F**. r* The above procedure maintains the proportionality of the original series, dtl r** = (r - r*)/f. It assumes, first, that factors causing revenues to vary operate homogeneously (that is, there is a proportional effect with respect to some expected mean revenue); and, second,that revenue varies according to the crop mixture of an activity and not according to the activity itself. The risk-aversion parameters (phi) were first posited intuitivelv; small- scale farmers facing the possibility of starvation in a bad vear were given a high phi of 2.5 and large-scale farmers, a low phi of 0 to 0. 5. Values of phi in this range were confirmed by the results of a special study of 200 farmers and sharecroppers.2' The risk parameters finally used were determined through sensitivity tests of the phi's in solutions of individual farm models. TIhat is, the posited values for phi were varied upward and downward to see if these variations would greatly change the farm model's ability to repro- duce observed farm plans for the different farm types. The results of these tests suggested that, for estates in the West, a phi of 1.0 was more appropri- ate than 0.5 and that, throughout the East, for all practical purposes, risk could be ignored (phi = 0), because of the high reliabilitv of rainfall there. Otherwise, the tests revealed no basis for rejecting the phi's originally posited. 29. SeeJohn L. Dillon and Pasquale L. Scandizzo, "Risk Attitudes of Subsistence Farmers in Northeast Brazil: A Sampling Approach," AmericanJournal of Agricultural Economics, vol. 6, no. 3 (August 1978), pp. 425-35. STRUCTURE OF THE PROGRAMMING MODEL 149 Demand Activity Coefficients Following Duloy and Norton, the demand activity coefficients are formed from the base-year prices and quantities marketed and from price elasticities. )0 The procedure adopted is straightforward except for the price elasticities, for which direct estimates were lacking. These were calculated from Engel elasticities and budget shares, using a procedure derived by Feder and Scandizzo:3' ,,i = -Ej(o - I - otEi and = - aj ( 1+ Ej/w), where E, = Engel elasticity ,qii = own-price elasticity of the ith good Tlij = cross-price elasticity ai = budget share of the ith good of the average consumer of Northeastern products (o = Frisch parameter. 30. Duloy and Norton, "Prices and Incomes," pp. 591-600. 31. See Gershon Feder and P. L. Scaridizzo, "A Two-Region Multisectoral Model for Brazil" (World Bank, Development Research Center, 1977; processed). The budget shares are taken from studies by the United Nations Food and Agriculture Organization and the Federal University of Pernambuco. 6 Results from the Base Solution B3efore the model can be used to simulate the effects of policy interventions, its results for the base year, 1973, must be validated against those obtained from the tabular and cross-sectional analyses. This "base case" version of the model can also be used to investigate the structural and resource-related constraints on development in the region. As will become clear in the course of this chapter, the validation process itself reveals useful insights into the workings of the model. When these insights are examined in the light of the survey data, results from the cross-sectional analysis, and other prior information, they often lead to new hypotheses and explanations of farm behavior. Validation Considerations For economic planning models, the rather small body of literature on validation criteria and testing procedures permits only three general conclu- sions: that testing and validation are important; that the validation criteria must depend on the uses of the model; and that the validation procedure must instill confidence in the model, on the part of both the analyst and the user of the results. Though there are no rules for accepting or rejecting a model of the type used in this study, which is designed to explore a wide range of hypothetical situations for which no benchmarks exist, it is worth- while briefly to review some precedents. Nugent explored the possibility of validation tests in the context of a multisector, multiperiod linear programming model of the Greek economy.' In that study, he indicated three reasons a model may not 1. Jeffrey B. Nugent, "Linear Programming Models for National Planning: Demonstration of a Testing Procedure," Econometrica, vol. 38, no. 6 (November 1970), pp. 831-55. 150 RESULTS FROM THE BASE SOLUTION 151 perfectly simulate an actual economy: The model's constraint set may have been inaccurately specified; the underlying structure of the economy may be incorrectly represented (functionally or numerically); and it may be incorrect to assume that agents are strict optimizers with a singularity of purpose-either because some of their behavior is nonoptimal or because the objective function does not capture all their objectives. Departures from simple optimization behavior remain exceedingly difficult to handle in a linear programming framework. Where the observed production and price levels differed from "optimal" levels, Nugent used regression analysis to distinguish the differences owing to incorrect representation of the under- lying market structure (using tests of the model's structure and objective function) from those owing to omissions of variables from the model. Kutcher proposed and undertook several validation tests for an agricultural model of Mexico's Pacific Northwest region.2 These tests, however, were applicable only to the production/technology part of the model and were undertaken primarily because, unlike in the present study, no primary data on technology were at hand. Egbert and Kim, in a study of Portuguese agriculture, employed Theil U2 coefficients to test their model's ability to simulate aggregate production levels before conducting aggregation experi- ments and using the model to indicate areas of priority for investment.3 Although this background is helpful, it is not sufficient to define an appropriate rigorous testing procedure, particularly in the present case where farm-level behavior is embedded in a sectorwide, market-clearing framework. Our approach will be, first, to compare the model simulations of aggregate production, output prices, and levels of employment with the available survey and census data and then to undertake sensitivity analysis, varying the model parameters, to investigate the possible sources of bias. We then examine whether the bias can be traced to the simulation of particular farm types. 2. Gary P. Kutcher, "Agricultural Planning at the Regional Level: A Programming Study of Mexico's Pacific Northwest" (Ph.D. dissertation, University of Maryland, 1972). 3. Alvin C. Egbert and Hyung M. Kim, A Developmental Modelfor the Agricultural Sector of Portugal, World Bank Staff Occasional Paper no. 20 (Baltimore, Md.: Johns Hopkins Uni- versity Press, 1975). The model and its validation are also described, in condensed form, in "Analysis of Aggregation Errors in Linear Programming Planning Models," American Journal of Agricultural Economics, vol. 57, no. 2 (May 1975), pp. 292-301. Coincidentally, Leuthold, in the same issue, argued that such uses of Theil coefficients are inappropriate and are likely to give misleading results. His alternative, however, is to use "more sophisticated techniques." Raymond M. Leuthold, "On the Use of Theil's Inequality Coefficients," American Journal of Agricultural Economics, vol. 57, no. 2 (May 1975), pp. 344-46. 152 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-1. Alodel Performance: Crop Production and Prices Production' Pricesb Demand QO QS pO PS price Crop Observed Simulated Observed Simulated elasticity Cotton moco 427,531 575,246 2.10 1.60 -0.4, - - Cotton herbaceo 214,367 290,005 2.00 1.50 -0.4, - - Cotton verdao 120,127 161,348 1.90 1.40 -0.4, - - Rice 308,767 318,711 0.76 0.68 -0.3 Bananas 12,156 13,462 3.80 1.80 -0.2 Sweet potatoes 14,225 14,437 0.23 0.22 -0.3 Babacu 621,852 619,814 1.00 1.00 -x Cacao 203,098 208,896 5.00 5.00 - Cashews 11,002 10,765 1.00 1.00 -X Sugarcane 35,488 42,809 36.00 17.50 -0.4 Carnauba 836,746 852,652 0.80 0.80 - x Coconuts 4,544 5,314 40.00 14.18 -0.27 Beans de arranca 68,901 83,605 2.50 0.52 -0.26 Beans de corda 126,006 128,513 1.60 1.49 -0.26 Oranges 8,600 10,094 10.00 5.60 -0.4 Manioc 3,546,204 3,847,328 0.18 0.09 -0.17 Corn 632,123 635,521 0.51 0.49 -0.14 Castor beans 23,428 23,449 1.50 1.50 - Sisal 103,766 83,095 1.00 1.00 -t Note: IQ0P =Cr$7,175,456 IQ' Po = Cr$ 8,156,485 YQOP' = Cr$ 5,450,249 YQ' P0 = Cr$6,074,674 a. In metric tons, except for sugarcane, which is in thousands of metric tons, and coconuts and oranges, which are in thousands of fruits. b. In cruzeiros per kilogram, except for sugarcane, which is in cruzeiros per metric ton, and coconuts and oranges, which are in cruzeiros per hundred fruit. Production and Prices Because the model's full solution involves the summing of solutions for five zones and three to four farm types, depending on the zone, our first concern is whether this summation yields realistic totals for production levels and the associated (endogenous) market prices. The simulated pro- duction totals by crop and average Northeast-wide prices are reported in Table 6-1.4 The simulated gross value of production of all crops, at 4. Weighted average prices are reported because for most crops and livestock there are price differentials based on distances and access to markets. RESULTS FROM THE BASE SOLUTION 153 observed prices, is Cr$8,157 million, or 13.67 percent more than the observed total at observed prices. The crops in the table are grouped into three categories: the price- endogenous food crops that are, according to the best information available, consumed almost entirely within the Northeast; the cotton varieties that face both a price-responsive domestic demand and a perfectly elastic export demand for their processed products; and the long-cycle crops, which face either price-endogenous or perfectly elastic export demands but whose production is subject to stock constraints. For most of the food crops, the simulated production levels are close to, though slightly above, the observed levels. The total simulated value of these crops, at observed prices, is Cr$3,274 million, compared with an observed value of Cr$3,078 million, a discrepancy of 6.4 percent. The only dramatic difference in this group is for sugar production, the simulated value of which is 20 percent higher than that observed, even though this crop is generally considered to have long lost its comparative advantage in the Northeast. Most of the overestimation of sugar production occurs in the Agreste, not in the Zona da Mata. The simulated production of all varieties of cotton is about one-third greater than that observed; indeed, the simulated total of just over I million tons is about 10 percent greater than production in any of the previous ten years. (Although from 1962 to 1972 there were moderate increases in the area under cotton, output virtually stagnated because of declines in yields.) There are several possible reasons for this discrepancy. First, a static model cannot account for historical considerations: Farmers may have expected the previous declines in yields to continue and hence have planted less than the optimal (in a comparative-static sense) number of hectares. Second, cotton production may be subject to constraints not considered in the model. (Processing capacity, constraints on which were excluded from the model, could be a limiting factor; but knowledgeable sources have assured us that sufficient capacity exists to handle the volume of production simu- lated, and capacity expansion has historically kept pace with output growth.) Either of these explanations may be valid to some degree, but the principal reason, discussed more fully below, is that the fazendas in the Sertao, where the bulk of cotton is produced, use labor and land far more intensively in the simulations than they are observed to do. The production levels of the constrained long-cycle crops are, of course, generally extremely close to those observed. The deviations are due either to numerical (round-off) errors in tabulating the survey data and organizing them for linear programming, or to the use of stocks of trees, by area, as the constraint on production, rather than a simple limitation on the volume of output. 154 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-2. Regression Results: Goodness-of-Fit Tests for Production Volumes of Price-Endogenous Crops Simulated hy the Model Parameter a Parameter b (95 percent (95 percent Zone confidence interval) confidence interval) R2 West -0.247 1.012 (- 1.010, 0.515) (0.943, 1.082) 0.998 Sertao -0.142 0.998 (-0.584,0.299) (0.958, 1.037) 0.997 Southeast 0.155 0.983 (- 1.618, 1.928) (0.828, 1.138) 0.993 East 0.207 0.979 (-1.183, 1.597) (0.817, 1.142) 0.992 Agreste -0.117 1.000 (-1.134,0.899) (0.904, 1.097) 0.993 Northeast 0.058 0.985 (-0.208,0.324) (0.960, 1.010) 0.995 The fact that for almost every price-endogenous crop the simulated volume of production is above the observed requires closer inspection. As an alternative to Theil inequality coefficients-which may be inappropriate and are also difficult to interpret-and simple correlation tests-which would not capture systematic deviations that may be present here-we use the following regression test: log QS = a + b log Q0, where Q5 is the simulated volume of production of a given crop at observed price levels, Q° the observed production, and a and b are parameters to be estimated by regression. If, apart from random error, the model perfectly simulated the production levels, a, the intercept, should not differ from unity. The log transformation is used to abstract the b coefficient (though not the constant) from the scale and unit differences among the crops. The R2 from the regression will indicate the goodness of fit. The divergences in production levels across farm types are discussed later, but first it is worth investigating whether there are zonal patterns. The results of these regressions for each zone and the entire Northeast are shown in Table 6-2. For all zones and for the Northeast as a whole, a is not significantlv different from zero and b is not significantly different from unitv at the 95 percent level. As with most crops, the simulated value of output from livestock is higher than that observed. Although livestock production faces perfectly RESULTS FROM THE BASE SOLUTION 155 elastic demands in the model and no restraints other than those on the production factors (pasture, labor, credit for purchased inputs) and hence might be expected to dominate a solution of this type of model, the simulated herd sizes are within 20 percent of those observed and the value of production is only 18.3 percent greater. About 60 percent of the total difference between the observed and simulated livestock output occurs in the Southeast, where the model nearly doubles the existing herd sizes. Interestingly, knowledgeable sources indicate that since 1973 stock raising has indeed shown rapid growth in this region of Bahia. Thus, the model may have captured a latent comparative advantage at about the same time producers began to expand. The volume of total crop and animal production simulated by the model is only 15 percent larger than that observed, and its composition is very similar. The survey tabulations indicate that 33 percent of gross output originates from animal sources. If all outputs in the model are valued at observed prices, livestock's share is slightly more than 28 percent; when these outputs are valued at model (endogenous) prices, its share is about 34 percent. The model gives a much less satisfactory representation of prices than of production volumes. Indeed, as Table 6-1 shows, most crops with down- ward-sloping demand functions have equilibrium prices significantly be- low those observed. The exceptions are beans de corda, corn, rice, and sweet potatoes. The simulated prices of several crops (bananas, beans de arranca, coconut, manioc, and sugar) are at least 50 percent below their observed prices. For all crops, including those with fixed prices, the Las- peyres price index is 76, and the Paasche index 74, the observed price level being 100. For the price-endogenous crops alone, both indexes are about 60. A regression test for prices, similar to that for output volumes reported in Table 6-2, failed at the 95 percent confidence level for the Sertao and for the Northeast as a whole and gave positive results for these two cases only when the confidence level was lowered to 90 percent. These prices are of concern inasmuch as they correspond to marginal costs of production, averaged over zones and farm types. Thus, if we have confidence in the model's treatment of production technology and in the other input parameters estimated from the survey data, the results indicate that for several important products it would pay to expand production up to the point where market prices have fallen to 50 percent of present prices or lower. Either there are substantial disincentives to such expansion, or the observed operations of many farms are far below economically optimal levels. We return to these questions later in this chapter. 1.56 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-3. Persons Occupied in Agriculture: Model Results South- AVorth- Type of worker West Sertao east East Agreste east Farmers and family workers 309.435 827,010 173,747 52,168 388,919 1,751,279 Sharecroppers and family workers - 969,714 - - - 969,714 Permanent workers 126,756 3,837 17,938 63,485 45,213 257,229 Temporary workers 426,608 147,318 643,613 66,608 235,994 1,520,141 Total 862,799 1,947,879 835,248 182,261 670,120 4,498,363 - Not applicable. Note: The derivation of these estimates is described in the text. Employment and Labor Utilization The model provides several alternative measures of emplovment and labor utilization. The number of persons occupied in agriculture (the definition used by IBGE for the 1970 agricultural census) is one such mea- sure. Table 6-3 reports the simulated totals by zone for the four broad categories of agents: farmers and their family workers, sharecroppers and their familv workers, permanent workers, and temporarv workers. For these totals an occupied person is defined as follows. Because, in the model solution, all farms are operational to some degree, all the farm family workers are assumed to be occupied (though not necessarilv fully em- ploved). Thus, the farm family entries in Table 6-3 are simply the sum of the number of farms of each type multiplied by the corresponding survey estimates of family workers; for sharecroppers, the number of share- cropped plots simulated by the model is multiplied bv the average size of the sharecropper's occupied family (2.2 persons). The number of occupied permanent workers is measured as the number of annual contracts (whether or not these workers were fully used throughout the vear), and the number of'temporary workers as the peak monthly hirings in each zone. Thus, the totals in Table 6-3 are in fact upper bounds on the amount of labor actively engaged. Even so, the Northeast total of about 4.5 million persons occupied in agriculture derived in this wav contrasts sharply with labor forcc estimates in the neighborhood of 6 million (see Chapter 2) and indicates unemploy- ment rates of 25 percent or more. Almost 40 percent of those occupied are farmers and their family workers (1.75 million), and only 27 percent of the remaining occupied persons have permanent or share-tenant contracts. The RESULTS FROM THE BASE SOLUTION 157 Table 6-4. Labor Requirements: Model Results (man-years) South- Nortb- Labor West Sertao east East Agreste east Farmers and family workers 174,571 487,111 148,605 34,920 292,349 1,137,557 Sharecroppers and family workers - 420,688 - - - 420,688 Permanent workers 126,756 3,837 17,938 63,485 45,213 257,229 Temporary workers 96,599 23,061 149,535 28,368 61,898 359,463 Total (1) 397,927 934,698 316,078 126,773 399,461 2,174,938 Survey estimates (2) 336,434 754,751 222,570 77,562 376,861 1,768,178 Percentage deviation(1)/(2) +18.3 +23.8 +42.0 +63.4 +6.0 +23.0 - Not applicable. Note: The derivation of these estimates is described in the text. other 3 3 percent of the occupied persons are temporary workers receiving a minimum of one month's employment a year. Because the typical employed temporary worker receives an average of two to three months' employment a year and most family farms and sharecropped plots exhaust half or less of the family's labor, it appears that well over half the agricultural labor force is either unemployed or unquestionably underenmployed, if we ignore nonagricultural employment. To derive more precise measures of un- and underemplovment from the model, actual labor requirements were tabulated from the crop and live- stock production activities. The results are shown in Table 6-4 along with the survey estimates of labor requirements reproduced from Table 2-4. For all zones and the Northeast as a whole, the model uses labor more intensively than was recorded by the survey. This squares with the model's tendency to overestimate production: The simulated labor requirements are 23 percent greater, whereas production is 15 percent greater. The discrepancies between actual and simulated labor use are largest in the Southeast, which is dominated by cacao plantations (42 percent), and in the East, which is dominated by sugar plantations (63 percent), whereas they are smallest in the Agreste. It seems clear that there is considerable scope for more intense employment even under existing farming systems. The labor requirements simulated by the model are substantially higher than in actual practice but may be used to investigate unemployment questions further if it is borne in mind that the employment rates they suggest are likely to be upper bounds. 158 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-5. Employment Rates, by Month and Zone: Model Results (percent) South- North- Month West Sertao east East Agreste east January 53.90 63.70 66.15 62.48 47.78 58.80 February 46.55 61.25 37.98 75.95 44.10 51.45 March 25.73 72.28 24.50 56.35 40.43 47.78 April 19.60 61.25 14.70 60.03 40.43 40.43 May 18.38 52.68 15.93 44.10 44.10 36.75 June 20.83 40.43 20.83 75.95 53.90 36.75 July 26.95 26.95 20.83 51.45 50.23 30.63 August 25.73 40.43 23.28 90.65 53.90 37.98 September 39.20 49.00 22.05 77.18 35.53 40.43 October 37.98 39.20 26.95 95.55 36.75 37.98 November 69.83 58.80 88.20 98.00 69.83 69.83 December 78.40 46.55 100.45 89.43 63.70 67.38 Annual average 39.20 51.45 39.20 73.50 49.00 46.55 Source: Based on survey estimates of total labor available reported in Table 2-4. Table 6-4 shows that full-time equivalent employment is generated for only 2.17 million, or 48 percent, of the 4.5 million persons occupied. Of the actual labor absorbed, 52 percent is supplied by farmers and their families, with sharecroppers and permanent and temporary workers supplying 19 percent, 12 percent, and 17 percent, respectively. For a nonpropertied labor force approaching 5 million, only slightly more than I million man- years of employment are generated. A good part of the underemployment problem can be traced to the seasonality of agricultural labor requirements. Table 6-5 reports the monthly use of labor by zone from the model results as a percentage of the labor force available. Of the nearly 6 million available in the work force, only 31 percent are engaged during July, and no more than 70 percent are ever engaged (during November). This variation is partly due to the model's exclusion of activities not directly related to cropping and stock raising (such as maintenance and postharvest processing of some products), but it nevertheless stresses the importance of seasonality in the underem- ployment problem. The likelihood that workers will be idle or assigned to marginal tasks for several months during the year probably also explains, in part, estate owners' reluctance to hire workers on annual contracts. Thus far, comparisons with the survey data have shown that the model slightly overestimates production, as might be expected of an optimization model; that it more strongly underestimates prices; and that it overesti- RESULrS FROM THE BASE SOLUTION 159 mates the intensitv of labor use. The following three sections examine some sources of bias and test the model's sensitivity to changes in parameters. Family Reservation Wage The model's deviation in labor use may be due partly to an incorrect assumption about the family reservation wage rate. This parameter deter- mines the minimum return to family labor, below which the farm would not be operated. Its value, in relation to the wages that must be paid to nonfamily workers, will in general influence the combination of the various labor sources employed on the farm; other things being equal, the higher (lower) the family's reservation wage, the more (less) attractive it will be for them to employ other workers. The reservation wage used in the nmodel, Cr$3 a day, is about half the average daily wage of hired laborers. The probabilitv that a farm family worker could find alternative agricultural employment is about 0.5. In varying the reservation wage to test the model's sensitivitv, little leewav was possible: Cr$2 a day is probably the bare subsistence minimum, and in the West and the Sertao even permanent laborers earn little more than Cr$4 a day. Table 6-6 reports the effects on the model solutions of varying the reservation wage. These sensitivity tests reveal that aggregate output, income, and emplov- ment rise somewhat but are largely insensitive to the reservation wage. Though the use of Cr$2 as the wage rate gives the most accurate representa- tion of these aggregates, it causes some misrepresentation of the composi- tion of employment, because family labor is substituted for permanent hired labor and, in particular, sharecropper labor. Indeed, sharecroppers all but drop out of this solution, leading us to reject the Cr$2 figure on these grounds alone. The Cr$4 figure must also be rejected because it leads to an unrealistically high use of sharecroppers and permanent workers and fur- ther departs from the output, price, and total emplovment levels of the Cr$3 case. Though this test does not establish Cr$3 as the correct reserva- tion wage, it allows us to conclude that it is more appropriate than Cr$2 or Cr$4. The increases in aggregate output and employment shown in Table 6-6 are in fact confined to the Sertao and can be explained bv the model's treatment of sharecropping in that zone: The landowner is assumed to be the decisionmaker (the producer in the objective function), with the proviso that the production plan he selects for each sharecropped plot will permit a minimum consumption bundle for the sharecropper's family, who supply 160 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-6. Effects on Model Results of Variations in the Family Labor Reservation Wage (percentage changes from Cr$3 a day) Category At Cr$2 a day At Cr$4 a day Sector income - 1.9 + t.6 Price index + 1.1 - 1.3 Gross output Livestock - 1.7 + 1.5 Crop - 0.4 + 1.6 Total - 1.1 + 1.6 Labor use Farm family + 8.4 - 15.8 Sharecroppers -81.6 + 51.9 Temporary + 5.6 - 5.5 Permanent -14.4 + 50.1 Total - 2.8 + 3.8 the labor to produce this bundle for themselves at no cost to the landowner. Thus, any labor employed for sharecropper family consumption is costless to the model, and because this production for own consumption does not enter the market it does not provide utility to the model in the form of consumers' surplus. Under the lower reservation wages, when the land- owner produces a good portion of the cash crop, cotton, he produces cotton alone because the additional labor required for interplanting food crops is not warranted. When the family reservation wage rate is increased, how- ever, and sharecroppers are substituted for family labor, the consumption requirements of the sharecropper, still of no concern to the landowner, induce greater labor use and more output. Price Elasticities of Demand The model's underestimation of food crop prices may be due partly to the low price elasticities of demand used for these crops, particularly when combined with the model's overestimation of production. These elasticities were shown in Table 6-1. They are short-run elasticities, because the base case of the model refers to a single point in time. Though they were estimated as rigorously as possible (see Chapter 5, Appendix C) it is ifnportant, for several reasons, to assess how strongly they affect the model results. First, the model results suggest that Northeastern agriculture is RESULTS FROM THE BASE SOLUTION 161 constrained by demand-a hypothesis on which the demand elasticities obviously have a direct bearing. Second, the magnitudes of the elasticities are particularly important in the evaluation of simulated policies to shift supply-for example, technical progress, employment subsidies, and poli- cies to promote land use. The structure of the model is such that unless demand is infinitely elastic (as it is assumed to be for export crops) increases in supply will depress prices, benefiting consumers but potentially harming farmers. The distribution of the benefits from supply-augmenting in- terventions thus depends partly on the magnitude of the demand elastici- ties. Finally, there is room to argue that the elasticities that have been estimated are too low in absolute value for a framework that will be used primarily for comparative-static analysis of equilibrium points several years apart. This topic will be addressed more fully in Chapter 7, but it should be pointed out here that longer-term demand schedules are likely to be more elastic (higher in absolute value). Higher elasticities yield less steep demand functions and-given the supply functions in the base case of the model-higher output levels and higher prices. In Figure 6-1, DIDI is a typical demand schedule under the rather low elasticities used in the base case, and SS is a typical supply function. Because DID, and SS intersect to the right of the base period quantity Q° (as is the case for most of the food crops considered), the new solution price P1 is lower than the base price P°. Increasing the elasticity (D2D2) has the effect of rotating the demand schedule about the point (Po, Q°), not about the price intercept of DID,, raising both quantity (to Q2) and price (to p2). It can also be surmised from Figure 6-1 that the opposite directional changes would result from an increase in elasticity if in the base case SS intersected DID, to the left of Q° and that no changes would occur if all supply functions intersected demand functions exactly at Q°. A sensitivity test was undertaken in which the demand elasticities were doubled, with all other aspects of the base case kept constant. The results of this test are not reported in their entirety, simply because there were no dramatic changes. Crop output rose by 3.2 percent, and livestock output fell by 0.8 percent, as some medium-quality land was shifted from pasture use to crops. Employment rose by only 1.3 percent, and incomes of all agents rose by 0.8 percent. Such magnitudes demonstrate that the key lessons from the base solution do not depend entirely on the demand elasticities, although the latter are of obvious importance for the distribu- tion of gains between consumers and producers under different policy conditions. The model's representation of prices and quantities cannot be made more accurate by manipulating the elasticities, because a closer fit of prices will correspond to a poorer fit of quantities. It remains, then, to 162 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Figure 6-1. Effects of More Elastic Demand Schedules: Model Structure Price D, Quantity 20 QI Q2 investigate the causes of the model's overestimation of production and resource use. Optimality Assumptions Much of the overestimation of production and the underestimation of product prices that is directly related to it may lie in the optimality assumptions embedded in the linear programming framework. In virtually all known examples based on reliable production and resource-endowment data, linear programming models are more "efficient" than the reality they are designed to depict. The cross-sectional analysis of the performance of farms of different sizes, reported in Chapter 4, strongly suggested that large-scale farmers do RESULTS FROM THE BASE SOLUTION 163 Table 6-7. Output Generation and Labor Use, by Farm Type: Survey Data and Model Results (cruzeiros of gross output per hectare of farm size; man-years of labor per hectare of farm size) South- North- Category West Sertao east East Agreste east Family farms Output per hectare Survey averages 231 246 280 417 506 344 Model results 140 124 347 375 480 288 (-10 percent) Labor per hectare Survey averages 0.090 0.085 0.088 0.102 0.130 0.102 Model results 0.042 0.036 0.095 0.108 0.175 0.095 (- 7 percent) Medium-size farms Output per hectare Survey averages 85 157 147 376 273 162 Model results 79 176 283 348 180 175 (+8percent) Labor per hectare Survey averages 0.024 0.032 0.025 0.056 0.046 0.032 Model results 0.016 0.036 0.046 0.074 0.052 0.036 (+ 12 percent) Estates Output per hectare Survey averages 24 69 68 75 153 63 Model results 50 84 85 150 103 83 (+ 32 percent) Labor per hectare Survey averages 0.004 0.010 0.008 0.101 0.028 0.010 Model results 0.015 0.020 0.019 0.026 0.023 0.019 (+90percent) not equate marginal costs and prices; that is, they do not maximize profits with respect to their productive land resources. Given the attention paid to the complexities of Northeastern agriculture in the design of the model and the painstaking analysis that lies behind the technical parameters used, the estate subsector is the most obvious of the remaining possible sources of bias. The comparisons in Table 6-7 of survey data and results from the model show that the model's overestimation of production and employment is indeed almost entirely traceable to its simulation of the estate subsector; on family and medium-size farms aggregate output and employment are much more precisely simulated. In all zones save the Agreste it would be optimal for the estates to employ far more labor and produce substantially more than they do currently. If they were to employ their fixed and variable resources so as to equate marginal costs with market prices (taking account 164 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL of risk penalties, and so forth), they would on average employ 90 percent more labor and attain about one-third more production. Table 6-7 also shows that the simulated characteristics of family and medium-size farms are biased in opposite directions, or that the behavior of these farms departs from the optimal in different ways. First, the typical family farm not only makes fuller use of family labor and produces more per hectare than other types of farm but actually uses more labor and produces more output (by about 10 percent in both cases) than the model indicates would be optimal. Second, the medium-size farm uses less labor than would be optimal and produces slightly below optimal levels (again in the range of 10 percent). These identified biases do not render the model inappropriate for most conceivable policy and project uses, as long as they are kept in mind in reviewing the results. The Pattern of Constraints Part of the power of linear programming lies in its ability to identify constraining factors and to provide marginal valuations (shadow prices) of those that are binding. The present model permits us to examine the constraints on output and the employment of resources at two levels: the farm and the sector. Constraints at the Farm Level The constraints at the farm level are family labor, which may be aug- mented by hiring permanent or temporary labor (and sharecroppers in the Sertao); land of three types (good and medium-quality cropland and pas- ture); working capital; and the existing stock of tree crops. Family labor is seldom a binding constraint on family farms, but on large farms, once family labor is exhausted, the expense of hiring additional workers often leaves gross margins on crops too small to warrant the cultivation of the whole farm. This pattern of behavior shows up vividly in the land constraints; see Table 6-8, which reports shadow prices for good and medium-quality cropland and native pasture. For cropland of both qualities, family farms generally show much higher shadow prices. An exception is the Southeast, where cropland is not exhausted for any of the farm types; all the excess is used as either planted or native pasture. Within zones, the differences in land shadow prices are the counterpart of farm employment patterns. Land RESULTS FROM THE BASE SOLUTION 165 Table 6-8. Shadow Prices of Land, by Farm Type and Zone: Model Results (cruzeiros per hectare) South- Farm type West Sertao east East Agreste Good cropland Family farms 83 90 16 850 531 Medium-size farms 6 42 16 511 44 Estates 10 58 16 477 28 Medium quality cropland Family farms 83 87 16 717 288 Medium-size farms 6 42 16 378 42 Estates 6 53 16 344 28 Native pasture Family farms 6 42 16 - - Medium-size farms 6 42 16 67 28 Estates 6 35 16 72 28 -INot applicable. shadow prices are highest on the family farms, reflecting the abundance of idle family labor, and generally decline as farm size increases, in direct relation to the higher percentage of hired labor used.' The differences in shadow prices further indicate distortions in the land market. Why is a land rental market not operating when family farmers in, say, the Agreste would pay up to Cr$5 31 a year for the use of one hectare of good cropland while the fazendas are realizing an annual marginal return of only Cr$28? The across-zone differences in land shadow prices closely follow the pattern expected from consideration of location (access to markets), soil quality, and the volume and timing of rainfall. The shadow prices are highest in the East, which has abundant rainfall and lies near the major coastal cities; are next highest in the Agreste, just inland from the East and not subject to periodic droughts; and are lowest (apart from the Southeast) in the drought- prone and more remote Sertao and West. Another constraint at the farm level is the stock of previously planted or native long-cycle tree crops: babacu, cacao, carnauba, cashews, castor beans, coconut, and sisal. The model harvests the entire stock of all except 5. The decline in shadow prices as farm size increases is not always monotonic: in several cases the largest farms have land shadow prices slightly greater than the medium-size farms. This result is due to different technologies employed in the livestock activities. Because average labor is more expensive for large farms, they tend to use labor-saving techniques, which results in a higher opportunity value of land. 166 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-9. Shadow Prices of Tree-crop Stocks, by Zone: Model Results South- Crop West Sertao east East Agreste Babacu 399 - - - - Cacao - - 3,568 - - Cashews 92 12 3 - - - Carnauba 510 3,782 - - - Coconuts - - - 1,008 - Castor beans - 333 96 - - Sisal - - - - 0 -Indicates that the crop is not grown in the particular zone. sisal, and the shadow prices on the stock constraints may be interpreted as the net income that would accrue from an additional hectare of the crop under consideration. Table 6-9 reports these shadow prices averaged over farm types and land qualities. In all but two of the farm models the simulated use of working capital is less than the actual use, as measured by farm expenditures taken from the survey data. Credit is a binding constraint for the livestock fazendas in the West and the cacao plantations in the Southeast. This finding appears to be completely unrealistic, inasmuch as these plantations have vastly larger capital stocks and vastly higher income levels (both actual and simulated) than any other type of farm. The shadow price of short-term capital for the fazendas in the West indicates that they would borrow at a real rate of interest of up to 30 percent. There are two possible reasons the other farm types, as simulated by the model, consistently use far less working capital than the expenditures recorded by the survey. Though it is likely that the model has failed to capture fully items such as maintenance and construc- tion materials that require working capital, it is also quite probable that farmers' responses during the survey did not make a sharp enough distinc- tion between the working capital required for agricultural production and that required for family consumption and services. Indeed, the discrepan- cies between actual and simulated expenditures are largest for the family farms. These results do not allow us to conclude that credit for working capital is not a constraining factor in the Northeast; we may conclude only that the model is not an appropriate tool for investigating the credit prob- lem. The working capital constraint was therefore transformed into an accounting row for crudely monitoring those simulations that might pose severe cash flow problems. The constraints that are binding at the farm level can now be summarized quite simply. For all farms in the applicable zones, the stock of most RESULTS FROM THE BASE SOLUTION 167 exportable tree crops is a binding factor. For family farms, the overwhelm- ing constraint is land, particularly land suitable for crops. Medium-size farms appear to be constrained partly by land-though much less critically than are the family farms-and partly by the availability of relatively cheaper family labor. For large farms, with the possible exception of the livestock fazendas in the West, which may lack sufficient credit, there are no clearly evident constraints, either in reality or in the model's representa- tion. Constraints at the Sectoral Level Possible constraints at the sectoral level are labor, processing capacity, credit (already dismissed), and the level of market demand for the price- flexible crops. As was evident from the earlier discussion, the stock of available labor is never exhausted, but the supply of labor at wage rates low enough to induce the full utilization of land may be a binding constraint. Temporary workers' wages are probably already at the bare subsistence level,6 and those of permanent workers and sharecroppers are not far above that level. Processing constraints on cotton and sugar are not enforced, although they might become important for cotton if production expanded dramatically. Although the level of demand for price-endogenous crops is not a con- straint comparable to the availability of fixed resources, the model clearly shows it to be an important element of the stagnant low-level equilibrium. As was noted earlier, in all zones the prices of these crops are equated to marginal costs at levels that are far too low to induce full utilization of the available land and labor, and at levels of output and employment that imply extremely low income levels for the millions of landless and small farmers. The treatment of demand in the model is discussed at some length in the following chapters, but here we investigate whether export sales could relieve the demand constraints for crops not currently being exported. A simple exercise, reported in Table 6-10, reveals this is not likelv. Table 6-10 shows the orders of magnitude and the components of domes- tic resource cost (DRC) coefficients, which are computed as the ratios be- tween the shadow prices given by the model (base year solution) and data on weighted f.o.b. border prices, again for 1973. Clearly, without considera- tion of marketing margins, transport losses, or quality, and even with 6. The observed wage rate for temporary labor is at most Cr$10 per day of employment. The minimum subsistence-level income is about Cr$2 a day. If paid US$2.13 a day a worker would earn US$100 a year working 300 days. Because most temporary workers are engaged for only 50-60 days a year, Cr$10 a day is probably the effective subsistence wage. 168 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 6-10. Domestic Resource Cost Coefficients of Food Crops, 1973 Farm-levoel domestic Domestic Border resource Transport resource prices cost b cost C cost Domestic (U.S. dollars (cruzeiros (cruzeiros (cruzeiros resource per per per per cost Crop metric ton) kilogram) kilogram) kilogram) coefficientd Rice 126.62 0.68 0.40 1.08 1.15 Beans 202.53 1.50 0.25 1.75 1.16 Manioc 74.08 0.45 0.30 0.75 1.36 Corn 76.76 0.49 0.25 0.74 1.30 a. Instituto Brasileiro de Geografia e Estatistica, Anuario estatistico Go Brasil (Rio dejaneiro, 1975). b. From base-year solution of the model. c. Conjuntura economica (une 1974), p. 61, table xxxi. d. Based on a shadow exchange rate of Cr$7.44:US$1.00. The domestic resource cost coefficient is computed as the ratio between the shadow cost of foreign exchange and the cost of generating foreign exchange by producing export crops-indicated by the border prices of those crops. allowance for Brazilian currency overvaluation, Northeastern food crops do not show potential for export. In comparison, the DRC coefficients for the Northeast's traditional exports, apart from sugar, are all below or equal to one. 7 The market prospects of traditional exports as of 1973 were not particu- larly encouraging. The model results suggest that it would have paid to expand production of cacao and cotton, diversify out of sugar, and mod- erately expand beef production. For both cotton and cacao the long-term demand prospects were good. The lack of suitable land, however, would have severely limited the expansion of high-quality cotton production. With the prevailing pattern of land tenure and high subsidies for sugar production, financial incentives tended to favor beef and sugar production rather than cotton on large farms. Small-scale farmers, for their part, may have been discouraged from expanding cotton production by the margins exacted by the marketing intermediaries and the gins. For cacao, the land tenure and marketing situation would not have presented such problems, but prospects for growth were confined to small areas of Bahia and, according to government expansion plans, selected areas of the Amazon. 7. For example, the coefficients are 0.6 for cocoa, 0.8 for cotton, 1.0 for beef, and 1.2 for sugar. 7 Policy and Project Options for Northeastern Agriculture In this chapter the model is used to explore alternative ways of promoting growth in agricultural production, employment and incomes, and improve- ments in income distribution. Despite the wealth of detail the model contains, it should be emphasized at the outset that the model is not a tool for detailed planning. It is suitable for the simulation of broad policy or project options and can roughly estimate their effects on the agricultural sector, but it is not suitable for the detailed specification of policies or individual projects. Costs, in particular, can be made explicit in only a few cases. Furthermore, some of the simulated experiments may not have counterparts in any readily defined intervention or policy instrument, even though their effects may be of great interest. Exogenous shifts in demand for food crops, or accelerated technical progress, for example, are difficult to induce directly. Nonetheless, the nature and size of the consequential changes identified by the model can help provide a perspective for the analysis of possible indirect measures to achieve these ends, such as invest- ment in marketing and transport infrastructure, agronomic research, or improved extension services. The analytical framework used is comparative-static. Most, if not all, of the simulated changes would require considerable time to make their effects felt, and hence the model's base case is projected forward five years to 1978. This projection suggests how Northeastern agriculture might have changed over time without interventions and provides a reference point against which the results of experiments can be contrasted. Though the five-year time span is inevitably arbitrary, it is judged long enough for most adjust- 169 170 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL ment to be completed, but not long enough for the basic structure of the economy to be substantially altered.' After the changes in parameters necessary to make this projection, as well as the main features of the 1978 base solution, are described, five types of simulated interventions are analyzed, alone and in combination: promo- tion of technical progress, reduction of risk, labor policies, promotion of demand, and agrarian reform. The promotion of technical progress is selected because it is emphasized in current policy toward the Northeast, both in Brazil and in World Bank operations. Farmers' ability and willing- ness to bear risk have an important influence on existing production pat- terns in the Northeast; farmers' attitudes toward risk are also an important factor in the spread of technical progress. Labor policies are simulated because employment is probably the single most important problem in the rural Northeast. Demand-related programs, often ignored in sector analy- ses, are considered because the model has suggested the hypothesis that the low-level stagnation may be due to limited access to markets. Agrarian reform is considered because it has been strongly advocated by others and has been legislated and funded, but we feel its potential effects on the agricultural sector at large have not been adequately evaluated. The Agricultural Economy Projected to 1978 In projecting the model's base from 1973 to 1978, three types of changes must be taken into account: behavioral, technical, and demographic. Behavioral changes occur on the part of both consumers and producers. Consumers respond to growth in their incomes and to price changes. The treatment of demand shifts caused by income growth will be described later. For food crops, price elasticities of demand are assumed to be higher in the long run than in the short, primarily because consumers take time to adjust their tastes and habits to changes in relative prices.2 Empirical verification of this phenomenon is scant, though Houthakker and Taylor were able to estimate both short- and long-run elasticities for several 1. Duloy and Norton, in the CHAC study, chose a six-year horizon for their comparative- static analyses of Mexican agriculture, but they were not explicit about the reasons for this choice. See John H. Duloy and Roger D. Norton, "CHAC: A Programming Model of Mexican Agriculture" and "CHAC Results: Economic Alternatives for Mexican Agriculture," in Multi- Level Planning: Case Studies in Mexico, ed. Louis M. Goreux and Alan S. Manne (Amsterdam: North-Holland Publishing Company, 1973), chaps. IV.I and IV.3. 2. See George J. Stigler, The Theory of Price, rev. ed. (New York: Macmillan Company, 1952). POLICY AND PROJECT OPTIONS 171 products. 3 In many cases, they found long-run elasticities to be about twice as high in absolute value as short-run elasticities. For want of information to the contrary, we shall, as in the parametric exercise described in Chapter 6, take the price elasticities for the longer-run case as double those estimated for the base case. Producers are assumed to adjust their output patterns and the intensity and use of resources so as to continue to maximize their annual profits, that is, to continue to equate marginal costs (including the risk penalty costs) to prices for each commodity.4 Thus, for them, projections to the 1978 base require no structural or parameter changes to the model. Agricultural practices in the Northeast have stabilized over perhaps centuries, and there is no available evidence to suggest that they would change much in the course of five years. But though progress in irrigation, mechanization, and the like can thus be ignored in the model, two other changes affecting output must be considered: the time trend in the yields of many crops, and the changes in the stocks of long-cycle or tree crops. The IBGE data in Table 7-1 reveal that yields of several important crops declined in the decade before 1973. (The possible upward bias in the IBGE data, discussed in Chapter 3, probably does not distort the time trends of the variations in yields.) Several students of the Northeast suggest that this decline was due to the expansion of cultivated area to less productive land, unaccompanied by technical progress.' Although the expansion of culti- vated area was much faster than that of population or income, the declines in yields left gross production growing at about the same rate as population. Indeed, virtually identical changes occurred in overall crop production growth and population growth over the twenty years preceding 1973. Because the model can represent demand shifts between different points in time, as well as different land qualities that permit different crop yields, these effects are treated endogenously. Expansion of demand should induce a more than proportional expansion of the cultivated area and declines in average yields per hectare. 3. Hendrik S. Houthakker and Lester D. Tavlor, Consumer Demand in the United States: Analyses and Projections (Cambridge, Mass.: Harvard University Press, 1970). 4. To the extent that farms do not optimize over the variables we consider, the results will be misleading; if, indeed, many large farms are held as status symbols or inflation-hedging assets, the simulated effects of, say, wage subsidies will be overestimated. But since the base case shows large farms as more efficient than they actually are, the model will underestimate the efficiency gains from structural changes, such as land reform, affecting these farms. 5. See, for example, George F. Patrick, "Efeitos de programas alternativos do governo sobre a agricultura do Nordeste," Pesquisa eplanejamento economico, no. I (February 1974), pp. 43-82. 172 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-1. Time Trends in Yields and in Area Harvested, Principal Crops, 1961-63 to 1971-73 Crops 1961-63 1964-66 1967-69 1971-73 Yield (kilograms per hectare) Cotton 339 293 292 276 Rice 1,504 1,332 1,360 1,295 Sugar 41,068 42,897 44,441 44,898 Beans 539 495 572 548 Manioc 11,833 12,058 12,528 12,143 Corn 784 743 814 724 Area harvcested (thousands of hectares) Cotton 2,268 2,586 2,923 3,187 Rice 553 669 766 878 Sugar 562 566 604 698 Beans 1,022 1,247 1,548 1,638 Manioc 706 788 976 1,036 Corn 1,482 1,743 2,084 2,184 Note: Three-year averages tabulated from IBGE data; 1970, a drought year, is omitted. The area under long-cycle or tree crops is taken as exogenous to the model, and we assume that the time trends in the areas under these crops obtained from 1973 to 1978 are the same as were estimated by IBGE for 1968-7 3. Thus, for the 1978 solution the area under coconuts is assumed to have increased by 16.67 percent, that under oranges by 26.35 percent, and that under castor beans by 13.2 percent; and the area under all other long-cycle crops is assumed to be the same as in 1973. Demographic changes concern the expansion of population and related growth in the labor force and in consumer demand. World Bank estimates place the Northeast's rural population growth rate between 1973 and 1978 at 1.2 percent a year, net of migration, and we assume that the labor supply also grows at this rate. This assumption has little consequence for the structure of the model because no conceivable policy changes would result in a labor shortage within the time period considered. Nevertheless, it is necessary for assessing the extent of unemployment, on the assumption that labor force participation rates do not change. The additional workers are assumed to join the pool of landless labor available for permanent, tempo- rary, and sharecropping employment. The principal effects of population growth in the model are on product demand. Population and income growth and changes in income distribution will raise the demand for different commodities by different factors, depending on the characteristics of the consuming groups and the Engel elasticities assumed. An unpub- POLICY AND PROJECT OPTIONS 173 Table 7-2. Percentage Shifts in Demand for Price-Endogenous Crops, 1973-78 Crop Shift Cotton 21.71 Rice 18.04 Bananas 21.71 Sugar 17.03 Coconuts 21.71 Beans 15.87 Oranges 20.14 Manioc 14.20 Corn 15.49 lished study by the United Nations Food and Agriculture Organization and the Superintendency for Planning, Ministry of Agriculture, projects de- mand for the principal agricultural products of the Northeast on the basis of the income shares of different consuming groups, and these estimates imply the shifts reported in Table 7-2. In summary, the changes made to project the model to 1978 are as follows: Product demand functions are shifted rightward, depending on the growth of population and income and Engel elasticities, and are made more elastic; labor availability is increased in line with population growth; the stocks of some long-cycle tree crops are shifted according to changes observed in the previous five-year period. The 1978 Base Case Solution Table 7-3 compares key indicators at the sectoral level from the 1978 solution and the 1973 base. Farmers' and workers' incomes and employ- ment rise, but consumers appear to suffer: Over the five years, the increase in total output (14.9 percent) is slightly below that of population (16 percent), and real prices rise by 3.2 percent. The substantial growth in the value of crop output (24.3 percent) is largely confined to the exportable crops, production of which expands by 46.2 percent. Output of food crops grows by only 7.7 percent. This minimal growth in food crop production combined with a decline of 11.5 percent in livestock output implies that the per capita agricultural consumption of Northeasterners would have de- clined over the five years considered, unless imports from outside the Northeast had increased substantially. The income gains accruing to farmers are quite skewed. Family farms attain only a 2 percent increase, which is wholly due to the rise in the price 174 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-3. Output, Income, and Employment: 1973 and 1978 Base Cases Percentage change, Category 1973 1978 197311978 Gross output' Crops 8,965 11,146 +24.3 Livestock 3,180 2,813 - 11.5 Total 12,145 13,959 + 14.9 Price indexb 91.4 94.3 + 3.2 Incomec Family farms 749 764 + 2.0 Medium-size farms 4,991 5,707 + 14.4 Estates 1,161 1,287 + 10.9 Sharecroppers 633 935 + 47.7 Temporary workers 962 923 - 4.1 Permanent workers 454 776 + 70.9 Total 8,949 10,392 + 16.1 Employment d Farm family 1,138 1,222 + 7.4 Sharecroppers 421 630 +49.6 Hired workers 617 770 + 24.8 Total 2,175 2,622 +20.6 Peak month 3,259 3,386 + 3.9 Note: The nurmberof farms, and their family workers, rernains constant from 1973 to 1978; the number of nonpropertied workers has increased by 6 percent. a. Thousands of cruzeiros in constant prices. b. Laspeyres index, 1973 observed price = 100. c. Money incomes in thousands of cruzeiros, at model-endogenous prices. d. Thousands of man-years; peak month in thousands of men. level. Without technological progress, small-scale farmers simply do not have enough land to expand their production to take advantage of expand- ing product markets. At 14.4 percent, the medium-size farm sector shows the highest growth in incomes. Because this rate closely compares with the growth in total output, the model suggests that this subsector is the most flexible and responsive to expanding markets. Although the estates show an income increase of 10.9 percent, virtually all this growth occurs in the Agreste and East in response to the shift in demand for cotton and sugar. In the other zones, the West, Sertao, and Southeast (cacao region excepted), output, employment, and income on estates grow very little. This is probably because of the inferior and less income-elastic crops grown there, and the general remoteness of the region from expanding markets. Employment and incomes of nontenured labor improve substantially. Sharecroppers' employment grows by 49.6 percent, although their per POLICY AND PROJECT OPTIONS 175 capita incomes decline by about 2 percent. The growth in sharecropper employment occurs almost entirely on medium-size farms in the Sertao where the average number of sharecroppers per farm doubles, almost perfectly in accordance with the expansion of cotton production (up 53 percent in the Sertao from the 1973 solution). The 70 percent growth in the employment and incomes of permanent laborers is greater than for any other population group. This growth occurs in all zones but is greatest in the Agreste where it triples. Temporary labor employment decreases slightly. Overall, the simulated employment situation in 1978 remains poor, however. The number of "fully" employed, nontenured workers (share- croppers and permanent workers) increases by 380,000 over the five-year period, but temporary employment declines and the rural labor force grows by about 350,000. Thus, in spite of substantial emigration from the North- east and in spite of the model's assumption that farmers optimize, these results suggest that the net increase in employment over the five years would have affected only about I percent of the un- and underemployed. A highly important first conclusion is that the employment problem will not be alleviated through time alone. The 1978 solution of the model reflects a continuing expansion of culti- vated area as revealed by the census data for preceding years, but it does not indicate a growing pressure on the availability of land. In the 1973 solution, 9.64 million hectares appear to be underused, of a total of about 79 million hectares of all types of land contained on agricultural properties. In the 1978 base solution, this falls only to 7.08 million hectares. In both solu- tions, the good and medium-quality lands are underused in roughly equal proportions. Even with the expansion of markets over time, in the 1978 solution 3 million hectares of good cropland are not cropped at all. This permits a second conclusion: Time and the expansion of markets cannot be expected, in the foreseeable future, to induce full utilization of the North- east's land resources. Other features of the 1978 solution will be brought out in the compari- sons of alternative policy and project options that follow. Technical Progress: The Effects of Improvements in Crop Yields The promotion of technical progress-increasing crop yields through better agricultural practices and the use of so-called modern inputs-is a central component of most rural development programs and agricultural development plans. Technical progress on small farms is the principal 176 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-4. Technical Progress Experiment 25 percent 1978 increase in Percentage base case crop yields change Category (1) (2) (2)1(1) Gross output' Crops 11,146 11,812 + 6.0 Livestock 2,813 3,080 + 9.5 Total 13,959 14,892 + 6.7 Price indexb 100.0 87.7 - 12.3 Income' Family farms 764 977 + 27.9 Other farms 6,994 7,439 + 6.4 Hired labor 2,634 1,992 -24.4 Total 10,392 10,407 + 0.1 Fmploymentd Farm family 1,222 1,283 + 5.0 Sharecroppers 630 210 -66.7 Temporary workers 342 386 + 12.9 Permanent workers 427 333 -22.0 Total 2,622 2,211 -15.7 a. Thousands of cruzeiros in constant prices. b. Laspeyres index, 1978 base case = 100. c. Monev incomes in thousands of cruzeiros at model-endogenous prices. d. Thousands of man-years. objective of a newly formed government agency, the Brazilian Agricultural Research Company (EMBRAPA), and is a key component of the World Bank's numerous state-level rural development programs in the Northeast. Be- cause the model is based entirely on observed technological practices, it cannot address the question of how to improve yields as such: The survey simply did not reveal enough instances of advanced cultivation practices or of use of modern inputs to model such technologies. Nonetheless, the model can indicate the effects of increasing vields even if not the mecha- nisms and costs of doing so. This experiment assumes that the yields of all crops rise by 25 percent over the five-year period or by 4-5 percent a year, a rate at the lower end of the range (4-7 percent a year) specified by a recent agricultural research program in Brazil financed in part by the World Bank. The results are shown in Table 7-4. Consumers clearly benefit, as the gross output of crops and livestock products rises by 6.7 percent and the overall price index falls b)y 12.3 percent. Crop output increases by only 6 percent, which implies POLICY AND PROJECT OPTIONS 177 that market restrictions may severely limit the benefits of technical prog- ress. The 9.5 percent rise in livestock production is surprising; though one might have expected some of the land under pasture to be brought into use for crops, technical progress appears in fact to release land from crops for the expansion of livestock activities. That technical progress is land- augmenting is attested by a significant increase of 2.24 million hectares in underutilized land, mostly on estates.6 The simulation also reveals that, on balance, technical progress is labor- augmenting as well. Overall labor use declines by 15.7 percent; use of farm family labor rises by 5 percent, but there is a 66.7 percent drop in share- cropper employment and a 22 percent drop in permanent labor employ- ment. Declines in sharecropping have occurred in most regions of the world where there have been significant advances in agricultural technology (either yield improvements or mechanization that is directly labor- augmenting). The fall in permanent labor employment of 22 percent simply reflects the labor-saving nature of the technical progress we have simulated and is perhaps by coincidence almost as large as the percentage increase in yields. If it is assumed that yields improve equally on all farms, the income gains are greater for small farms: In the simulation, family farms realize income gains of 27.9 percent, whereas the medium-size farms realize a 5.2 percent gain and the large farms, 3.5 percent. In sum, the simulation suggests that technical progress produces sub- stantial gains in consumers' welfare and in small farmers' incomes but that on large farms land is used less intensively, fewer sharecroppers and permanent workers are contracted, and only moderate income gains are realized. If all disadvantaged groups are to benefit from technical progress, then other means of protecting the employment of nontenured labor or of inducing large farms to increase output and employment must be taken into account. The Influence of Risk It has long been conjectured that risk is an important factor limiting the volume of crop production and the adoption of new varieties and tech- 6. The 1978 base solution recorded 7.08 million hectares of underutilized land, down from the 9.64 million hectares of the 1973 base. The introduction of technical progress, which brings the amount back to 9.32 million hectares, virtually eliminates this gain in land-use intensity. 178 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL niques. The periodic droughts, some of which have been catastrophic, are the most vivid reminders of the effects of unpredictable and untimely rainfall in the Northeast, and they undoubtedly induce safety-first, subsis- tence-ensuring behavior on the part of many farmers of small plots in much of the Northeastern interior. The riskiness of crop production elicits the following tendencies: The more risky the crop, that is, the higher the coefficient of variation in a time series of revenues per hectare, the less important the crop will be in the farm plan; other things being equal, two (or more) crops whose output has typically deviated from the norm in opposite directions will be likely to appear in the same farm plan and jointly to form an important part of it; and the more risky the environment for most or all crops, the less intensely will most or all resources be employed. The first of these tendencies is easy to understand. The second is common in the Northeast in the form of consortiated planting, particularly on small farms. Corn and beans require timely and sufficient rainfall, for example, whereas some varieties of cotton can still produce an acceptable yield in arid years; hence, this combination is one of the most prevalent in the interior of the Northeast. The third tendency, toward a less intense use of resources, can be explained if one considers risk as a form of costs. The Hazell-Scandizzo formulation, in which these costs are made explicit as a penalty on more risky farm plans, yields a supply function higher than might otherwise be supposed and an equilibrium with lower output and higher prices.7 It is possible to quantify the influence of risk on farm plans, aggregate output, and resource use by setting the risk penalties in the model to zero. The implication is that farmers will consider only the expected values of revenue per hectare and will ignore their experience of deviations from normal yields and the correlations among deviations. Short of a fully guaranteed income based on the farm plan, there are no readily definable interventions that could induce such behavior, though price supports would presumably assist. The results of the simulation should thus be viewed as the maximum potential gains from risk-reduction policies in general. These gains are unremarkable. Aggregate crop output, livestock output, and employment all change by 1 percent or less. Some medium-quality 7. It is doubtful, however, that farmers' aversion to risk accounts for more than a small part of the underutilization of resources in the Northeast. The smallest farms, which are theoreti- cally and observedly the most risk-averse, use their land and labor resources very intensively, whereas the bulk of the underutilized land is on the estates, many of which are large and wealthy enough virtually to ignore risk considerations. POLICY AND PROJECT OPTIONS 179 land (about 4 percent) was transferred from pasture to crops, a shift that very slightly raised crop output and employment at the expense of livestock production. Possible reasons for this lack of response include the low price elasticities of demand, which restrain increases in food crop production under most supply-augmenting experiments with the model, and the insen- sitivity to risk of production on estates. A more important reason, however, appears to be that the evolution of crop varieties and cropping methods has gone so far in reducing the effects of risk that the behavioral aspects of farming have little remaining scope for improvement. These results should not be taken to mean that risk in general can be ignored in the Northeast. Even if catastrophic droughts are ignored, year- to-year variations in yields owing to the vagaries of the weather do cause substantial fluctuations in income, which can be devastating for small-scale farmers. Changes in cropping patterns and resource use will not reduce these fluctuations sufficientlv. Short of a complete income guarantee pro- gram, the best hope for reducing the risks faced by small-scale farmers lies in the identification of varieties and crops less susceptible to insufficient rainfall, combined with provision of sufficient land to cushion the effects of poor years. Employment Policies Severe agricultural underemployment coexists with extensive underuse of land. Before examining the role of demand we investigate whether labor policies could significantly reduce the employment problem. If land on large farms is underused because labor is too expensive relative to the marginal returns from cropping and livestock activities, then making labor cheaper should promote employment and output growth. Labor policies in the Northeast cannot be considered without addressing sharecropping, so this section also examines the implications of abolishing share contracts. We first consider subsidies on the wages of permanent laborers who, apart from sharecroppers, appear to be the only nontenured workers with a chance of earning a reasonable income. Temporary workers are not consid- ered because subsidies on their wages, which are paid on a daily or monthly basis, would be extremely difficult to administer. Table 7-5 shows the results of subsidizing permanent laborers' wages by 10 percent and 20 percent. Even under the 20 percent subsidy, crop output is virtually unchanged, and gross output rises less than 2 percent. Livestock output rises by about 9 percent-a result that reflects the relationships in the model; the workers employed for livestock maintenance typically have Table 7-5. Alternative Employment Policies Nvo share- cropping, 1978 10 percent Per- 20 percent Per- Io Per- 20 percent Per- base wage centage wage centage share- centage wage centage case subsidy change subsidy change cropping change subsidy change Category (1) (2) (2)1(1) (3) (3)1(1) (4) (4)1(1) (5) (5)1(1) Gross output' Crops 11,146 11,158 +0.1 11,165 +0.2 10,145 -9.0 10,660 -4.4 Livestock 2,813 2,919 + 3.8 3,059 +8.8 3,018 +7.3 3,194 + 13.5 Total 13,959 14,078 +0.9 14,224 + 1.9 13,163 -5.7 13,854 -0.8 Output index 100.0 100.9 101.9 94.3 99.2 Price indexh 100.0 99.7 96.9 104.6 97.1 Income' Family farms 764 754 - 1.3 748 -2.1 831 +8.8 802 + 5.( Otherfarms 6,994 7,011 +0.2 7,006 +0.1 6,8(03 -2.7 6,890 -1.5 Hired labor 2,634 2,786 +5.8 2,935 + 11.4 2,189 -16.9 2,849 +8.2 Total 10,392 10,551 + 1.5 10,689 +2.9 9,823 -5.5 10,542 + 1.4 Cost of wage subsidy 0 243 271 0 461 Employmentd Farm family 1,222 1,161 -5.0 1,133 -7.3 1,289 +5.5 1,119 -8.4 Sharecroppers 630 629 -0.1 604 -4.1 - -100.0 - - 100.0 Temporary workers 342 330 -3.5 360 +5.3 411 +20.2 380 +11.1 Permanent workers 427 542 +26.9 609 +42.6 640 +49.9 1,107 + 159.3 Totalnontenuredworkers 1,399 1,501 +7.3 1,573 + 12.4 1,051 -24.9 1,487 +6.3 Total 2,622 2,661 + 1.5 2,706 + 3.2 2,340 - 10.8 2,606 -0.5 Employment index 100.0 101.5 103.2 89.2 99.4 Average number offull-time workers on estates West 8.13 8.35 8.52 8.13 8.52 Sertao 12.94 13.06 12.82 10.93 12.47 Southeast 0.83 0.80 0.80 0.83 0.80 Fast 20.53 20.53 20.57 20.53 20.57 Agreste 27.72 27.62 27.24 27.72 27.24 -Not applicable. = a. Thousands of cruzeiros in constant prices. b. Laspeyres index, 1978 base case = 100. c. Money incomes in thousands of cruzeiros at model-endogenous prices. d. Thousands of man-years. 182 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL permanent contracts, there being little seasonal variation in labor demand for this activity. Sector income also shows little change, rising by 1.5 and 2.9 percent under the respective subsidy rates of 10 and 20 percent. The distributional effects of the wage subsidies are substantial. Perma- nent workers' employment and incomes rise by 26.9 and 42.6 percent under the 10 and 20 percent subsidies, and consumers gain somewhat because of the increase in output, which reduces the price index by 3 percent in the 20 percent subsidy case. Farmers' income as a wvhole is virtuallv unchanged. That of family farms declines slightlv because thev, not using hired permanent labor, are not much affected by the subsidy yet receive slightly lower prices. The overall effects on the agricultural sector, although disappointing, show that wage subsidies probablv have a positive cost/benefit ratio. Although the administrative costs cannot be estimated, the gains in sector income alone roughly offset the direct cost of the subsidies, leaving the gains in consumers' welfare as net benefits. The lower part of Table 7-5 shows nonetheless that wage subsidies cannot be relied upon to improve the performance of estates. Permanent employment increases slightly on estates in the West and the Sertao; in the other zones it remains the same or declines. On fazendas in the Sertao, where sharecropping predominates, there are no net additions to employment; about three sharecroppers per fazenda simply have their share contracts transformed into permanent worker contracts. Although wage subsidies reduce the marginal costs of production on larger farms, the effect on output from family and medium- size farms induces price declines that are large enough to continue to equate the marginal value product of labor with (subsidized) wage rates at about the same level of employment as before. Hence, wage subsidies alone will not significantly affect performance and employment in the region. The treatment of sharecropping in the model, described in Chapter 5, depends on the agreement of landowners and sharecroppers on existing farm plans, which are considered to be Pareto-optimal. It is not possible to simulate enforced changes to share contracts, but it is possible to simulate the abolition of these contracts. The changes that follow from so doing, reported in Table 7-5, take place almost exclusively in the Sertao, the main sharecropping zone. Crop output in the Northeast as a whole falls by 9 percent; crop output in the Sertao falls by 20 percent. Livestock replaces sharecroppers; average livestock output and herd size increase by 15 per- cent in the Sertao and by 7.3 percent in the Northeast as a whole. Prices rise by 4.6 percent, making consumers worse off and family farmers marginally better off. About 22,000 fully employed workers in the Sertao would lose their jobs if sharecropping were abolished. In the base case, fazendas in the POLICY AND PROJECT OPTIONS 183 Sertao employ an average of 12.28 sharecroppers and 0.66 permanent workers; when sharecropping is abolished they replace them with an average of only 10.93 permanent workers. The far-right column of Table 7-5 reports a simulation combining the abolition of share contracts with the 20 percent subsidy on permanent laborers' wages. Permanent workers replace almost all the sharecroppers, which results in much the same aggregate output, income, and emplovment levels. The only significant change from the base case is the substitution of livestock for crops, mainlv in the Sertao. These results reinforce the conclusion that intervention in share-tenancy arrangements would have no perceptible benefits. They also indicate that consumers' and hired laborers' welfare would almost certainly decline as a consequence. The Promotion of Demand Analysis of the base case solution suggested the hypothesis that the Northeast is a demand-constrained economy. The model can be used to give a broad idea of the effects on the sector at large of increases in demand for the principal crops consumed in the Northeast and of the distribution of gains that would result. Before undertaking such experiments, however, we must clarify what we mean by a demand constraint: It is easy-and naive-to recommend that a depressed economic region find unrestricted outlets for its products at higher prices. Interpretations of a Demand Constraint To review the analytics of the maximization procedure used, see Chapter 5, which describes how the model maximizes the sum of consumers' and producers' surpluses, such that the equilibrium price is equal to the mar- ginal cost of production (Figure 5-1). First, it is clear from Figure 5-1 that the economy would be better off either from a rightward (upward) shift in the demand schedule or from a rightward (downward) shift in the supply schedule. On the face of it, there is no reason to assume that either a demand constraint or a supply constraint is operating to prevent this. Second, the magnitude of the demand elasticity assumed, which fixes the slope of the demand schedule about the point P°Q°, has no relevance to producers' welfare in this static case: Variations in the elasticity will not alter the measure of producers' surplus defined as the area below the Po price line and above the supply curve. Third, where the supplv schedule represents an aggregation of different producers, the maximand acts as an allocation 184 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Figure 7-1. Supply Schedules for Price-Endogenous Crops of Different Classes of Farms: Model Structure Price S3 I I l I SI SI Quantity device. T he most "efficient" producers-those with the lowest average costs, including shadow costs of resources-supply the earlv portions of the aggregate demand schedule until their capacity is exhausted, and they are followed by the next most efficient producers, serially, until the surplus is exhausted and prices are equated to marginal costs. (It is important to note that this use of the term "efficiency" need have nothing to do with tradi- tional optimality conditions; differences in their supplv schedules may simply reflect differences in producers' resource endowments and in their access to input markets.) Figures 7-1 and 7-2 show stvlized representations of the Northeastern supply schedule for the price-flexible crops. In the first of these, SIS1 is the supply schedule of family farms; S2S2 that of medium-size farms; and S3S3 that of the estates. S,S, has a low price intercept, because the family farms rely largely or totally on "cheap" family labor, and a very short run on the quantity axis until land is exhausted, at which point the supply schedule becomes vertical. S2S2 has a somewhat higher price intercept, because family labor must be supplemented by more expensive hired labor, and it also has a longer run on the quantity axis before land is fully used. S3S3 has POLICY AND PROJECr OPTIONS 185 Figure 7-2. Alternative Equilibrium Points for Price-Endogenous Crops: Model Structure Price D, D, D3 I S Quantity Q0 Q2 Qp the highest price intercept, because the estates have the highest expendi- tures on hired labor, and a much longer run on the quantity axis, given their abundance of land. These three supply schedules are aggregated in Figure 7-2. A less simplistic view would result in a rising schedule that was smoother but still discontinuous in places because of the blending of different sources and types of labor contracts, different qualities of land, and so forth. Also shown in Figure 7-2 are three demand schedules. If DID, applies, the static equilibrium quantity will be Q'. This quantity is supplied by small farms up to the limit of their land, and the remainder by medium-size farms. The position of DID, in relation to the quantity axis results in a static equilibrium price too low to induce the medium-size farms to use all their land, and at which it is not profitable for the estates to produce these crops at all. If D2D2 applies, the equilibrium quantity will be Q2, which occurs at the discontinuous point where the resources of medium-size farms are fully utilized but the resultant price is still too low to bring the estates into food crop production. D3D3 appears to typify the Northeast equilibrium for many crops: The resultant price is just high enough to bring some, but not 186 THE AGRICUTLTURAL ECONOMY OF NORTHEAST BRAZIL all, of the large farms' land into crop production. A demand constraint, in the static sense considered so far, thus refers to a position of the demand curve in relation to the quantity axis at a point that yields a price too low to induce full utilization of the land and labor resources available. Before we discuss demand constraints in a less restrictive framework, it is worthwhile to compare the effects of price supports with those of demand shifts in the static framework. If DID, or D3D3 in Figure 7-2 applies, very small supported price increases will cause fairly large increases in supply (with risk and other factors that might be constraining ignored). If D2D2 applies, however, even a quite substantial supported increase in price may have no effect at all on the quantity supplied; it could indeed be equivalent to an income subsidy to those farmers currently producing the crops in question (with the usually substantial costs of administration, product collection, and disposition ignored). A shift in demand, by contrast, will almost always raise the quantity supplied,' and thus the intensity with which resources are used, but will not necessarily raise price. When the model is solved for different points in time, the price elastici- ties assume critical importance. Most agricultural projects are designed to increase production of agricultural commodities, and the elasticities of (demand have a direct bearing not only on the gains to the economv as a whole but also on the gains to the producers concerned. The relevance of price elasticities can readily be seen from Figure 7-3, which portrays a supply-augmenting intervention in the face of three demand functions, differentiated solelv by their elasticity (slope). In this figure, S.%. represents the prevailing supply curve of all producers combined,9 and S1S1 applies after the intervention, once all responses by producers have been worked through to a new static equilibrium. The three demand curves all correspond to the same equilibrium price and quantity conditions in the static case, represented by the point P°Q0. Depending upon which demand curve is applicable, the effects of aug- menting supply will differ dramatically. If the demand for the product is highly inelastic (DID), the intervention reduces prices substantially but 8. The exception is a small range on the quantity axis around Q2. Although it is accepted procedure to value incremental outputs at border prices (which implies infinite elasticity of demand), it must be realized that this theoretically attractive approach is of little relevance for the price-flexible crops considered here, which for a variety of reasons noted earlier are almost entirely consumed within the Northeast. 9. The reader may wish to examine the results of this analysis under alternative supply functions, such as those shown in Figure 7-2, and under alternative supply-oriented interven- tions, such as one that affects only output at the margin and does not affect the supply curve intercept, unlike the case illustrated in Figure 7-3. POLICY AND PROJECT OPTIONS 187 Figure 7-3. Effects of Augmenting Supply under Alternative Elasticities of Demand: Model Structure D, D0 D II I 2Ql Ql 2 Q3 increases the quantity supplied only slightly. This may be a desirable developmental strategy (at the extreme, it corresponds to taxation of farm- ers and subsidization of consumers), but it is not appropriate where the overriding considerations are to stimulate the use of resources and raise production and producers' incomes. If the intermediate, relatively elastic curve D2D2 applies, the intervention results in both moderate production increases and declines in price. Depending on the shape of the supply curves, producers may or may not be better off. As they are drawn, producers' surplus is about the same both before and after the intervention. If demand is perfectly elastic as depicted by D3D3, all of the gains consid- 188 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL ered accrue to the producers (the increase in producers' surplus is greatest), and the gains in production (and, other things being equal, in resource utilization) are the highest. These relations have no direct implications for policy, not onlv because it is difficult to influence demand elasticities through policy measures but because interpersonal utility comparisons between producers and consum- ers are required. Even in the case of an export crop, represented by the perfectly elastic curve D3D3, the gains from increasing supplv mav accrue largelv to the wealthier farmers, who are the main producers of the crop, whereas small-scale farmers, lacking access to the investment capital needed to establish stocks and arrange marketing, are relegated to the production of locally consumed food crops, demand for which is repre- sented bv D,D1. To sum up, when the effects of interventions to increase supply are being evaluated in a comparative-static framework, a demand constraint means that price elasticities of demand are so low as to restrain the response of production, and price declines are large enough to negate expected in- creases in producers' incomes. The final meaning of a demand constraint that we wish to consider is only partly relevant to the time frame of this study but could be extremely important over the longer run. This concerns the role of income elasticities. In projecting the model to a 1978 base, the position of the demand curves was shifted rightward on the basis of population growth, income trends, and income elasticities. If higher income elasticities had been assurned, the production increases and the gains in producers' incomes over the projec- tion period would have been larger. Although this effect is probably of onlv slight importance in a relatively short planning period, it is of substantial importance in the long run. As population and incomes grow, the resultant demand "pull" will have a direct bearing on how resources are employed to meet those demands and on the welfare of the producers employing those resources. Again, it is not a policv option to alter income elasticities, but interventions affecting production possibilities must be made in awareness of the longer-term demand for the products produced or producible in the region. Demand Interventions With this background we now turn to two experiments for the promotion of demand for the price-flexible crops: rice, beans, corn, manioc, sugar, and cotton, which together account for about two-thirds of the gross value of cr'op production in the Northeast. 'I'he first involves guaranteeing or sup- POLICY AND PROJECT OPT'IONS 189 Figure 7-4. Effects of Augmenting D)emand: Model Structure DD IPI QS QOQI porting the prices of these crops at their observed levels of the base period (1973). In Figure 7-4, this is represented by making the demand curve perfectly elastic at the observed price Po. The simulated price in the base solution, PS, is slightly lower than P° as was noted in Chapter 6; making demand perfectly elastic at Po should simultaneously raise both output and prices above the levels simulated in the base case. This experiment is reported in column (2) of Table 7-6. The directions of the resulting changes are as predicted, but the magnitudes are disappointing; crop output and employment grow by only 1.4 percent. In most zones, only beans and manioc are affected; output of both types of beans rises by 9.3 percent, and that of manioc by 7.8 percent. These increases are, however, sufficient to induce a 5.2 percent increase in the overall price level, which accounts for most of the increase in farmers' incomes: 17.3 percent for small-scale farmers and 12.5 percent for others (mostly medium-scale farmers). Whether such a price guarantee program would be desirable depends upon how one values increases in farmers' income as opposed to consumers' 190 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-6. Results of Promoting Demand 1978 Guaranteed Per- 25 percent Per- base prices for centage demand centage case food crops change shift change Category (1) (2) (2)1(1) (3) (3)1(1) Gross output' Crops 11,146 11,301 + 1.4 12,466 +11.8 Livestock 2,813 2,785 - 1.0 2,566 - 8.8 Total 13,959 14,086 + 0.9 15,032 + 7.7 Output index 100.0 100.9 107.7 Price indexb 100.0 105.2 111.2 Income' Family farms 764 896 + 17.3 849 + 11.1 Other farms 6,994 7,871 + 12.5 7,801 + 11.5 Hired labor 2,634 2,678 + 1.7 3,057 - + 16.1 Total 10,392 11,444 +10.1 11,706 +12.6 Employmentd Farm family 1,222 1,239 + 1.4 1,292 + 5.7 Sharecroppers 630 642 + 1.9 615 - 2.4 Temporary workers 342 342 0.0 345 + 0.1 Permanent workers 427 437 + 2.3 600 + 40.5 Total 2,622 2,659 + 1.4 2,853 + 8.8 Peak month 3,178 3,382 + 6.4 3,457 + 8.8 Employment index 100.0 101.4 108.8 a. Thousands of cruzeiros at constant prices. b. Laspeyres index, 1978 base case = 100. c. Monev incomes in thousands of cruzeiros at model-endogenous prices. d. Thousands of man-years, except peak month, which is thousands of men. w elfare. The costs of such a program in the Northeast would undoubtedly be prohibitive, however; effective price stabilization requires control of sizable storage stocks and a substantial administrative structure, neither of which exists in the Northeast or indeed in most developing countries.'" Furthermore, the resources required to establish these prerequisites would very probably yield a much higher return if used for other programs in the Northeast. For these same reasons, we shall not investigate direct measures to raise prices. The other demand promotion experiment considered is the more round- about one of inducing shifts in the demand functions; this is also illustrated 10. John W. Mellor, The Economics of A4gricultural Development (Ithaca, N.Y.: Cornell University Press, 1966), pp. 208-09. P'OLICY AND PROJECT OPTIONS 191 in Figure 7-4. The demand curve shifts from DD to DID1, inducing a supply response to Ql and raising the price level to P'. Such a shift, which we take as exogenous, could be caused by various interventions, such as nutrition programs, urban incomes policies, and even those agricultural policies designed to raise the employment and incomes of hired laborers. The results of this experiment, in which all of the price-endogenous demand functions are shifted rightward by 2 5 percent, are reported in column (3) of Table 7-6. Total crop output rises by 2.2 percent, on the average, implying quite high supplv elasticities. The supply of the affected crops, however, increases partly at the expense of livestock production, which falls bv 8.8 percent. Overall land use is much more intense; underutilized land falls from 9.64 to 4.81 million hectares. The effects on crop output realized through this shift in demand are much larger than those achieved with the price guarantee experiment. The two instruments are not comparable means of augmenting output, how- ever, because thev have verv different effects on prices: The crop price index rises by 0.9 percent under the price guarantees but by 8.8 percent under the demand shift. Nevertheless, the results of the demand shift experiment do suggest the magnitude of the changes in demand and prices that would be needed for significant gains in output and welfare. As regards the distribution of those gains, it is not possible strictly to compare consumers' welfare before and after the shift. The increase in crop production implies greater consumption, but overall prices rise by 11.2 percent. The major beneficiaries of demand promotion are farmers and permanent workers. Of these groups, the land-constrained family farmers gain the least, 11. 1 percent, entirely because of increases in price, not output. Sharecroppers and temporary workers are little affected, but permanent workers' employment rises b' more than 40 percent. The effects of the simulated shifts in demand strongly support the hypothesis that the agricultural economy of the Northeast is constrained by demand. Better outlets for Northeastern production would mean substan- tial gains in farm incomes and wage employment, although how outlets can be improved is another question. Development Strategies in the Absence of Agrarian Reform The model results thus far suggest that at least in the medium term the Northeast problem will not take care of itself; intervention is required. Before addressing the question of land reform, we simulate alternative 192 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-7. Alternative Development Strategies in the Absence of Agrarian Reform Solution 1978 base case Category (1) (2) (3) (4) (5) (6) (7) Instrument ' Technical progress X X X 25 percent demand shift X X X X 20 percent wage subsidy X X X Gross ototputb Crops 11,146 1,812 12,466 11,165 13,727 12,668 13,631 Livestock 2,813 3,080 2,566 3,059 2,806 2,736 3,103 Total 13,959 4,892 15,032 4,224 16,533 15,404 16,733 Output index 100.0 106.7 107.7 101.9 118.4 110.4 119.9 Priceindexc 100.0 87.7 111.2 96.9 92.5 104.9 87.3 Incomed Family farms 764 977 849 748 1,001 809 974 Other farms 6,994 7,439 7,801 7,006 8,267 7,759 8,118 [fired labor 2,634 1,992 3,057 2,935 2,613 3,455 3,005 Total 10,392 10,407 11,706 10,689 1,880 12,023 12,097 Employmente Farm family 1,222 1,283 1,292 1,133 1,288 1,164 1,105 Sharecroppers 630 210 615 604 422 530 175 Temporary workers 342 386 345 360 377 341 374 Permanent workers 427 333 600 609 485 954 1,012 Total 2,622 2,211 2,853 2,706 2,573 2,989 2,667 Peak month 3,178 3,281 3,457 3,675 3,950 3,802 3,957 Employment index 100.0 84.3 108.8 103.2 98.1 114.0 101.7 a. As defined in the text. b. Thousands of cruzeiros at constant prices. c. Laspeyres index, 1978 base case = 100. d. Money incomes in thousands of cruzeiros at model-exogenous prices. e. Thousands of man-years, except peak month, which is thousands of men. packages of the development measures so far considered to examine whether, under the existing agrarian structure, thev could induce substan- tial growth in production, employment, and the incomes of the poor. The simulations reported in Table 7-7 show the results of different combinations of technical progress (that is, a 25 percent increase in crop yields), a 25 percent rise in demand, and a 20 percent subsidy on the wages POLICY AND PROJECT OPTIONS 193 of permanent labor.1" The 1978 base case and the results of the interventions simulated individually are repeated for reference. The results of the pack- ages are difficult to evaluate without a social welfare function and, indeed, without an account of the costs involved. If, for example, the employment and incomes of nontenured workers are the main concern, then solution (6), combining demand shifts and wage subsidies, is the best set of instruments. If Northeastern consumers are the principal target group, then solution (2), involving only technical progress, is the best strategy. For family farmers, solution (5), combining technical progress and demand shifts, results in the highest levels of income (up 31 percent from the base solution). When all three instruments are combined, all the agents-consumers, farmers of all types, and nontenured workers-are better off than in the base case. Output in solution (7) is nearly 20 percent higher, prices are about 13 percent lower, and farm incomes rise by between 10 and 20 percent. Nonpropertied labor income increases by 14 percent, and about 800,000 more jobs are created in the peak month. Though these benefits are substantial, they are quite inadequate in the face of the employment problem: Overall labor use increases by less than 2 percent, and about 2 million people are still left unemployed even in the peak month. A package of measures such as solution (7) cannot be recommended without examining the performance of the estates before and after its introduction. Given their control of about half the agricultural land and their much less intense use of land and labor than that of family and medium-size farms, it is safe to conclude that a developmental strategy can be successful only if this sector responds dramatically to the interventions or is otherwise restructured. Table 7-8 reports the gross production and labor use on estates in each zone in the 1978 base case and in solution (7), in which all three instruments are applied. Except for the fazendas in the Sertao, the results are not encouraging: Estates in the Northeast as a whole increase production by an average of 14 percent (compared with 20 percent for the region if all types of farms are included) and labor use by about 9 percent (above the regionwide average but far short of what is needed). The vastly improved performance of the fazendas in the Sertao reflects changes in the profitability of cotton, the predominant crop and one of the most 11. It should again be stressed that the model is not a suitable tool for investigating other potentially powerful instruments such as agricultural credit, irrigation, marketing improve- ments, and additions to on-farm capital in the form of structures, equipment, storage facilities, or long-cycle crops. 194 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 7-8. Responses of the Estate Subsector to Policy Interventions Technical progress, 1978 demand sbifts, base and 20 percent Percentage case wage subsidy change Category (1) (2) (2)1(1) Productiona West 40,010 45,606 + 14.0 Sertao 39,502 67,106 +69.9 Southeast 65,334 65,317 - 0.1 East 243,993 267,900 + 9.8 Agreste 108,875 125,231 + 15.0 Northeast average 14.0 Employmentb West 11.62 12.67 + 9.0 Sertao 13.43 15.74 + 17.2 Southeast 14.75 14.71 - 0.3 East 35.05 36.65 + 4.5 Agreste 33.49 36.48 + 8.9 Northeast average 9.0 Net full employment increase' West 7,430 Sertao 17,168 Southeast 191 East 1,734 Agreste 7,339 Northeast total 33,480 a. Cruzeiros per farm. b. Man-years labor use per farm. c. Aggregate increase in number of permanent workers plus sharecroppers for all estates in the zone. labor-intensive grown in the Northeast. Both the vield of cotton and the demand for it increase by 25 percent, and labor is subsidized. Table 7-8 shows that the number of those fully employed (permanent workers and sharecroppers) on estates in the Northeast grows by only 33,480-an average increase of less than one worker per estate, which is insignificant compared with the aggregate emplovment problem. Whether the managers of estates optimize in practice appears doubtful. Even if they did so, however, and even with the substantial policv interventions consid- ered, these simulations reveal that the estates cannot be relied upon to solve the employment problem. 8 The Land Reform Alternative Land reform in the Northeast has been studied intensely, debated widely, legislated strongly, funded sufficiently, and all but abandoned. We did not set out to address the land reform issue in this study, and indeed we do so reluctantly. Agrarian reform has had mixed results in various parts of Latin America and the issue has had a stormy history in Brazil. A reform on the scale suggested by the sheer dimensions of the Northeast problem would undoubtedly tax the administrative capacity of the agencies involved, and disruption costs would be large, though difficult to measure. Nevertheless, the analysis has indicated that the land reform issue cannot be dismissed. Land and labor markets have failed to compensate for un- equal distributions of resources. Land rental markets are not operating- only 3 percent of the agricultural land is under rental contract, when vast differences in the shadow values of land indicate that this market should be vigorous. Estates, in particular, employ labor far below optimal levels in the face of widespread underemployment. Sharecropping, a contractual relation used throughout the world to alleviate land and labor distortions in a manner that satisfies both parties, is common only in the cotton- producing interior of the Northeast and is largely rejected by landowners in all other regions. Given the unequal distribution of assets in the Northeast, such market failures are severely impeding agricultural development. The changes in employment and income that could be wrought by the tradi- tional policy instruments so far considered are wholly inadequate to meet the problem. Structural interventions to reduce the distortions in land and labor markets may in fact be a prerequisite for agricultural development in the Northeast. The model's representation of farms of different types and sizes and its incorporation of aggregate constraints on demand make it possible to estimate both the aggregate and the distributional effects of agrarian re- form. Though it cannot directly address questions of costs or trace the dynamics of adjustment over time, the model gives more realistic results 195 196 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL than could be obtained, for example, by extrapolating microeconomic studies. Before the model's simulation of land reform is presented, it is necessary to review the legislative experience and briefly describe the official plans that have been put forward. Legislative History The current Brazilian land reform law, the Estatuto da Terra (Law no. 4504), was introduced in November 1964 and is an elaborate extension of Article 146 of the 1946 Federal Constitution, which enables the president of the republic to decree land reform measures.' Among its stated goals is to develop and strengthen efficiency and equitv in the agricultural sector. Its features include: commitment to private enterprise; recognition of a need for government intervention; creation of instruments to implement agri- cultural change on privately held properties; establishment of performance norms for such properties; isolation for remedial action of regions in which a large proportion of the properties fail to meet these norms; planning and implementation of agricultural colonies to absorb the potential surplus population; creation of agencies to administer the above programs; and establishment of a land inventory. The statute's aims are to be achieved through progressive land taxation and curtailment of the right to private property. All farms are to be taxed, depending on their size, location, production relationships, and social conditions, so that large and inefficient farms are taxed progressively. The highest potential tax rate, however, is only 3.45 percent of a property's unimproved land value, and the lowest is 0.02 percent. Ludwig and Taylor recognize that this range is "probably too low to evoke a response from affected land holders."2 The statute recognizes four types of holding, which are defined in terms of modulos rurais (land reform modules), a concept of fundamental impor- tance. A module is a unit of land capable of providing the equivalent of four minimum annual salaries or requiring the labor of four working adults and of supporting them at a standard of living consistent with overall goals of 1. The section "Legislative History" was originallv written by Maria Helena de Castro Silva. The principal source in English is Armin K. Ludwig and Hlarry W. Taylor, Brazil's New Agrarian Reform (New York: Praeger, 1969). A more recent study focusing on the political aspects of the topic is Marta Cehelsky, Land Reform in Brazil: Tbe Management of Social Change (Boulder, Col.: Westview Press, 1979). 2. Ludwig and Taylor, Brazil's Vew Agrarian Reform, p. 81. THE LAND REFORM ALTERNATIVE 197 economic and social progress. The types of holding are: minifundia, of less than one module; "latifundia bv size," that is, greater than 600 modules;' "latifundia by use," according to the degree of inefficiency with which the land is used and ranging between 1 and 600 modules; and empresas rurais, or rural enterprises, which may also be between I to 600 modules and which are distinguished on performance criteria. Holdings of the first three types lie under threat of expropriation or abrogation of the right of free disposal. Land that is expropriated is to be redistributed into modules and given preferentially to landless workers. Recipients need only be twenty-one years of age or over, sane, and of good background or demonstrating the potential for rehabilitation. Recipients are eligible to borrow an amount up to one minimum annual salary for infrastructure investment and initial working capital at a 6 percent rate of interest (presumably nominal and thus highly negative) for up to twenty years. The statute authorizes funds with which to compensate landowners and delineates priority areas for reform, which in the zonal delineation used in this study are in the Sertao and the East. The agencies initially established under the statute were absorbed into INCRA (National Institute of Colonization and Agrarian Reform) in 1970; since then this agency has been largely responsible for implementing the land reform law. By its third year, INcRA had a budget of US$ 100 million and 7,000 employees. Cline has documented a series of ill-fated or half- hearted attempts by INCRA to establish family farms (only 563 of 8,000 target families were resettled in 1973, mostly in colonized Amazonian areas).4 INCRA has, however, completed two of the prerequisites of a land reform: a comprehensive cadastral survey, which is periodically updated,' and detailed specifications of land reform modules. The Land Reform Module As defined by INCRA, the module farm should be of sufficient size and capital endowment to fully employ four adults and yield each of them a net annual income equal to the minimum wage,6 plus a return of 15 percent on 3. In most areas of the Northeast, this criterion implies a farm larger than 20,000 hectares. 4. William R. Cline, Economic Consequences of a Land Reform in Brazil (Amsterdam: North- Hlolland Publishing Company, 1970). 5. The INCRIA cadastral survey of 1972 provided the basic information on landholdings used for the SUDENE/World Bank survey (see Appendix B to this volume). 6. About US$450 in 1974. 198 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL the capital invested in it. The SUDDENE/World Bank survey shows that in most of the Northeast existing farms need to be around 150 hectares before they can fully employ four adult workers, whereas the average productivity per fully employed worker is only around US$400 annually. Nevertheless, INCRA applied the module definition in great detail to each of the 1,365 municipios in the Northeast, taking account of population density in the locale (as a proxy for access to markets), existing patterns of land use and production, soil quality and other ecological conditions, and overall poten- tial productivity. The module farm sizes that resulted from that exercise range from less than five hectares (sometimes as small as two) in the coastal areas near large urban markets, to more than 100 hectares in areas of the sparsely populated West where stock raising is the main activity. Potential for Redistribution of the Estate Subsector The five physiographic zones used for the model in this study certainly cannot match the very detailed disaggregation used by INCRA; nor can the very broadly defined farm types in the model be used as a proxy for INCRA'S classifications. But it is at least possible to simulate the interpolation of average-size INCRA module farms into the existing agrarian structure, zone bv zone, because the model is capable of reallocating land into different farm sizes and types. The average size of the INCRA module is calculated for each state and the figures are then transformed to average sizes for the zones used in the model: West, 63.15 hectares; Sertao, 31.30; Southeast, 19.90; East, 35.90; and Agreste, 23.20. In the experiments reported below, the land composing our estate subsector is divided into farms of these sizes.' Though this is a somewhat crude procedure, the results of the cross- sectional analysis reported earlier revealed that very few of the existing estates would not be classified as latifundia, either by use or size, if the criteria in the land statute were applied to them.9 Table 8-1 shows that nearly 800,000 module farms could be formed from the land on existing estates; this would yield a total of about 1.5 million 7. Ludwig and Taylor, Brazil's New Agrarian Reform, p. 16. 8. Estates in the Humid Southeast, comprising about 5,000 plantations in Bahia and less than 3 percent of total Northeastern agricultural land, are ignored throughout the land reform simulations because they were inadequately covered bv the survev. 9. Although the majority of minifundia are too small by almost any criterion, we do not adfdress the question of farm consolidation; administrative attention to the 1.4 percent of the land contained in properties smaller than ten hectares could oniv detract from the required attention to the 50 percent of the land on farms larger than 500 hectares. THE LAND REFORM ALTERNATIVE 199 Table 8-1. Potential Number of Module Farms in the Estate Subsector South- North- Item West Sertao east a East Agreste east Number of existing estates 14,569 11,221 4,766 1,314 5.168 37,038 Average size (hectares) 794.57 676.43 763.47 1,086.50 634.56 742.80 Total area (thousands of hectares) 11,576 7,590 3,639 1,428 3,279 27,512 Module size (hectares) 63.15 31.30 19.90 35.90 23.20 34.84 Potential number of module farms 183,31(0 242,492 182,864 39,777 141,336 789,779 a. Cacao plantations are excluded, as explained in note 8 to the text. farms (family, medium-size, and module), or nearly double the present number. Performance of the Module Farms Table 8-2 summarizes the characteristics of the average module farms when they replace the existing estates. Because these performance figures are from the solution for the whole agricultural sector, reported in Table 8-3, all the sectorwide constraints in the model are operating. By contrast, when a model of a single farm (or farm tvpe across the zones) is solved in isolation, the behavior of the farm does not affect any market equilibrium relations, and hence the solution is apt to give an unrealisticallv favorable impression of its performance. The results of the module experiment confirm that, with existing tech- nology, practices, and markets, farms of this size range cannot provide full employment for four working adults. In all zones, the average modules barely absorb the labor of one family, assumed to be equivalent to 2.2 adults, and do so for a maximum of only one or two months a year. The modules in the East use the most labor on a year-round basis, at two man-years (600 days) a year, but the amounts they absorb vary from month to month; a maximum of 2.54 men are employed in the peak month. In the West, family labor is fully occupied in only one month of the year and never needs to be supplemented with temporary hired labor. In none of the zones 10. For example, another simulation of a comprehensive land reform that ignored demand constraints and other factors revealed a scope for expanding output up to 80 percent as opposed to our figure of about 15 percent, reported below. See R. Albert Berry and William R. Cline, Agrarian Structure and Productivity in Developing Countries (Baltimore, Md.: Johns Hopkins University Press, 1979), p. 130. 200 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table 8-2. Characteristics of Land Reform Modules South- Characteristics West Sertao east East Agreste Size (bectares) 63.15 31.30 19.90 35.90 23.20 Good cropland 5.37 5.49 7.40 4.17 4.33 Medium cropland 5.82 8.23 4.36 5.10 6.23 Native pasture 15.64 10.48 2.53 6.70 11.28 Output and incomes (cruzeiros) Crop sales 3,519 1,927 3,758 7,762 3,699 Livestock sales 759 2,229 1,272 1,238 2,046 Head (number) 7.37 11.98 11.78 3.99 9.70 Value of consumption 518 767 770 785 934 Gross output 4,796 4,922 5,800 9,785 6,679 Money costs 191 520 464 362 805 Net money income (gross output minus money costs) 4,605 4,402 5,336 9,423 5,873 Net real income 3,302 3,336 4,397 7,676 4,256 Labor use Total labor use (man-years) 1.45 1.20 1.09 2.00 1.86 Peak month family labor use (man-equivalents) 2.20 2.20 2.20 2.20 2.20 Peak month temporary labor hire (man-equivalents) 0.00 0.15 0.52 0.34 0.43 Sbadow prices (cruzeiros) Good cropland 126 52 38 510 104 Medium quality cropland 83 52 38 394 65 Native pasture 6 42 38 67 28 Performance (cruzeiros) Output per hectare 76 157 291 273 288 Output per man 3,308 4,102 5,321 4,893 3,591 Labor per hectare (man-years) 0.023 0.038 0.055 0.056 0.080 do the modules need to hire permanent workers, even though the observed annual wage for permanent workers that is used in the model is less than half the minimum individual income as specified by INCRA. Very little temporary labor is hired, the maximum being about two man-weeks in the peak month on module farms in the Southeast. Even if the modules do not meet INCRA'S stated employment objectives, they are quite consistent across zones in the amount of family labor they absorb. Family labor is fully employed in all zones in at least one month and supplies the following proportions of total labor use: West, 100 percent; THE LAND REFORM ALTERNATIVE 201 Table 8-3. Results of a Land Reform in the Estate Subsector Percent- 1978 Land age base case reform change Category (1) (2) (2)1(1) Number of farms 779,519 1,532.260 + 97.0 Number of farmers and family workers 1,751,279 3,914,476 + 124.0 Outputa and incomeb Crop output 11,146 12,630 + 13.3 Livestock output 2,813 3,361 + 19.5 Gross output 13,959 15,991 + 14.6 Farm income 7,758 10,531 + 35.7 Nontenured labor income 2,634 1,462 - 44.5 Sector income 10,392 11,994 + 15.4 Employment Man-vear equivalents 2,622 3,034 + 15.7 Peak month (thousands of men) 3,178 4,274 + 34.5 Trough month (thousands of men) 1,868 2,119 + 13.5 Sector performance Price index 100.0 94.7 - 5.3 Output per hectare (cruzeiros) 177.1 202.9 + 14.6 Output per man-year (cruzeiros) 5,324 527.1 - 1.0 Labor per hectare (man-years) 0.033 0.039 + 16.6 Underutilized land (millions of hectares) 7.08 3.90 - 45.0 a. In thousands of cruzeiros at constant prices. b. Money incomes in thousands of cruzeiros at model-endogenous prices. Sertao, 99 percent; Southeast, 96 percent; East, 97 percent; and Agreste, 97 percent. The simulated income levels of the module farms are also far short of those specified by INCRA. As Table 8-2 shows, the highest money income is realized in the East and is equivalent to only about two, rather than four, INCRA minimum salaries (with the "return on capital" requirement ignored). Total money incomes from the module farms in all other zones average about the INCRA minimum individual salary. Even though much lower than the INCRA standard of CR$20,000, these incomes are substantially above those currently earned by Northeastern family farmers, sharecroppers, and permanent workers and are thus acceptable. Although they do not meet INCRA'S standards for absorbing labor and generating income, the modules perform very well as family farm units. 202 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL They are capable of providing acceptable levels of income and employment for rural families, and they are large enough to take advantage of improving markets and advances in technology. We can find no reason to alter the INCRA farm-size specifications for a potential land reform and can offer no better criteria for defining module farms. Potential Effect of Land Reform Apart from the social, political, and distributional objectives, a land reform should contribute to economic development by increasing produc- tivity. Not all land reforms have done so, and to counter the apparent preference for commercial agricultural enterprises in Brazil, the potential gains in productivity from newly created family farms need to be clearlv demonstrated. Table 8-3 summarizes the results of the land reform simulation for the agricultural sector as a whole. The gains in output are far short of what might be expected: Crop output rises by 13.3 percent, and gross output by 14.6 percent. Though livestock output might have been expected to de- cline, it increases by almost 20 percent because the production of the estate subsector is much more livestock-intense than that of existing small farms. rhis reflects the demand limitations, which result in a 5.3 percent decline in the overall price index. The gains in gross output arise from additions to the production of food crops for family consumption on the modules and additions to marketed production, the modules being more efficient than the existing estate sector. Much of the increase in crop output is of rice, beans, manioc, and corn, which the module farms, like familv farms, grow mainly for on-farm consumption, but the production of other crops does not decline. This suggests that a land reform will not have a detrimental effect on the production of export crops and that the production of consum- ables, both crop and animal products, will increase. The additional value of subsistence production is relatively small: Cr$572 million, or 28 percent of the increase in gross output of Cr$2,032 million. In the sense that the model "forces" family farms to produce a minimum consumption bundle, and the land reform simulation increases the number of these farms, the consumption gains from this source are somewhat artificial. There is, however, no reason to assume that module farms would not produce at least this amount of food crops for their own consumption. The remaining increase in output is "pure" in the sense that it arises from the reallocation of labor to land and the resultant decline in THE LAND REFORM ALTERNATIVE 203 production costs. This increase is still greater than could be brought about by most of the traditional interventions reported in Table 7-7, the excep- tions being those combining demand shifts with technical progress (5) and also with wage subsidies (7). The emplovment gains from a land reform are more substantial than the output gains. Overall labor use rises by 15.7 percent, more than under any of the alternatives considered in Chapter 7 (even those that most favor labor-demand shifts and wage subsidies). More important, peak-month employment rises by more than I million workers. The gains to consumers from a land reform are difficult to measure because the size and characteristics of the consuming population relying on the markets change as families on the newly formed modules produce their own food. The price decline of about 5 percent, however, indicates that all consumers would benefit. This follows, in part, from the output increases of about 15 percent in food crops and livestock products. The number of families who are assured of being able to produce and consume the mini- mum food crop "bundle" rises from 667,995 in the base case to 1,420,696 under the land reform, an increase of 113 percent. A land reform, then, is likely to benefit all Northeast consumers substantially, though to varying degrees, and ensure adequate consumption for an additional 750,000 fami- lies. More livestock is produced under land reform than under any other intervention considered, but land in general is used much more intensively. Underutilized land falls from 7.08 million hectares in the base case to 3.90 million under the land reform-a far greater decline than under any other intervention. This is due partly to the increase in crop production and partly to the modules' tendency to use cultivated pasture, which is more labor-intensive and land-saving, whereas livestock maintenance on the estates in the base case was based mostly on native pasture. Although the aggregate effect of a land reform would not be large, the micro effects are striking, as shown by a comparison of the output and employment of the large farm subsector before and after redistribution (see Table 8-4). Note that all other things, particularly the land endowments by quality, stock of tree crops, and so on, are identical. In this light, the potential increases are more convincing. For the entire subsector, crop output more than doubled, and gross output and employment nearly doubled. Whereas this subsector accounted for 22.1 percent of Northeast output in the base case, its share rose to 37.4 percent, slightly more than the 34.5 percent of the land these properties comprise. Thus, because of the land reform, the subsector's performance rose from far below average to become one of the most productive. Table 8-4. Performance of the Fstate Subsector before and after Redistribution Crop outputa Livestock outputa Gross output'a Labor use b Per- Per- Per- Per- centage centage centage centage Zone Estates Modules change Estates Modules change Fstates Modules change Estates Modules change West 480.19 740.02 + 54 106.65 139.13 + 30 586.85 879.15 + 50 169.29 265.80 + 57 Sertao 432.16 653.27 + 51 201.90 540.51 +168 634.51 1,193.79 + 88 150.70 290.99 + 93 Southeast 4.44 848.85 + - 306.94 232.60 - 24 311.38 1,081.46 +247 70.30 199.32 + 183 East 289.28 445.26 + 54 31.33 49.24 + 57 320.61 494.51 + 54 46.06 79.55 + 72 Agreste 389.24 686.04 + 76 173.43 289.17 + 67 562.66 975.22 + 73 173.08 262.88 + 52 Northeast 1,595.76 3,373.46 + 111 830.25 1,250.67 + 52 2,416.02 4,624.12 + 91 609.42 1,098.55 + 80 Note: TIhis table concerns only the estate subsector; it thus gives a more favorable impression than the net results for the agricultural sector as a whole that appear in Table 8-1. a. In thousands of cruzeiros at constant prices. b. In thousands of man-years. THE LAND REFORM ALTERNATIVE 205 The employment gains are also striking; the subsector, after redistribu- tion, provided full employment (300 days a year) for nearly an additional half-million. Recall that under the 20 percent wage subsidy experiment employment increased by less than 30,000 on the large farms. Table 8-4 also suggests the areas that might be prime prospects for a land reform. Cline considered the Zona da Mata (roughly corresponding to our East zone) as having priority, and INCRA included parts of the Sertao as well as the East. Their considerations, however, were highly influenced by the degree of social tension prevailing, whereas ours are only in terms of output and employment effects. Under these criteria, the Southeast, comprising most of the state of Bahia, is potentially the most profitable area, followed by the Sertao. The East is, in fact, below the Northeast-wide average gains in output and employment. The vast potential gains (of 50 percent or more) in output and employment in all zones, however, indicate that none should be ignored. The most impressive gains from the land reform simulation are distribu- tional. If we consider family incomes of US$456 (the bottom of ENDEF'S third income class) to be at least acceptable," all of the 790,000 module farms fall into this category. Presumably, the beneficiaries of the module farms would be temporary and permanent workers' families as well as sharecroppers' families-the vast majority of whom are earning substan- tially less than this figure. The land reform could thus raise the incomes of as many as 40 percent of these families to acceptable levels. The 1978 base case showed only 680,000 families or farms earning incomes greater than US$456 from agriculture. In the land reform simula- tion, this figure nearly doubles to 1.32 million, for a net gain of 640,000. Although this gain is not impressive in view of the extent of poverty among nonlandowning families, it is impressive when compared with what could be accomplished through interventions other than land reform; none of the policies or packages of policies described in the previous chapter raised the number of families earning more than US$456 by as many as 100,000. Potential Beneficiaries Because there are about 2 million nonlandowning families depending on agriculture, there can be no shortage of beneficiaries for the approximately 700,000 module farms considered in this experiment. Most of the family heads probably meet INCRA's basic criteria (aged 21, sane, and of good 11. See the incomes discussion in the appendix to Chapter 3. 206 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL background or demonstrating the potential for rehabilitation). Given the technological level, minimal contacts with markets for both inputs and outputs, and simple practices observed for perhaps centuries, we discount the contention that a land reform would fail because of managerial short- comings and illiteracy. If, however, managerial experience were to be a criterion for selection, the roughly 250,000 existing sharecroppers and roughly 1 million permanent workers, who are often moradores operating their own family plots for food crops, would be the best candidates. If such a criterion were exclusively applied, however, it would limit the effect of the reform on income distribution, because those with the lowest income levels are temporary workers. Estimates of Cost Probably the most difficult aspect of evaluating a land reform is to estimate its cost. 12 In the early 1 970s, Cline used an estimate of US$2,500 as the nonrecoverable cost per module,'3 which SUDENE later updated to US$2,722 to take account of inflation. A 1978 unpublished SUDENE study places the figure at US$5,000 per module, of which US$3,000 for prepara- tory land work and installations would be "recoverable." None of these sources provides a detailed breakdown of the costs involved, but they suggest rough orders of magnitude. Each of the several costs of land reform has to be analyzed differently. First, there may be anticipatory costs: If a landowner feels that appropria- tion is imminent, he may alter the operation of his fazenda, possibly evicting sharecroppers and workers whom he may view as potential ben- eficiaries, causing real losses in employment and output and in his own income. In view of the long debate on the subject in Brazil, such costs have almost certainly been incurred, and they will be minimized only by deci- sive action one way or the other. Second, there are the transactional costs of legal services, surveying, title registration, and the like. Third, transfer payments will normally be made to compensate landowners. Whether this compensation comes from public coffers or from the beneficiaries makes little difference except in its distributional impact. Fourth, there are the investment costs of outfitting new farms with structures, equipment, ani- 12. All dollar figures in this section have been converted from cruzeiros at the spring 1974 exchange rate of Cr$6.40= US$1.00. 13. In World Bank, "Rural Development Issues and Options in Northeast Brazil," report no. 665a-BR (Washington, D.C., June 23, 1975; restricted circulation). THE LAND REFORM ALTERNATIVE 207 Table 8-5. Estimated Partial Costs of Establishing Land Reform Modules (1973 U.S. dollars) Average value South- Aortheast per module West Sertao east East Agreste averagea Land 1,781 1,756 4,766 7,193 3,685 3,073 Structures 1,504 1,636 2,494 3,864 2,087 1,995 Equipment 138 121 117 245 107 128 Animals 1,959 1,304 931 1,562 1,699 1,453 Total nonland value 3,601 3,061 3,542 5.671 3,892 3,576 Total value 5,382 4,817 8,308 12,864 7,527 6,649 a. Weighted by potential number of modules in each zone (see Table 8-1). Source: SUDENE/World Bank survey. mals, working capital, and facilities (mainlv access roads) which enable them to purchase inputs and market their products. Finally, there are the costs of extension services, credit provision, and other administrative ser- vices, depending on how much the government is involved in the operation of the modules. The sUDENE/World Bank survev data permit estimates of some of these costs. Table 8-5 reports the average value of land (a proxy for the cost of acquisition) and of structures, equipment, and livestock, by zone, for farms about the size of the INCRA modules, in UL. S. dollars to ease comparison with the estimates quoted above. The land values per module were calculated from survey data based largely on farmers' own judgments. To check these values, the marginal returns to land on all estates were averaged for each zone, with weights corresponding to the quantities of land of the various classes of productivity. This vielded for the Northeast an average annual net income of about US$240 per module. Because half or more of the land on existing estates is not under productive use, and adjustment is made for differences in the shadow prices, the value of this stream over twenty years, discounted at 10 percent, is slightly over US$2,000, or about two-thirds of the average declared land value. The value of structures, equipment, and animals amounts to about US$3,500 per module. SUDENE considers the costs of equipping new farms to be recoverable, and the model results suggest that they should be, because annual net money income, after family consumption requirements have been met, averages about US$ 1,000 per module. Assuming, however, a 50 percent default rate and interest charges that are just sufficient to cover 208 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL administrative charges and opportunity costs, the cost of installations to be publicly financed would average US$1,750 per module. Recent World Bank estimates of the annual cost of recommended exten- sion services average about US$120 per family farm in the Northeast. Over, say, five years, presumably a time sufficient to achieve the perfor- mance estimated in the model, this amounts to US$600 per farm, or less than US$500 if discounted at 10 percent. These considerations appear to confirm Cline's figure of US$2,500 in the early 1970s as an average for the real, nonrecoverable cost per module. On this basis, the establishment of the 790,000 modules considered in our experiment would cost US$1,975 million." The additional gross output of US$318 million a year that could be produced in the Northeast if the estate subsector were reformed amounts to 16 percent of the land reform costs. Discounted at 10 percent, this stream of output increases would enable the land reform to break even in ten years. To the extent that the model overestimates the efficiency of the existing estates, this output gain is understated, and the reform is more attractive. Alternative Development Strategies in Conjunction with a Land Reform Despite these benefits, a land reform alone could not solve the problems of the rural Northeast. In this section we therefore examine the possible effects of a land reform combined with the instruments described in the previous chapter. Table 8-6 shows that if the estate sector were reformed the output of the Northeast could be raised by up to 35 percent and employment by almost 30 percent, depending on the combination of instruments. These are much larger gains than could be achieved by the interventions alone. The differ- ent strategies are difficult to evaluate without a social preference function, but a comparison of the results in this table with those in Table 7-7 shows that the traditional policy interventions-particularly the promotion of technical progress and measures to expand demand-would be more effec- tive if the estate sector were reformed than they would be otherwise. Prices paid by consumers would always be lower, the incomes of all farmers higher, and total employment greater. 14. By 1976, PROTERRA funds amounted to roughly double this figure. THE LAND REFORM ALTERNATIVE 209 Table 8-6. Alternative Development Strategies in Conjunction with a Land Reform 1978 base case Category (1) (2) (3) (4) (5) (6) (7) (8) Instrument' Technical progress X X X 25 percent demand shift X X X X 20 percent wage subsidy X X X Land reform X X X X X X X OutpUtb Crops 11,146 12,630 13,279 14,173 12,656 15,212 14,286 15,151 Livestock 2,813 3,361 3,727 3,003 3,474 3,396 3,118 3,561 Total 13,959 15,991 17,006 17,176 16,130 18,608 17,405 18,712 Output index 100.0 114.6 121.8 123.1 115.6 133.3 124.7 134.1 Price index 100.0 94.7 82.0 104.6 92.8 86.9 99.6 84.8 Income' Family farms 764 4,751 5,563 5,635 4,626 6,154 5,392 5,863 Other farms 6,994 5,780 5,884 6,162 5,777 6,281 6,096 6,324 Hired labor 2,634 1,462 543 1,525 1,696 798 1,864 1,316 Total 10,392 11,994 11,990 13,322 12,099 13,234 13,353 13,504 Employment' Farm family 1,222 2,215 2,274 2,464 2,099 2,448 2,309 2,196 Sharecroppers 630 587 0 576 587 99 576 6 Temporary workers 342 181 172 182 164 196 151 172 Permanent workers 427 51 43 96 223 46 346 525 Total 2,622 3,034 2,489 3,318 3,073 2,790 3,381 2,899 Peak month 3,178 4,274 3,952 4,148 4,401 3,999 4,315 4,478 Employment index 100.0 115.7 94.9 126.5 117.2 106.4 128.9 110.6 a. As defined in the text. b. Thousands of cruzeiros at constant prices. c. Money incomes in thousands of cruzeiros at model-exogenous prices. d. Thousands of man-years, except peak month, which is thousands of men. Two groups might, however, require special attention to prevent their incomes from falling to a poverty level as the result of a land reform. Existing medium-scale farmers, though they are better off than small-scale farmers and landless workers, could suffer a decline in income of up to 20 percent because of the price declines and because, in particular locales, only less skilled labor might be available for hire after a land reform (a possibility not recognized by the model but probably of some importance). Strategies to promote demand would reduce these income losses. Sharecroppers 210 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL might also suffer unless they are among the direct beneficiaries of a land reform: Under the alternative programs considered, their employment fluctuates dramatically from 587,000 (slightly below the number in the 1978 base case) to zero. Conclusions The legislation, background information, and ample funding exist for a land reform like that simulated. The Land Statute, on almost all accounts, is precise, realistic, and to the point of the problems; the INCRA criteria for defining farm sizes and types are valid and consistent; and, even with allowance for inflation, existing PROTERRA funds are double our estimates of the probable cost. The analysis of farm performance has shown that there is little, if any, economic justification for the existence of most estates, given the vastly superior performance, by almost any economic criterion considered, of farms of module size or smaller. Furthermore, a reform involving 2.7 million families at a proposed cost of US$13,500 million was recently drafted in Brazil.15 Two central questions must, however, be borne in mind in any inter- pretation of the land reform simulations in this chapter. First, could a reform of the scale considered here actually be undertaken, given the political and administrative challenges it poses? Second, is a development program encompassing such a redistribution sufficient? The administrative capacity to implement such a reform is at least as important as legislation and funding. To minimize disruption costs, any reform must be carried out swiftly and effectively. The bureaucratic night- mares associated with some of the resettlements attempted by the irrigation aigency (DNOCS) are certainly not typical of current administrative capability in Brazil, but such examples emphasize the importance of administrative arrangements in any intervention. At a minimum, the administrative re- (quirements for a land reform need not go much beyond measures to put the landless Northeasterner on the land. Farming techniques are simple and have been handed down from father to son for centuries. Though we do not (lispute the need for extension services, research and dissemination of new varieties, a functioning credit network, and marketing facilities, we argue strongly that in the Northeast of Brazil such activities are subsidiary to the need to give better access to land to the multitude of available agricultural 15. As reported in "Agrarian Reform: A Time-Worn Subject Returns to the Headlines," I'eja, December 7, 1977. THE LAND REFORM ALTERNATIVE 211 workers. Unless their access is extended, both human and land resources will continue to be underused, and the objectives of agricultural develop- ment will continue to be frustrated. A land reform could increase production and employment and raise a significant number of people out of absolute poverty. But conditions in the rural Northeast leave the sufficiency of a land reform much in doubt. Even if a reform of the existing estate sector were supported by the successful stimulation of demand and technical progress and were accompanied by a subsidy on wages, more than I million families would remain without entrepreneurial access to land and near absolute poverty. It is clear that the full solution to the Northeast problem does not lie within the agricultural sector alone. But it is equally clear that a land reform holds the best prospects for directly improving productivity and the distribution of in- comes within the rural Northeast. 9 The Northeast Problem and the Future of Brazilian Development This studv concludes in a quite sobering manner: The Northeast problem is probably more acute than previously imagined; it is not likely to dis- appear over time of its own accord; direct policy and project interventions are not likely to have much effect; thus, there is a need for sweeping structural change within the Northeast's agricultural sector. Yet even such a move will only partially alleviate the severe relative-and possibly abso- lute-poverty persisting in the region. Because this study focused on a single sector in a single region and relied largely on a single cross-sectional farm survev, the above conclusions wvarrant further reflection and raise many questions. If Brazil can resume its "miracle" growth record, will the benefits eventually trickle down to the rural Northeast? Are there other, less politically sensitive solutions avail- able? What should be the role of the agricultural Northeast in the future lBrazilian economy? What are the likely ramifications if the Northeast problem is not solved? Miracle Growth and the Rural Northeast From the end of World War II to the early 1960s the Brazilian economy grew at rates envied by most of the developing world. This growth has been well documented and virtually unrefuted. Whereas most students would agree that the poor of Brazil have not experienced absolute declines in their welfare during the growth period, all indicators point to a decline in the income share of the poorest 40 percent of the population. Although two- thirds of the Brazilian poor are rural residents and half reside in the Northeast, the trickle-down argument is muted bv the recent performance of the Brazilian economy in the 1970s and realistic forecasts for the 1980s. 212 FUTURE OF BRAZILIAN DEVELOPMENT 213 Whereas the economy grew at rates averaging 10. 1 percent from 1968 to 1974, the rate has slowed to an average of 6 percent since 1974.' With increasingly limited opportunities for import substitution, skyrocketing oil prices (Brazil is straitjacketed by its dependence on imports for 80 percent of its supply), increasingly tight external credit because of Brazil's massive outstanding foreign debt, and further curtailment of export markets as a result of trade limitations and lagging demand in the industrialized world, the growth prospects for Brazil in the 1980s are not optimistic. Thus, there are two serious doubts in the equation: Do benefits actually trickle down, and will there be sufficient growth from which the trickle may come? Beyond the 1980s is another matter. Brazil's natural resource base pro- vides good cause for longer-term optimism. The vast-and largely un- tapped-resources in the Amazon basin could well provide for another period of miracle growth. And, given the proximity of the Northeast, the army of available labor could easily be absorbed. But will this occur, given the growing sensitivities about the international ecological ramifications of the Amazon basin? When will this occur, and will the Northeasterner have acquired the education and skills to make him employable in such ventures? Alternative Solutions Certainly, alternative solutions are possible and, in some cases, feasible. A few of these options are discussed in this section. Education and skill training to prepare workers for out-migration are probably the best possibilities for a longer-term solution in the absence of a structural reform. Land, natural resources (except oil), and unskilled labor are Brazil's most abundant resources. The shortage of an educated and skilled labor force has been, however, and will continue to be a critical constraint on the economy's growth. The underutilized human resources of the Northeast constitute a potential supply source. Can the Northeasterner be educated or trained to fill the need? Within a generation, the answer is probably no. Such a goal is not likely to be achieved in the near future. Decades of nutritional deficiencies, illiteracy rates of 75 percent, and sadly neglected educational infrastructures have culminated in a monumental socioeconomic challenge. That "there is a sense of urgency about education 1. Werner Baer, "The Brazilian Growth and Development Experience: 1964-1975," in Brazil in the Seventies, ed. Riordan Roett (Washington, D.C.: American Enterprise Institute for Public Policy Research, 1976), pp. 46-47; and World Bank data. 214 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL in Brazil" is recognized,2 but how long will such a program take and how much national, human, and physical capital must be diverted to accomplish the goal? A human resource solution is, at best, decades away, and the national sacrifice of diverting skills and infrastructural investments from other, more pressing endeavors does not seem to be on the horizon. A more favorable national policy toward agriculture could increase the incentives toward fuller utilization of land as well as increase employment. These strategies, in turn, would include devaluing the cruzeiro (to make the export crops more profitable), stabilizing inflation, providing investment incentives to agriculture, and intensifying research to develop new higher- yielding and more drought-resistant varieties. Of these various steps, the first two involve macroeconomic policies that would have sweeping effects on all sectors and regions. The narrow focus of this study does not warrant recommendations on such policies. It seems clear, however, that they will be addressed in terms of the wider goals of Brazilian developmental prob- lems rather than in terms of the narrower Northeast problem alone. The promotion of investment in agriculture is a multifaceted topic requiring a far more detailed analysis than is possible here. Agricultural investment in Brazil has been highly subsidized through loans with often negative real interest rates, but these policies have had mixed effects on both productivity and income distribution. In the Northeast, although subsidized investment has benefited the large-scale farmers, there is no evidence that it has increased productivity and resource utilization (particu- larly labor). The small-scale farmers and sharecroppers have probably benefited only from public investment in infrastructure. To recommend increased investment, therefore, raises the question of which type of invest- ment is desired-labor-saving in an area of surplus labor, or land-saving in an area of underutilized land? Leaving this question open, this study finds that the problem is concentrated in distortions in the markets for the two factors that are most abundant-land and labor-and not in agricultural capital. Agricultural research has been advancing vigorously since the inception of EMBRAPA and promises to yield favorable results. The question, however, is still this: Will the gains be sufficient to alter behavior patterns so mark- edly that land will be more intensely utilized and labor more intensely employed, given the proclivity of large-scale farmers to react less than optimally to economic opportunities? The technical progress experiments reported in Chapter 7 indicate that this may not be the case because small- 2. RobertJ. Havighurst andJ. Roberto Moreira, Society and Education in Brazil(Pittsburgh, Pa.: University of Pittsburgh Press, 1965), p. 251. FUTURE OF BRAZILIAN DEVELOPMENT 215 and medium-scale farmers mav increase production of the price-flexible crops, whereas large landowners may restrict land and labor use to roughly maintain current income levels. Those results depended on the assumption that the new seed varieties and technology would be adopted by all farmers in the region. Yet the reluctance of small-scale farmers and sharecroppers- whose decisions to adopt new varieties may be dominated by safety-first criteria and who do not have the access to credit or the desire to borrow for new technology-may severely limit the potential benefits of agricultural research. This reluctance, combined with the motivations of large land- owners, could well mean that successful agricultural research would bring income gains to the more progressive medium-scale farmers, consumption gains to consumers, and no gains at all in the welfare of the small-scale farmers. In short, agricultural research is strongly warranted, but it cannot be a panacea, given the structure of the agricultural economy of the North- east. Nonagricultural activities in the rural Northeast could also alleviate underemployment and poverty in the region. Certainly, the dispersion of the 34/18 investments in industry away from the coastal cities, accompa- nied by more labor-intensive criteria for the subsidies, could absorb much of the labor force and induce secondary, multiplier effects. Apart from the agricultural processing industries, however, what comparative advantage, from a macro view, lies in such a dispersion? The paucity of nonagricultural resources, the location of markets, and the inadequate transport system undoubtedly override the advantages of cheap labor. On economic grounds, dispersion of industry into the rural Northeast makes little sense; on distributional grounds, there are other means available that need not run counter to regional comparative advantage. The possible solutions to the Northeast problem mentioned above could well transform the rural Northeast from an impoverished, largely illiterate, underproductive agricultural region into one that is economically diversi- fied, self-sufficient in food production as well as in educated and skilled labor, and profitable in terms of foreign exchange. The lack of national commitment to do so, however, the many "ifs" involved, the decades of time and enormous amount of effort required, and, from a national point of view, the far more pressing developmental needs competing for the scarce capital and human resource skills make this a remote possibility. There is no short-term solution. But there is a short-term problem. More than half the natural annual increase in the Northeast's population migrates to the cities of the Northeast, the Center-South, and Southeast. These migrants, or retirantes, are unskilled and unprepared for absorption into the economies of Belo Horizonte, Rio, and Sao Paulo and tax the already 216 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL overburdened services of these cities. Leaving families behind, thev are frustrated and unwanted. They find refuge infavelas (squatter settlements) and obtain menial employment at best. And then, with news of "rain in the Sertao," the retirantes often return to the Northeast. Potential Role of the Rural Northeast The nearly I million viable farming units and sharecropped plots clearly clemonstrate that the agricultural Northeast does provide a standard of living of at least acceptable standards for those favored with land. The nearly 30 million hectares of underutilized land of similar if not superior quality on the estates indicate that nearly another million families could achieve comparable living standards. On egalitarian grounds alone, a land reform is strongly warranted and provides the best prospect for swift rises in the living standards of the most impoverished groups in the rural Northeast. Egalitarian arguments, however, are not the chief determinants of policy in dynamic, resource-rich economies with virtually unlimited potential, especially when protagonists of the trickle-down theorv are vociferous. Nor should egalitarian principles necessarily shape policy, unless political ex- igencies so dictate. The solution to the Northeast problem needs to be considered in the context of the long-term development of Brazil. True, the Northeast problem has existed and intensified for over a century, but the coming decades will undoubtedly alter all the justifications, motivations, and apolo- gies for past policies toward the Northeast. A new role for the rural Northeast must be found, and that role must be compatible with Brazilian developmental goals and, subsequently, contribute to them. That role will not lie in the monocultural production of crops for export. Sugar no longer has a comparative advantage. Cacao and the oil palms stiffer from limited markets. Cotton remains a good prospect, but its labor intensity makes it most attractive to family farms or to estates using sharecropped labor. Nor will the Northeast become a major livestock producer; natural pasture is too poor to support more head of cattle under current or envisaged practices, and more advanced production of cultivated grass for fodder raises questions about alternative uses of land-such as for food crops. The green revolution has to this dav barely touched the Northeast, as is the case with most semiarid tropical zones in the world. If varieties and FUTtURE OF BRAZILIAN DEVELOPMENT 217 techniques are found that are amenable to Northeastern conditions, it is doubtful whether Northeastern farmers would adopt them, given their risk-averse behavior and lack of funds for investment. Newly discovered methods would have to be promoted and subsidized to a far greater extent than hitherto evidenced in Brazil. The role of the rural Northeast, for the foreseeable decades ahead, reduces to a simple, yet vitally important one: to increase the traditional foodstuff production of corn, beans, manioc, and rain-fed rice to supply the rural and urban population of the Northeast and, ultimately perhaps, permit some excess to be exported to the South; to continue present levels of production of cacao, cotton, and oil palm for domestic consumption and exportation; and, finally, to increase livestock production through more intensive management of pasture. These objectives can be achieved simply through the more rational allocation of labor to land by substantially increasing the number of family farms of the type defined by INCRA. And this move can be abetted by strengthening the infrastructure and advancing research on appropriate technologies. These would be the direct benefits of a land reform. The indirect benefits, however, may be far more important to the economic and political stability of Brazil. Even when the Brazilian economy was growing at an annual rate of 10 percent, it could not absorb the masses of migrants to its Center-South cities. Now that growth has slackened and opportunities for migrants returning to the Northeast have improved little if at all, the urban problem continues to grow more acute. According to a recent estimate, 1.2 million people live in the Rio favelas,3 and probably far more live in the Sao Paulo favelas. Most have come from the Northeast. The reasons behind such migrations are complex but generally encompass economic motives: The realities of life in the rural Northeast make the uncertain yet potentially better life in the Center-South cities attractive enough that those who can move do so. Unless and until the migrants are fully prepared for productive lives in modern, urban Brazil, however, they would probably be better off remaining in the Northeast. To this end, the most effective incentive is the provision for sufficient, productive land. A reduction of rural-urban migration is an economic, if secondary, benefit of land reform. A noneconomic benefit of potentially as great importance is the lessening of political tensions. Such tensions have dimin- 3. "Brazil's Outcast Shanty Dwellers Offered a Helping Hand at Last," New York Times, October 22, 1979, p. A4. 218 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL ished considerably since the 1950s and early 1960s, and in the 1970s were not widely publicized. Yet they are still manifested in a sense of frustration, though not necessarily as a potentially disruptive force.4 Land reforms in other countries have bought time during which develop- ment in other regions and political reform could begin to rectifv the structural problems and make further calls for land reform redundant. Mexico is a prime example: The reform bought time and a measure of political stability, and as an additional benefit agricultural production increased.' Whether Brazil needs to buy time for a Northeast solution, and whether rural-urban migration and its effects pose important problems, are ques- tions for Brazilian policymakers to assess. This studv has concentrated on the economic aspects of the problem, and the economics of a land reform in Northeast Brazil can be summarized very briefly. Four percent of landowners own more than 50 percent of land in agri- cultural properties; only one out of four agriculturally dependent families owns the land it works. The smallest farms emplov twenty-five times more labor per hectare on their land than do the largest farms and obtain vastlv higher productivity levels. The smaller farms (less than fifty hectares), though only 10 percent of the agricultural land, produce over 25 percent of the region's sugar, cotton, and rice and 40 percent of the beans, corn, and manioc. Yet 2 million agriculturally dependent families own no land at all while an area of land the size of France is un- or underutilized. 4. "Osvaldo Galvao, a rancher in [Piaui] was asked if he was encouraged bv the announce- ment last week from Brasilia [that the government was undertaking a US$330 million program of well drilling and small dam construction]. le pulled a used soft drink bottle from a rack, held it upside down over a counter and watched several drops form a rivulet. 'That's how much we'll get, the rest will stay in the south as it always has. ""Brazil's Arid Northeast Skeptical as Hlelp Is Promised Once Again," ANew York Times, September 21, 1979, p. A13. 5. Shlomo Eckstein, Gordon Donald, Douglas Horton, and Thomas Carroll, Land Reform in Latin America: Bolivia, Chile, Mexico, Peru, and Venezuela, World Bank Staff Working Paper no. 275 (Washington, D.C., April 1978), p. 72. Introduction to the Appendixes The four appendixes that follow describe how the survey was carried out, from the design of questionnaires and samples to the fieldwork and the organization, validation, and retrieval of the computerized data. They are by no means intended as a manual for agricultural surveys in general; they simply record the methodology for this study. Although some members of the research team had strong backgrounds in statistics, they were all economists and, aside from one short-term consul- tant, none had more than a passing acquaintance with the mechanics of large surveys. In retrospect it appears that the team's complementarity of skills largely overcame its lack of experience in such endeavors. The Brazil- ian members included agricultural economists and agronomists with inti- mate knowledge of Northeastern agriculture, and computer programmers who could identify potential computational bottlenecks. The World Bank members had expertise in sampling theory, mathematical planning tech- niques, and procedures for organizing and manipulating data. Most of the team had studied problems of agricultural economics, and some had taken part in large agricultural-sector planning studies. This combination of skills meant that many potential pitfalls were avoided. On the advice of computer experts, for example, questionnaires could be designed to facilitate the computational organization of the data once collected; with the help of a logistics expert, the sample was designed to ensure that interviews could actually be conducted as planned. 219 Appendix A Questionnaire Design 'Though we had originally intended to use a questionnaire used in a 1972 pilot survey by SUDENE and the Institute of Economic and Social Planning ([PEA), Ministry of Planning,' we decided, for several reasons, to construct an original one instead. First, the IPEA/SUDENE questionnaire did not yield the detailed input-output information necessary to permit the use of linear programming techniques at the farm level. Second, we wished to use a precoded system, both for the names of crops, implements, and so on, and for the actual data themselves: To speed processing and eliminate the step of transferring the information from the questionnaire to coding sheets, the questionnaire itself had to be in the form of coding sheets so that it could be keypunched directly. Third, as we consulted other institutions that might have an interest in the survey, we found that the informational content of the IPEA/SUDENE questionnaire was inadequate.2 Design Criteria The new questionnaire was designed to maximize the information obtained subject to constraints on ease of processing and cleaning, inter- viewee fatigue, and logistics. Other considerations related to the diversity of ecological factors, cropping patterns, tenure relations, and ease of train- ing enumerators also had to be recognized. Some of these constraints affected the sample design and are discussed in Appendix B. The general approach was to group the desired information into tables (for example, on land use, production, distribution, equipment), the col- umns of which corresponded to fields on an IBM data card and the rows of which corresponded to individual data cards. The tables were organized 1. IPEA was originally to have assisted SUDENE in a studv of the "typical size of the unit of production in the Northeast," or "tamanho tipico." Staff shortages were largely responsible for its withdrawal. 2. Lee Bettis, consultant to the World Bank, assisted in drafting this section. 220 QUESTIONNAIRE DESIGN 221 into a unified questionnaire by proceeding from the general to the specific. Thus, one of the first tables contains questions on land use, subdividing the farm into cropland of different types, pasture, and so on. A following table records the use of cropland by crop, showing the age and structure of perennial crops, gross output, and so on. Then data on the outputs are tabulated to show how they are distributed among consumption, sales, pavment in kind to laborers, inventory changes, and so forth. This method of organization not only permits a logical sequence of questions but also allows information obtained at various times during the interview to be cross-checked. These cross-checks were deemed necessary to ensure the consistency and quality of the data. Figure A-1 is a representative table from the questionnaire covering sources and uses of the land on the farm. The table columns correspond to IBM card format; the small numbers 1-80 are the columns in which the data are directly keypunched. Each row of the table, if not blank, would be punched on a different card. Each card has a unique identifying number in columns 78-80; for example, for every farm surveyed, card 002 will always be the one for "total area." Columns 72-77 contain codes for state, muni- cipio, and farm, so that each card from the entire survey will have a unique identifier to permit computerized sorting and provide insurance against card "shuffling." The items across the top, numbered in parentheses, are sources and modes of administration of each of the categories of land that are defined in the rows. Column (1) is the size of the property as given by the INCRA cadastral survey from which the farms in the SUDENE/World Bank survey were identified. If other properties belonging to the same owner were close by and were managed by the same person, they would appear in column (2). Columns (3) and (4) are for parcels rented or squatted upon, respective- ly, if, again, they are part of the same entrepreneurial unit. The administration of land is divided into "direct," "sharecropper," "morador," and "rented out."' The farm is defined as the sum of the land under the first three types of administration, and this definition is used consistently throughout the study. A within-card consistencv check is built into the table because total area under the sources, (1) + (2) + (3) + (4), must equal that under the administration types, (5) + (6) + (7) + (8). Rows 2 through 11 allocate the total land area in row 1 to the possible uses. Here, across-card consistency checks are possible because the sum of rows 2-11 must equal row 1. A later table on crop-specific data provides a 3. Moradores are permanent workers who reside on the farm and are often allocated small plots on which their families plant subsistence crops. Figure A-i. Representative Table from Questionnaire I .iDn ~0J ld J 4,id,s¢mono/ o x Ares VrLa A>a ()her I)~~06- i-cl By ,0)hj, ~~~~~ ,ooood ~ -pw d fo moe propenyv p..p0rO0s in ,.c i.ord .nppe.. m~or dooz, (1) (2) 05) (5) I I 0)(5) (7) /.X.1. ' t1'1 4 PI-IdO I- 0 Ak.~,h 000, 10 0-riao(b, -1,o 10~ 000 I--- - - 9z F"i-.. -)0 QUESTIONNAIRE DESIGN 223 check across tables, because the total area in crops identified there must equal the sum of rows 2, 3, and 4 in the table shown as Figure A-i. Field Tests Field tests were carried out by members of the World Bank, SUDENE, and the Federal University of Pernambuco in the state of Pernambuco, which has areas representative of the three major physiographic zones of the Northeast-the Zona da Mata, Agreste, and Sertao-and all the types of farms likely to be encountered, ranging from minifundia to large sugar plantations, and including all known tenurial types. The most significant change in the questionnaire dictated by the field tests was a substantial reduction in its size, as the first version, it was found, could take up to eight hours to apply! Some tables were eliminated and others combined. These first field tests also pointed out the need to enlarge and make more precise the activity and other codes and to sharpen the definition of terms used. Though doing this reduced enumeration time itself, the questionnaire still needed an average of four hours for each interviewee. We adopted, as a rule of thumb, a limit of two hours for the interview, both to prevent interviewee fatigue, and thus possible distortions, and to permit each enumerator at least two interviews a day. Hence, the questionnaire was divided into two parts that could be applied independently but could be linked through overlapping tables. On the assumption that both parts were applied to identicallv drawn random samples from the same stratified universe, this overlap allowed the two parts to be linked later. The questionnaire was divided according to the use that was to be made of the data: The elements of a farm-level programming model can conveniently be divided into (1) the objective function and vector of resource constraints and (2) the technology matrix. Furthermore, we assumed that, for a given population of farms, the various data related to the objective function (output and input prices, transportation costs, and so on) and resource constraints (land of various types, equipment credit) would show greater differences among farms than would the technology data relating outputs to inputs. This assumption appears entirely plausible in the context of the Northeast, where technology is virtually all of a fixed proportion (man/hoe). It permits a much smaller number of questionnaire 11 (technology-oriented) than of questionnaire I to achieve a given level of statistical significance.' 4. For a precedent for this type of sample division, see United Nations, Food and Agricul- ture Organization, Coffee in Latin America, parts I and 11 (Rome, 1960). 224 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Questionnaire Content Questionnaire I contains questions on the identification and location of the farm, land tenure, off-farm income, production incentives, hired labor supply, and prices. Responses to these questions assist in establishing the setting within which the farm operates. The information from question- naire I would also allow the construction of individual farm budgets, though this did not form part of the study. Table 0 (zero) of questionnaire I provides information on the interview- er's travel time and mode of transportation to the farm and the duration of the interview. These data permitted cost and productivity analysis of the survey itself. The last table, Table 14, is a census of sharecroppers from which to select units to survey separately later (see Appendix B). Table 2 deals with transactions in farmland in 1972 and 1973. It includes purchases, sales, and rentals in terms of area and value. This table is useful for further characterizing the farm setting and for checking the internal consistency of the responses. Table 2A provides information on the land tenure system and distribu- tion of the total farm area among broad categories of use, measured in area and value. This table gives the interviewer an immediate overview of the type of farm he confronts and compels him to be consistent about the total area of the farm throughout the later tables. Table 3 contains data for estimating the stock of fixed capital including houses, barns, pens, roads, and dams; it captures the size, useful remaining life, market value, and annual service flow of each structure. Table 4 provides information on the stock of farm machinery and equip- ment and the annual flow of services from it. Together, Tables 3 and 4 provide the data for calculating the amount and value of capital owned and rented by the farm. Table 5 details information about land use according to cropping patterns and type of administration, that is, whether the farm is managed directly by the owner or by sharecroppers and hired laborers. Individual crops and consortiums (interplantings) are specified with great accuracy. This table provides the basis for several tables in questionnaire II and is thus of critical inportance to the survey. Table 6 records the disposition of the crops harvested in 1973 (which were specified in Table 5) showing the quantities used for family consump- tion, nonwage payment, feed, processing, and those retained as stocks. The QUESTIONNAIRE DESIGN 225 table also indicates the quantity and market value along with unit price per kilogram of each crop marketed in 1973. Table 7 contains data about the stock of livestock and poultry on the farm, actual livestock transactions during 1973, and all other possible disposition alternatives. Both quantity and value measures according to the breed and age of the animals are specified. It was difficult to obtain accurate information for all parts of this table because of the primitive nature of livestock markets in the Northeast. The aim was to collect the minimum essential information required by the model without overtaxing the inter- viewee's ability to respond or jeopardizing the completion of the table. Table 8 records the quantity and value of agricultural commodities processed on the farm, such as flour, meat, and milk and other dairy products, and their disposition, as in Table 6 for crops. The table also records the amount of processed products sharecroppers produced and delivered to the owner. Tables 9, 10, and 11 cover the sources and uses of labor in detail. Table 9 begins with a census of all persons occupied at least one day of the year on the farm, in any of eighteen categories of labor. It also provides the number of days worked on the farm and off and the monthly pattern of off-farm work. Remuneration, both monetary and in kind, is also recorded. Table 10 gives a monthly breakdown of the number of workers and man-days worked for several types of temporary workers and sharecroppers. Table 11 relates the cropping pattern recorded in Table 5 to the use of labor for crops and maintenance of pasture. For each cropping activity, the number of man-days employed is obtained for subactivities: preparing soil, planting, cultivating, fertilizing and applying insecticides and pesticides, irrigating, harvesting, and transporting to market. Table 12 lists each specific input (for example, fertilizer, seeds, pesti- cides, forage, veterinary services) used on the farm during the previous year, showing the quantity produced on the farm, the quantity obtained from elsewhere, and its price, if purchased. It also records whether the use of the particular input had increased, decreased, or remained the same over the previous year. These input data are recorded only in gross terms at this stage; Table 19 in questionnaire II identifies them by activity. Table 13 collects detailed information about the sources and value of credit used for farm production, according to activity and the term and length of the loan, and for loans used for operating expenses or for invest- ment, according to the type of loan. To summarize, the thirteen questionnaire I tables on the farm's produc- tion and exchange relations can be grouped under six headings: 226 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL -land tenure system and land use (2, 2A, 5) -improvements, machinery, equipment, and animals (3, 4, 7) -crop production and on-farm processing of livestock and crop products (5, 6, 8) -labor and other inputs (9, 10, 12) -financial capital (13) -technical coefficients (11). The first group describes in quantitative terms the relations among the various production agents, as revealed by the allocation of farmland among them. The second group records information on fixed capital. Together, the first two groups provide estimates for the constraint vector of the linear programming model. The third group quantifies the various production activities to be represented in the model. The fourth group covers the sources and uses of labor, which is probably the crucial element of the model and of the entire project. The information on labor was used to define the monthly requirements for labor, from which we determined the extent of seasonal unemployment. Information collected on the use and costs of fertilizers, pesticides, and so forth were combined with the gross revenue coefficients of the objective function to calculate net revenue. The information on financial capital recorded by the fifth group was used for several purposes; for the linear programming model it shows the equity limits and the areas of investment for financial capital (credit) prevailing in the Northeast. The information from Table 11 about input-output coef- ficients for the main production activities was used in the definition of the criterion used in the imputation of a more detailed technology to the farm. Questionnaire II is designed to collect the information necessary to define the input-output matrix of coefficients for all production activities conceiv- able on a farm in the Northeast. It records very detailed data on yields, on inputs such as labor, fertilizer, seeds, machinery, equipment, and fuels, and on other operating costs, including interest on outstanding loans. Some tables are common to both questionnaires: 0, 2A, 5, 7, 9, 12, and 14, which appears as Table 23 in questionnaire II. They were used in the question- naire II sample to delineate the structural context within which the indi- vidual activities are undertaken, and to provide the main elements for imputing a matrix of techniques to each farm in the first sample. Tables 15 through 20 of questionnaire II collect data on specific activities for use in building up the technical coefficients. Table 15 deals with the labor, machinery, and equipment requirements of individual crop and pasture activities. The number of man-days supplied by different catego- ries of workers and the pieces of equipment used in each month for each QUESTIONNAIRE DESIGN 227 activity are specified for broad categories of a production activity, for example, preparing soil, planting, and harvesting. Table 16 asks for information on the production of livestock and poultry on the farm for each month of 1973, including the number of work animals, cattle, and cows pastured and fed silage and the quantity of feed consumed by each animal. Questions are also asked concerning the quantity of swine, sheep, goats, and fowl slaughtered annually. Table 17 records the labor requirements for each month in 1973 for processing agricultural commodities. Labor is disaggregated by category in Tables 15 and 9. Table 18 collects data on the monthlv labor requirements (also disaggre- gated by category of workers) for constructing and maintaining fixed capital on the farm. Table 19 complements Tables 15 and 16 with data on the use of produc- tion inputs other than labor, crop, and livestock activity. Inputs include seeds, feed, fertilizer, veterinary medicine, fuel, and other operating ex- penditures. Table 20 asks for information on the number of hectares used and the total output of each crop per activity from land of high, medium, and low productivity. Information on land quality, vital for a well-specified linear programming model, was sought in the following way: First, a pedologist studied soil quality in the sampled municipios; second, each farmer was asked to classify the land on his own farm according to his perception of its profitability by crop or consortium. By merging the results from the two sources, soil classes could then be delineated within each sampled area. Tables 21 and 21 A contain data on marketing and storage activities for each product (raw or processed), showing the quantity sold, price, form of payment received, and mode of transportation and distance to each type of buyer (whether landowner, intermediary, cooperative, processing plant, or final consumer) and also, for each product, the amount stored in owned or rented quarters, the type of storage structure, the storage period, average loss from storage, and sales price. Finally, Table 22 furnishes information on water availability and quality and its influence on production. Questions are asked about the rainy months in 1973 in the area; whether there was a dry spell or flooding and which crops suffered most; what sources of water were used; in which months they were dry; how much land was lowland, hillside, and highland; and which crops did best on these types of land. Appendix B Sampling Procedures Several critical factors had to be considered in designing the sample: The Northeast is a very large area (1.548 million square kilometers), and trans- portation and communications can often be quite difficult; ecological condi- tions and agricultural production vary widely, as do farm sizes and tenurial systems; and, because there were two questionnaires, two samples were needed. To reduce logistical problems, we sampled clusters of farms scattered throughout the region. Our primary sampling units were the administra- tive subdivisions called municipios or counties, of which there are 1, 365 in the Northeast. There were two broad choices for the selection of munici- pios: purposive sampling and stratified random sampling. Several past surveys in Brazil have employed the former method, selecting "typical" inunicipios on the basis of available data, intuition, and the judgment of field experts. ' This technique is likely to introduce biases into the analysis, however. It will tend, in particular, to select larger primarv units, as near as possible to the observers' concept of the "average," with a much smaller proportion of extreme cases than would be obtained in a random selection. A nother shortcoming of purposive or other nonrandom sampling methods is that no statistical theory exists for measuring the reliabilitv of their results.' Because of these shortcomings, we adopted a stratified random sampling procedure that determines the strata state by state. In this procedure the primary units (municipios, or clusters of farms), are sampled in two stages: First, the primary units are selected by weighted random sampling in a two-wav stratification; second, the units are selected by weighted random sampling in a one-way stratification. Subunits (farms) are then selected from each of the selected primary units by stratified random sampling. 1. For example, recent surveys coordinated by the Ministry of Agriculture and IPEA of the Ministry of Planning. See Ruy Miller Paiva and George F. Patrick, Acesso a terra no Nordeste, Colecao Relatorios de Pesquisa (Rio de Janeiro: IPEA/INPES; processed). 2. See Morris Hansen, William tlurwitz, and William G. Meadow, Sampk Survey Metbods and Tbeory, vol. I (New York: John Wiley and Sons, 1953), pp. 71-73. 228 SAMPLING PROCEDURES 229 In the first stage of the sample, we divided the 1,365 municipios of the Northeast into strata according to (1) their ecological characteristics, using the 114 ecological microregions of the Northeast defined by IBGE, the Instituto Brasileiro de Geografia e Estatistica, and (2) their production characteristics, by considering the types of crops (or livestock) that are their main sources of agricultural revenue. A weight was then assigned to each municipio to reflect its economic size or importance as an agricultural producing unit. Using this ecological/production pattem classification and the weights, we selected a sample of municipios at random. The number of municipios selected in each state depended on: the number of ecological/ production pattern types represented in that state (we required the sample to include at least one municipio of each type in each state); our enumera- tion capacity; and the extent to which average farm size varied among the municipios of that state. The first-stage sample of municipios that resulted was too large for our enumeration capacity, and thus a second-stage random sample was drawn, stratified only by ecological microregion, not by pro- duction pattern, again using weights corresponding to economic size. Having selected the municipios, we identified a preliminary sample of farms-"preliminary" because several modifications were made once the early results had been evaluated. The available data consisted of a cadastral survey of agricultural properties undertaken by INCRA, the National Insti- tute of Colonization and Agrarian Reform, in 1972 and estimates of the enumeration time of questionnaires I and II from the pretests. Farm size was taken as a proxy for the range of factors limiting farm performance- the principal consideration of questionnaire I-and we assumed, initially, that the variation in the technical coefficients-the principal consideration of questionnaire lI-could be related proportionally to the variation in size. To stratifv the sample of farms, establishments smaller than 500 hectares in each selected municipio were divided into five size classes,3 all of which were equal to one another in total area. Establishments of 500 or more hectares were placed in a sixth stratum. For each state, an enumeration team was available for eight weeks. The problem was to allocate their total enumeration time among the municipios selected in that state; among the strata (farm size classes) within each municipio; and between questionnaires I and II. The allocation rule that resulted for each state in sample S,S (where m denotes the municipio, q the questionnaire, and s the stratum) depended on: the relative enumeration time needed for questionnaires I and II; the 3. Up to 9.9 hectares; 10-49.9 hectares; 50-99.9 hectares; 100-199.9 hectares; 200-499.9 hectares. 230 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL variance of farm size within each size class in each municipio; the number of municipios selected; and the total enumeration time available. A random sample for each state was drawn with this rule, together with a backup sample of equal size, to be used for replacing farms that, though recorded by the cadastral survey, could not be located or did not exist, and for extending the sample in cases where the estimates of enumeration time 'were found to have been too high. Thus, the actual sample could be greater or less than Smqs, depending on the actual time spent in traveling, locating farms, and conducting interviews. The size of the actual sample in each state also depended on the estimated differences in the degrees of variance in the two groups of items being investigated by questionnaires I and II. As explained below, these differences determined the appropriate ratio be- tween the sample size for questionnaire I and that for questionnaire II, and they were estimated with progressively greater accuracy as completed questionnaires were analyzed. To summarize, the procedure followed used the general rule that sam- pling should be concentrated where the sampling unit is larger, where the sampling unit is more variable internally, and where the sampling is cheaper in the sampling unit. Selection of Municipios For the first-stage selection of primary sampling units, we hypothesized that the variances of the random characteristics of interest to the study are (letermined by two basic factors: ecological conditions (climate, soil, lati- tude, altitude, and so on) and the orientation of production. Ecological differences would be responsible for the variance between groups of farms that are not contiguous. Differences in production patterns would be responsible for the variance within such groups. Stratification according to both factors would then ensure that the sample of farm clusters would reflect both the ecological and the agricultural diversity of the universe, that is, the state. A two-way stratification, of municipios based on these two factors was made using the microregions defined in physiographic terms by IBGE5 and 4. For a discussion of two-way stratification, see William G. Cochran, Sampling Techniques (New York: John Wilev and Sons, 1963), pp. 126-28. 5. Instituto Brasileiro de Geografia e Estatistica, Dadospreliminaresgeraisdo censoagropecuario, VIlII: Recenseamentogeral, 1970 (Rio deJaneiro, 1972). Our subareas do not correspond exactly to the 1972 census-defined microregions; in some cases we aggregated these microregions, in others the definition was altered in the final, published census. SAMPLING PROCEDURES 231 Table B-i. Production Patterns Monoculhures A. Arroz (rice) B. Algodao (cotton) C. Cana de acucar (sugarcane) D. Coco (coconut) Consortiums E. Mandioca, arroz, milho (manioc, rice, corn) F. Algodao, arroz, mandioca (cotton, rice, manioc) G. Algodao, arroz, milho (cotton, rice, corn) H. Algodao, arroz, feijao (cotton, rice, beans) 1. Algodao, feijao, milho (cotton, beans, corn) J. Algodao, feijao, banana (cotton, beans, banana) K. Algodao, feijao, mandioca (cotton, beans, manioc) L. Feijao, mandioca, mnilho (beans, manioc, corn) M. Outros (other, mixed cropping) N. Pecuaria (livestock) an unpublished set of SUDENE data on the main agricultural activities at the municipio level for 1964, because no more recent data were available at this level of disaggregation. The municipios were classified according to the following production patterns, listed in detail in Table B-1: -monoculture, where a single crop was the municipio's main source of agricultural revenue -consortium, where a single consortium of intercropped products was the main source of agricultural revenue -livestock, where livestock was the main source of agricultural revenue -other, where there was a diverse production pattern in which no single activity yielded more than 50 percent of the municipio's revenue. Because the municipios are administrative units, they differ significantly in their importance as agricultural producing units. To take account of this, we applied a technique suggested by Hansen, Hurwitz, and Meadow, whereby the municipios are given a probability of selection Z, that is proportional to their degree of importance as agricultural producing units.6 In determining these probabilities, we faced two problems: First, the Zi were not readily definable and, second, since the purpose of the survey was to seek information on many different variables, it was necessary to con- sider various possible measures of economic importance in selecting the 6. Hansen, I lurwitz, and Meadow, Methodse and Theory. 232 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL units to be sampled. For cases in which size Zi is some aspect of the physical size of the unit, Cochran suggests using the measure of size that is most nearly proportional to the unit totals of the principal items of the survey.' We can thus assign to each municipio within each ecological/production pattern group a probability weight that is proportional to an index of a group of variables related to the principal variables of the survey. Prelimi- nary data from the 1970 census were used to make the following selection of variables, where i indicates the municipio and j the stratum (ecologi- caUproduction pattern group): Qyj = population; Hij = area; PIj = em- ployment; and Fj = number of farms. The weight of the ith municipio in thejth stratum is: (B I) Wi/=~~ /4Pi + Hij + Pij + Fijl (B. 1) WH PE1! Q H P F/ where the bars indicate statewide means. Then the probability of drawing the kth municipio in thejth stratum is defined as: Pkj = Wkj k- Wkj k = I where M is the total number of municipios in the jth stratum. The first-stage sample that resulted from this procedure consisted of one inunicipio from each ecological/production pattern group. It was much too large for our enumeration capacity. In reducing the sample, we decided that it was important to maintain coverage of all the microregions within each state but that the differences in production patterns within microre- gions were less important. I A second-stage random sample was thus carried out and yielded one municipio per microregion per state. Anv excess enumeration time that permitted more municipios than this to be sampled was allocated to municipios in those zones that showed the highest variance in certain key variables that were to be sampled. Because we had no previous data about which were the key variables, available proxies were taken from the preliminary 1970 census data: establishment area, employ- 7. Cochran, Sampling Techniques. 8. The relative gains from stratification by zone, calculated in terms of Hansen's analysis, (Hansen, Hurwitz, and Meadow, Methods and Tbeoty, pp. 200-01), proved to be substantial and are reported in Gary P. Kutcher and Pasquale L. Scandizzo, "Northeast Brazil Regional Studies Project: Sampling Procedures" (World Bank, Development Research Center, 1974; processed). SAMPLING PROCEDURES 233 ment per establishment, and farm density (number of establishments per square kilometer).9 From the microregion that showed the highest total variance of all these measures, a second municipio was selected.'o This selection procedure was repeated, using the next highest variance, until the expected enumeration time was fully accounted for. The derivation of the probability weights used in the procedure is described in the last section of this appendix; the municipios finally selected are shown in Table B-2. Selection of Farms The framework established for the selection of farms thus ensured that all the ecological microregions of the Northeast would be represented by at least one cluster of farms and that those microregions in which agricultural activities are more diverse would be represented by more than one cluster of farms, in proportion to their degree of diversity. The farm sampling problem was then to allocate the fixed amount of enumeration time among the clusters of farms (municipios); farms of different size classes within the clusters; and the two questionnaires. We consider the last of these items first. Of the two broad groups of variables of interest in the survey, question- naire I covered the limiting factors on farm performance or, in a linear programming framework, the parameters of the farm's objective function and its constraint vector. Questionnaire II covered the technical coefficients and technological relations, or the matrix elements in linear programming terminology. We assumed that the information relative to the two groups of variables can be conveniently summarized by two proxy variablesy, andy2. These two variables may be an aggregate index of the variables in the groups, or they may simply be highly correlated with them. To give an example, for the first group a variable of this kind would be the total area of the farm because one would expect a close association between area and the endowment of most of the other limiting factors. For the second group, a possible choice is an index of the diversity of production or land use on the farm. 9. The Census Bureau defines an agricultural establishment as an unbroken parcel of land under a single management. 10. Total variance is defined as W n W, where fl is the variance/covariance matrix of the four variables considered, and W is a vector of normalizing weights, taken in this case to be equal to the inverse of the means of the same variables. 234 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table B-2. Selected Municipios, by State Alagoas (CIMN) Babia (DCKIWN) Ceara (BIJKGMN) 1. Coruripe (1) 1. Santana (1) 1. Araripe (1) 2. Cha Preta (2) 2. Sento Se (2) 2. Brejo Santo (2) 3. S. Jose da Leje (2) 3. Ibitiara (3) 3. Independencia (3) 4. Junqueiro (3) 4. Tapiramnuta (4) 4. Jaguaretama (4) 5. Sao Sebastiao (4) 5. Itambe (5) 5. Itapipoca (5) 6. Traipu (5) 6. Condeuba (5) 6. Acarau (5) 7. Agua Branca (6) 7. Senhor do Bonfim (6) 7. Aquiraz (5) 8. Barro Preto (7) 8. Itapiuna (6) 9. Jequie (8) 9. Ipu (6) 10. Mucambo (6) 11. Tiangua (6) 12. Poranga (6) Maranbao (AEFGMN) Paraiba (BIMN) Pernambuco (CBILMN) 1. Godofredo Viana (1) 1. Rio Tinto (1) 1. Jaboatao (1) 2. Pinheiro (2) 2. Pilar (2) 2. Limoeiro (2) 3. S. Benedito 3. Areia (3) 3. Sertania (2) do Rio Preto (3) 4. Coroata (4) 4. Campina Grande (4) 4. Aguas Belas (3) 5. Sambaiba (5) 5. Frei Martinho (5) 5. Tabira (4) 6. Barra do Corda (6) 6. Uirauna (6) 6. Cabrobo (5) 7. Imperatriz (7) 7. Mirandiba (6) 8. Moncao (8) 8. Serrita (7) Rio Grande do Norte Piaui (BFHIGMN) (BKIMN) Sergipe (DCKIMN) 1. Parnaiba (1) 1. Ceara-Mirim (1) 1. Porto da Folha (1) 2. Buriti dos Lopes (1) 2. Santo Antonio (2) 2. Aquidaba (I) 3. Aroazes (2) 3. Acu (3) 3. Propria (2) 4. Luzilandia (3) 4. Pedro Avelino (3) 4. Pedrinhas (3) 5. Capitao de Campos (4) 5. Areia Branca (3) 5. Araua (3) 6. Altos (5) 6. Caico (4) 6. Malha dos Bois (4) 7. Canto do Buriti (6) 7. Florania (4) 7. Frei Paulo (5) 8. Sao Joao do Piaui (6) 8. Apodi (5) 8. ltabaiana (5) 9. Jerumenha (7) 9. Sao Miguel (6) 10. Santa Luz (7) 10. Alexandria (6) Note: The letters in parentheses refer to the production patterns (see Table B-i) in each state. The numbers in parentheses refer to the microregion that each municipio represents. Microregions represented by more than one municipio are those that exhibited the greatest heterogeneity, as explained in the text. We assumed that the objective of the data collection within each state was to minimize the variance of the average measure ofy1 andy2, subject to the statewide constraint on our enumeration capacity. This objective may be expressed as: SAMPLING PROCEDURES 235 MLi [ I SIh + (B.2) min V(5, + y2) = . Y.W Zi L nih Nih mih Nihb] subject to M Li (1cib nih + 2Cih Mih) = C, i=I h=1 where Wi = a weight inversely related to the probability of selection of the ith municipio Nih = number of farms in the bth size class of the ith municipio L = total number of farms in the municipio Ni = 'Y Nih, where Bi = the number of farm size classes in the ith municipio Zih = NihINi 1S2th= relative variance" of yj in the hth size class of the ith munici- pio,j = 1, 2 nih = sample size (number of farms to be selected) in the hth size class of the ith municipio for the variables of group 1 mih = the same variables of group 2 M = total number of municipios in the state Cjh= cost (= time) of applying the questionnaire, relative to the group of variables represented byy1, hth size class, ith muni- cipio 2Cih = the same for Y2. Solving equation (B.2) with the usual Lagrangian techniques yields the following expressions for the optimal sample size in the hth farm size class of the ith municipio: (B. 3) nth = V;.S s n* i = 1, 2, . . . ,M i bi h = 1, 2, . . . B 11. In order to make the minimization in equation (B.2) meaningful, the two variablesy, and y2 or their variances have to be normalized so that their differences are independent of the unit of measure. This normalization was performed by using estimates of the mean of each variable. 236 THE AGRICIJLTURAL ECONOMY OF NORTHEAST BRAZIL (B. 4) m?bs = g 2Sihl 2Ch_ i = 1, 2, . M iM Lt , 2 h = I, 2, Li M Bi , I X 'g2Sbsl 2 Cib (B.5) * =i=lb=] n* M Li i-I h _IgV ISb hCh where gih = W Z,h; nih and m* represent the optimal allocation of the sample for each municipio and farm size class; and n* and m* the optimal overall sample size for a given state. To illustrate some of the implications of this analysis, consider its ap- plication to the state of Pernambuco, for which, as the first state to be surveyed, only minimal information was available on the values of the parameters. For Pernambuco there were sixteen people in the interviewing team, each available for only forty-eight days. Their employment during this period did not involve any variable cost; the limiting factor was time. During the pretest it had been established that questionnaire I took an average of about 0.65 day and questionnaire 11 roughly one day to apply. Although the variance of the variables in group 2 was actually expected to be smaller than the variance of the variables of group 1, for lack of direct information we first assumed it to be equal, that is, that I Sih = 2Sih for every i and h or, dropping the other subscripts, that SI = S2. Under these assumptions, equation (B.5) becomes simply (B.6) m = V7 n* = 0.8n, where t1 and t2 indicate the time taken to apply questionnaires I and 11. The cost equation can be written as: (B.7) t1n + t2m = 0.6n + m = 768. Solving equations (B.6) and (13.7) for Pernambuco gave the following overall sample sizes for questionnaires I and II: n* = 561, m* = 449. A general expression for the optimal size of the two samples, n* and m*, as a function of S2/51 (the ratio between the degrees of variance they are to measure) can be written as: (B3.8) n*= 768 ; m* =7768 _ 499.2 (B. 8) n* S2 M*=78-0.8&+0.6 0.8 - + 0.65 05 + 0.65 SI SI SAMPLING PROCEDURES 237 Table B-3. Change in the Optimal Size of Two Samples, as a Function of the Ratio between the Degrees of Variance They Are to Measure S2/S, n* m* 0.20 948 152 0.30 864 207 0.50 731 293 0.60 679 327 0.70 634 356 0.80 596 381 0.90 560 404 1.00 530 424 The variation of n* and m* with S2/1S is illustrated by Table B-3. Thus, the number of observations to be allocated to the two samples for Pernambuco was first determined by guessing that the ratio S2/51 was one. Consideration of the equations (B.8) and of the values in Table B-3 sug- gested that when S2/1SI is less than 0.7 changes in the ratio cause only minor variation in the sizes of the samples. Hence, in the initial phase of the survey, lacking quantitative information on the variance of the limiting factors, we used the sample allocation corresponding to 521SI = 0.7. On the basis of the early results from the first three of the nine states, S2/1S was reduced to 0.6 for the second three states, and finally to 0.56 for the last three. Let us now consider the problem of allocating a specified number of observations to the municipios and the farm size classes. The actual time needed to apply the questionnaires was thought likely to vary among the municipios and the different sizes of farms, but we were unable to predict how. Nonetheless, it was possible to calculate the differences in the variance of y, and Y2 (or a related variable, total area) by municipio and by farm size class using data from the INCRA cadastral survey. Consideration of equations (B. 3) and (B.4) then suggested the following two-stage proce- dure. First stage: Allocate n* and m* to the municipios under the assumption cib = tjb, for all i, h, according to: (.b 2Sih * (B. 9) n~,~ = _______ n__ M Li I Y. ISih Ngh i=1 h=I M L Y. 1 2Sih m* i=1 h=i Table B-4. sUDENElWorld Bank Survey: Farm Sample Size Questionnaire I Questionnaire 11 I and 11 together Total Pro- Pro- Pro- farms State grammed Realized grammed Realized grammed Realized observed Rio Grande do Norte 509 261 247 126 48 85 472 Paraiba 400 280 195 122 108 111 513 Pernambuco 583 328 283 176 98 80 584 Alagoas 399 409 137 143 82 96 648 Sergipe 442 283 154 105 94 97 485 Bahia 836 664 317 249 97 138 1,051 Piaui 783 571 297 210 149 160 941 Ceara 970 768 377 320 107 151 1,239 Maranhao 637 392 251 166 124 150 708 Maranhao (IPEI)' - 118 - 5 - 162 285 Iotal 5,559 4,074 2,258 1,622 907 1,230 6,926 -Not applicable. Note: Questionnaire 1, 5,3(04; Questionnaire 11, 2,852; total, 8,156. Overall percentage realized = 79.39 percent. a. A special survey of squatters undertaken with the Institute of Economic and Social Planning for the state of Maranhao. SAMPLING PROCEDURES 239 where kS,b, k = 1, 2, is the estimate of kS,b (the degree of variance in the group I or 2 variables in the size class and municipio in question) derived from the cadastral survey. From equations (B.9) and (B. 10) one can derive the time to be spent by the enumerators in each municipio and size class of farm: Bi E (0.6 n Ab + m,3b) (B.ll ) r=b=1 Ei where Ti is the estimate of the number of working days Ti to be spent in municipio i, size class h, by the enumerating team, and Ei is the number of interviewers (the size of the team) working in the same municipio. Second stage: Once the allocation had been made and put into effect, there were three possible outcomes: (1) Ti = Ti. In such cases the allocation was exact and no action needed to be taken. (2) Ti < Ti. In such cases the time for the application of the questionnaire in municipio i had been overestimated. The enumerator was required to expand the sample ran- domly, maintaining the same proportion between ni and mi until the predetermined time was exhausted. (3) 7; > Ti. In such cases the actual time needed was longer than had been estimated. The action was simply to accept the sample produced in the predetermined amount of time. To summarize, the original sample was modified in the field according to differences between actual and estimated enumeration time, and revisions in the estimated variance ratio S21S1 as early results were evaluated and, therefore, according to adjustments in the questionnaire I/questionnaire II sample ratio. Table B-4 compares the original allocation with the modified one that takes account of these factors. Supplementary Samples Three supplementary samples were drawn up: a sample of sharecrop- pers; a sample of farms assisted bv ANCAR, the National Agricultural Extension and Credit Agency; and a sample of squatters in a large frontier area of Maranhao. The cadastral survey on which the main sample was based was a survey of agricultural properties and does not contain information on subdivision into sharecropped plots, which is particularlv common in the interior dry lands. Both questionnaires I and II obtained an inventory of sharecroppers (if any) on each property sampled, and this provided the universe for a field-determined sample of sharecroppers, using the following rules: 240 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL -After one-third of the enumeration time for a given municipio has elapsed, consider the census of sharecroppers thus far obtained: n,- where i indicates the farm size class. -Find the largest sharecropped plot (Ari) and the smallest (AT') in each size class. -Determine the sharecropper sample size mf for each size class: M ° 29 (AG _- AT) mI Si i where Si is the standard deviation of establishment size, and mi is the regularly drawn sample size for questionnaire II. The numerator of this expression corresponds to the estimate of the standard deviation of the size of sharecropped plots, on the assumption that the latter does not vary with the size of establishments. -Select, at random, mf sharecroppers from the n,. The above procedure thus uses the information obtained from a field inventory of sharecropping units to allocate the (previouslv determined) sample for questionnaire II between sharecroppers and proprietors. The procedure is based on the assumption that the size of sharecropped plots is uniformly distributed, because this is the simplest hypothesis when there is no previous information on the distribution. The survey offers a unique opportunity to study, through cross-sectional analysis, the success of extension services provided through ANCAR. To ensure sufficient observations of establishments receiving special ANCAR assistance, a subsample was drawn in the field from ANCAR lists so that the percentage of ANCAR-assisted establishments sampled was the same as the percentage of all establishments sampled in the rnunicipio. If, for example, the regularlv drawn sample was 12 percent of the total number of establish- ments, then additional randomly drawn establishments from the ANCAR lists were sampled so that 12 percent of the ANCAR-assisted establishments were surveyed. In a large area in western Maranhao, a transitional zone bordering on the Amazon, a large number of squatters were encountered. Because few of the parcels of land they operated had titles, the INCRA cadastral survey was not a satisfactory base from which to draw a sample of them. The Institute of Economic and Social Planning for the state of Maranhao (IPEI) was particu- larly interested in studying these squatters, so an agreement was made by which a separate sample could be drawn. Based on cartography possessed by IPEI, a grid was drawn over the area, and cells were drawn at random, which were employed to locate nonpropertied farmers. These cells were then combined with the property sample for those INCRA-registered prop- SAMPLING PROCEDURES 241 erties to make a more complete sample."2 The sample size is shown in the last row of Table B-4. Notes on Sample Expansion For those microregions in which more than one municipio was sampled, the extrapolation from the sample to the microregion required weights that corresponded to the probabilities of the municipios sampled being selected. These probabilities are derived below. Probabilities of Municipio Selection As explained earlier, for each state of the Northeast the sample of municipios was drawn up in two stages, and probabilities of selection, proportional to the municipios' economic size (that is, their importance as agricultural producing units) were computed after each draw of the sample. The stage 1 sample was stratified by ecological microregion and by produc- tion pattern; municipios were selected using a set of probability weights that corresponded to the municipios' economic size. At stage 1, the prob- ability of the ith municipio being selected is given by: (B. 12) pi= v for allj, Awi where W is the economic size of municipio i, and Nj is the total number of municipios in the jth ecological/production pattern group. In the second stage, the strata were constituted simply by microregions, production patterns being ignored, and the sample was reduced initially to one municipio per microregion per state. In each of the eight states where this sample size did not exhaust the enumeration time expected to be available, a second municipio was added in the microregion that exhibited the highest variance of three measures suggestive of economic size. This procedure was repeated, and the municipio exhibiting the next highest variance was selected, until the enumeration time for the state was fully accounted for. In this iterative process, the probabilities of municipio i 12. For a more complete description of the procedure, see Institute for Economic and Social Planning (nEi), "Metodologia da amostragem por areas geograficas de unidades produtoras" and "Pesquisa por areas geograficas de unidades produtoras" (Sao Luis, Maranhao, 1974; processed). 242 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL being selected were derived as follows. The conditional probability that municipio i is selected in the first draw of stage 2, given that municipio i has already been selected in stage 1 and that the cth combination of m - 1 municipios" of the microregion in question was selected, is: (B.13) p2 = nm- Wi + Ij wjc The conditional probability that the ith municipio is selected in the first draw of stage 2 is: il c Wi (B. 14) P21 = rn I Wi + Y. Wjo c*j where C is the total number of possible combinations of municipios belong- ing to the microregion and the different production pattern groups.14 The joint probability of municipio i being selected at stage I and at the first draw of stage 2 can then be obtained as follows: (B.15) P'21 =211lC,W l !IIpC' where P-Ic is the probability of selection of the ith municipio, in combination c, at stage 1. The conditional probability for the ith unit at stage 2, draw 2, given that the kth municipio was selected at draw 1 is: P2 2 if i and k are not in the same microregion and in (B. 16) Pt = 1 -pk the same production pattern group; = otherwise. Therefore, the probability of being selected at stage 2, draw 2, is: Piz P22 1 (B. 17) p22 *=21 _p12 Expression (B. 17) is recursive and can be generalized for any draw r: 13. From stage I there will be M possible candidates for the selection of a sample of size m. 14. Note thatp, = p#pj. for municipios belonging to different production groups or different regions or both, but p = 0 if i and j are in the same production group and also in the same microregion. SAMPLING PROCEDURES 243 (B. 18) rn-I p 2rrl p k=1 -pk-~ Therefore, the probability that the ith unit will be selected in both stages is: m (B. 19) II, P2g For municipio i, the conditional probability of being selected, given that k is also in the sample, is: if i and k are not in the same microregion po. and in the same production pattern group (B.20) II/k = --I 1pk (case 1); = ° otherwise (case 2). Therefore, the joint probability of selection of the ith and kth municipios is: m P'2g m ____g k (B. 21) nik = l _ P2 g l k P 2g in case 1;' fli ={g P2gP~ gll P2g = 0 in case 2. Table B-5 lists the overall probabilities of selection for the municipios in those microregions in which more than one municipio was selected in each state. Weights for Sample Expansion For certain variables in the survey it was desirable to obtain estimates over broader areas than the clusters sampled and also to estimate population totals for such variables as employment or output. To proceed with these estimates, we needed to determine a system of weights for the municipios sampled and to estimate the population of farms by strata. For those microregions represAented by only one municipio, n, thq esti- mate for the microregion mean Y of a sampled variabley was simply Pi, the estimate of the mean for that municipio. Where more than one municipio was sampled in a microregion, we desired a system of weights wi such that n a n (B. 22) Y X ii Y~ S 1 where i is the index of the sampled municipios in that microregion. Our rationale for using a normalized system of weights is as follows. For subsampling units (farms) of unequal size and with unequal probabilities of 244 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Table B-5. Municipio Weights for Sample Expansion Probability Micro- of State region Aunicipio selection a Weight Rio Grande 3 Acu 0.1333 0.756 do Norte 3 Pedro Avelino 0.2991 0.164 3 Areia Branca 0.1945 0.080 4 Caico 0.9640 0.093 4 Florania 0.0531 0.907 6 Sao Miguel 0.4257 0.327 6 Alexandria 0.0984 0.673 Pernambuco 2 Limoeiro 0.0406 0.380 2 Sertania 0.0312 0.620 Ceara 5 Acarau 0.5815 0.156 5 Aquiraz 0.0456 0.760 5 Itapipoca 0.9347 0.084 6 ltapiuna 0.6219 0.007 6 Ipu 1.0000 0.012 6 Mucambo 1.0000 0.006 6 Poranga 0.1369 0.028 6 Tiangua 0.0164 0.947 Piaui 1 Parnaiba 1.00(0 0.171 I Buriti dos Lopes 0.4815 0.829 6 Canto do Buriti 0.5508 0.020 6 SaoJoao do Piaui 0.0154 0.980 7 Jerumenha 0.0909 0.805 7 Santa Luz 0.1638 0.195 Alagoas 2 Cha Preta 0.0820 0.311 2 Sao Jose da Laje 0.0519 0.689 Sergipe I Porto da Folha 0.2671 0.499 I Aquidaba 0.3114 0.501 3 Pedrinhas 0.0072 0.822 3 Araua 0.0723 0.178 5 Frei Paulo 0.0327 0.598 s ltabaina 0.3550 0.402 a. These probabilities would not apply if more than one municipio were selected to represent a single production pattern within the given microregion. selection (our situation), an unbiased estimate of the population microre- gion mean is:' (B.23) y= - I 15. See Cochran, Sampling Techniques, pp. 261, 305. SAMPLING PROCEDURES 245 where, in our case, M, is the number of municipios in the microregion, and Zi! = Ziln, where Zi is the absolute probability of selection derived above. Although this estimate of the microregion mean is unbiased, its variance may be very large. In terms of expression (B.23), a further related problem is that because, in general, 17= M1/nMo Zt # 1, that is, the sum of the weights of the sample municipio means in the population mean is not unity, the estimate of Y will be outside the range of the municipio mean estimates Y. Let us consider now the alternative estimator: - I n j:3 (B.24) Y= M i t M2 i=1 Zi Indicating with ti a random variable with a value of one if the ith municipio is in the sample and zero otherwise, we can write expression (B.24) as: N y. 15=i= Z! (B.25) N j The probability limit of ti will be equal to Zi = nZ!, that is, to the absolute probability of selection of municipio i. Therefore, by the Slutsky theorem, Y is a consistent estimator of Y: £ plim.1 t,iI - =I Z. n M - - (B. 26) plim Y = = E -Y, = Y. n M i= aMO plim ti i=l Estimating the Farm Population The procedure described above can be used to estimate the average population of each stratum (farm size class) and municipio. To extrapolate these results to the entire Northeast, we need an estimate of the correspond- ing stratified division of the Northeast population. The most natural auxil- iary variable for this purpose, given the type of survey conducted, is the number of farms. Once estimates have been made of the number of farms in each size class in the Northeast, it is a simple matter to obtain estimates for those size classes of any other farm level variable, by using ratio estimates as described by Cochran.'6 16. Ibid., pp. 154-88. 246 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL The definition of "farm" used for this studv-a parcel or group of close-lying parcels of land (not necessarily contiguous) under the same man- agement-differs from the definition of agricultural "properties" used by INCRA for their cadastral survev, which we used to identifv properties for our sample. INCRA had defined a propertv as an undivided parcel under the same ownership. The interviewers for our survey were instructed, upon arrival at a property, to discover the number of farms it comprised. (The propertv may not have been a farm at all, or it may have been parceled into two or more farms, or it may have been onlv part of a larger farm.) For the ith sampled municipio, an estimate of the total number of farms in the kth size class is: (B.27) Mk = bikMAk, where MFk = estimate of number of farms, municipio i, size class k Ark = number of properties in the INCRA cadastral survey, municipio i, size class k bik = ratio of actual sample size of farms to sample size of properties, municipio i, size class k. For nonsampled municipios, we did not have the complete distribution of properties, only the total numbers. We first estimated the number of properties in each size class and municipio for a given microregion: (B. 2 8) Alpk = dkMP. where ak wi aik, aik =Mf,lM; . then the estimated number of farms in size class k, municipio i, is: (B.29) = A A p where bk = w wbik. Before the formulas above were used to calculate the total number of farms per size class in the Northeast, it was necessary to test whether the number of properties was consistent with estimates that could be made of average property size. We knew the total number of properties per muni- cipio, MP, and the total agricultural area of the ith municipio, Ai. The desirable consistency conditions of the estimates APik of the number of properties and Sik of the average property size are: L (B.30) M k = Y. AL, k=l L (B. 31) Ai = Y. Yaik Sik k = I SAMPLING PROCEDURES 247 Such consistency checks may fail either because the size distribution of properties (both within and across the size classes) differs among munici- pios in the same microregion or because the average farm size of a given size class differs among municipios. Other than in the last size class (more than 500 hectares), there is little room for variation in the farm sizes within the size classes, because each is bounded from above and below in a relatively narrow range. (The consistency of average property and farm sizes within the first five strata had been verified by casual inspection.) Therefore, we permitted a variation of 10 percent in the consistency requirement (B. 31) to allow for minor differences in the sizes and sought an adjustment rule for those municipios that still did not satisfy the requirement. A practical procedure is to seek an estimated distribution closest in an appropriate sense to that estimated in expression (B.28) but meeting the requirements (B. 30) and (B. 31). Using the least squares criterion, one could construct a quadratic programming problem to find a distribution of prop- erties consistent with the a priori information, the Mj.17 Minimize (B.32) (A4p Afk) subject to (B.33) Y A4: Alp, (B.34) E M k = A, Ak Ž 0, for all k. In our case, however, this procedure would have entailed solving up to 1, 300 quadratic programming problems, a task that is simply out of the question. If, however, we ignore, for the moment, the nonnegativity constraints, we can convert (B.32)-(B.33) into a Lagrangian, (B.35) L = E (k - Mk- k- ffP) - X2 (Mkpkp - A), which has the partial derivatives: (B. 36) a (aMk = _ aAFk - XI - A2Sk = O, (B.37) aL x Mkp - MP = 0, 17. We have deleted the i subscript in what follows because any procedure must be repeated for each municipio that fails the consistency check. 248 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL (B, 3 8) dL = YMkSk - A = 0. aX2 Summing (B.28) over k and rearranging, we have (B. 39) A, -X2 Sk N where N is the number of size classes. Multiplying (B.28) by Sk, and summing over k, that is, multiplving the number of properties by the average property size, for all size classes, and rearranging, we have A - A (B.40) -2= A 2 Nk where A = AMp Skp (the estimated total agricultural area is equal to the estimated total number of properties multiplied by their average size). Thus, (13.41) At = Mk + XI + X2Sk.I In computational tests of this procedure, it was found that occasionally the Mp would be negative for either the first stratum (k = 1) or the last (k = 6). In such cases, we simply set the negative number to zero and solved the problem again for the other strata. The estimated distribution of properties for all municipios in the Northeast grouped by regions is given in Chapter 2. For nonsampled municipios, the distribution of farms was estimated from Mpk and (B. 2 7). Again, we used bik on a microregionwide basis with the weights defined above. The complete estimated distribution of farms is also shown in Chapter 2. 18. Afpis a consistent estimate of MU'because: plim XI = 0 (as -o) and plim X2 = (A - A)/ZS2 = 0, since A = plim A. Notice that if EA = A, the estimate is also unbiased. Appendix C The Fieldwork Perhaps the ideal enumerator would be a person familiar with both the agronomy and the agricultural economics of the region being surveyed, possessing sufficient statistical training to understand the causes and con- sequences of biases in response, a personality conducive to conducting a good interview with a strange farmer, and the stamina to endure the hardships of traveling in a remote region. Instead of finding or attempting to train such people, we opted for a team approach, combining people with complementary skills. Selection and Training of Enumerators Each enumeration team comprised five to eight people, depending on the size of the sample of farms and municipios, and had members selected from the four principal Brazilian participating institutions: Federal University of Pernambuco; Northeast Development Superintendency (SUDENE); the National Agriculture Extension and Credit Agency (ANCAR) for each of the nine states; and the Agricultural Planning Commission (CEPA) for each of the nine states. The Federal University of Pernambuco, which was con- tracted by the World Bank, provided six master's level economics students who served as enumeration team supervisors.' SUDENE provided several agricultural economists and agronomists from its staff, who, together with the university supervisors, provided continuity to the teams. ANCAR and CEPA provided agronomists and planning coordinators to the teams in their respective states. Thus, part of the team was constant throughout the survey as the teams moved from state to state, and part of the team, from ANCAR and CEPA, changed with each state. Although this required continual training of new enumerators, the benefits of having persons with intimate knowledge of their own areas outweighed the training costs.2 1. The university also loaned the services of Professor Yony Sampaio to the project during the sample selection and enumeration stages. 2. Not only the diversity of ecological conditions and farming practices but other differ- ences, such as units of measurement, made ihe inclusion of local specialists a necessity. 249 250 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL The training of the enumeration teams for the survey of the first state (Pernambuco) began immediately after the revisions to the questionnaire, suggested by the first field tests, had been completed and a training manual produced. This training manual contained a discussion of principles of enumeration and a description of the types of analysis to which the data would be applied. It described the logic of the questionnaire, giving a detailed description of each table and showing how the tables fitted together to permit a "flow" of information and simultaneously to allow the answers to be cross-checked. It provided a conversion table for those instances in which measurement units used locally differed from regionwide official units. The training sessions began with three days of classroom study, followed by exercises in filling out questionnaires under simulated conditions in the classroom. Subsequently, small groups of enumerators under the supervi- sion of the project leaders were sent to nearby farms to conduct interviews under real conditions. The results of these test applications were reviewed back in the classroom, and further instruction given where needed. The Paraiba and Rio Grande do Norte training sessions followed the same pattern as those in Pernambuco, except that the SUDENE and universi- ty counterparts took over the training. During these sessions, the pattern of two class days, one and a half field days, and one-half class day was established. In retrospect, this much training was not sufficient for such a complicated questionnaire; the testing and cleaning procedures later re- vealed that much more intensive training would undoubtedly have saved many months of data cleaning. It also has to be noted that the real period of training lasted longer than these four days, because the first several days of the actual fieldwork were considered a further training period, but without the interviewers' knowledge. This allowed for observation of the interview- ers' performance under realistic and exacting conditions, while there was still an opportunity to refine their techniques. Finally, it has to be stressed that the training was continued in the field not only through the daily interaction between the supervisors and the enumerators but also through constant contact between the field teams and the research group based in Recife. As will be explained below, the exten- sive traveling of the coordinators allowed them to transfer experience from one team to another team, anticipating difficulties and proposing uniform solutions to most of the problems. FILDWoRK 251 Organization and E-valuation of the Fieldwork The actual field enumeration of the questionnaires was carried out state by state in three phases: Pernambuco, Rio Grande do Norte, and Paraiba in October-November 197 3; Ceara, Piaui, and Maranhao in December 197 3- January 1974; and Alagoas, Sergipe, and Bahia in Februarv-March 1974. An average of eight municipios were sampled per state, and four enumera- tion teams worked approximately one month in each municipio. The state-by-state procedure enabled us to take full advantage of the cooperation of the state agricultural planning agencies, which provided some of the interviewers and coordinated state-specific aspects of the fieldwork. It also permitted intense supervision of the enumeration by the joint SUDENE/ ANCAR/university team. Within a given state, phasing was necessitated by the seasonality of agricultural activities. Clearly, with such an intricate organizational structure cutting across several institutions and such a large area (the state of Bahia alone is the size of France), problems could be expected. There were, of course, delays, arising from bureaucratic difficulties and in the printing of questionnaires. The fieldwork was only eleven days late in starting, however. This mini- mum delay can itself be considered a success, since the design of the project was finalized only in May, the full team formed in late Julv, the question- naire finalized in August, and the sample of farms drawn in early September. If we discount the initial frictions and the slight problems caused by changes in schedule in the first fifteen davs, the experience of the fieldwork was clearly a success. Altogether, more than fifty people were in the field from the university, ANCAR, SUDENE, and CEPA, working together as a group with no bureaucratic or institutional problems in spite of different institu- tional backgrounds, wage differentials, strains of daily interactions, and hardships encountered in the field. The field teams also developed good relations with local people such as mayors and officials of city planning agencies, who showed a strong desire to work closely with state and federal agencies in the agricultural planning of the Northeast. The supervisors were invited to use local radio and television stations to publicize the survey and participate in debates of general interest. The scheme of work adopted for the coordinators was such that each team was visited by at least one member of the central team each week. This permitted the training to continue in the field and helped to unify and accumulate the experience of interpretation, application, and analysis of the 252 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL questionnaires. These frequent visits, together with the active presence of the supervisors, also gave the enumerators the incentive to work from sunrise to sundown and at night in revision of questionnaires. In general the questionnaire proved to be both a convenient and a feasible base for the interview. The results of the pretest on the average length of the interview were, in general, confirmed during the fieldwork: Questionnaire I took between three and five hours, and questionnaire 11 did not exceed seven hours. Though the time needed for locating the farmers had often been underestimated, between 1.2 and 1.8 questionnaires were applied per interviewer per day. After three or four days of fieldwork the enumerators and the supervisors were already sufficiently familiar with the questionnaire to be able to remember most of the codes, which made the interview less formal and more flexible. The inclusion of extension agents on the teams proved particularly fortunate, not only because they could translate the questions into terms familiar to the farmers and translate back the answers, but also because they were genuinely interested in using the knowledge obtained from the interviews in their fieldwork. Almost all farmers reacted positively to being interviewed, and cases of diffidence or open suspicion were very rare. Especially in the Sertao, farmers are very articulate and enjoy the opportunity of talking about their activities. Although excessive enthusiasm on the part of the farmer may itself make for distortions in the data collected, it can be said that the most successful interviewers understood the need to let the questionnaire be the basis for a widely ranging discussion of the characteristics of the farm. Farmers in drought areas where the crop seemed to be scarce were more friendly and forthcoming than farmers in areas that had had a good year. It would have been valuable to analyze this difference quantitatively, because it could lead to distortions. Bad-year farmers could have been hoping for material help from SUDENE (whose cars were well known); or they might have been looking only for sympathy. The interviewers as well as the supervisors acquired familiarity and confidence in the questionnaire very rapidly. On certain critical topics, however, it was discovered that the interpretation of the question was a matter of opinion; no clear-cut solution was possible for this. One of the main purposes of the coordinators' traveling was to provide a unified interpretation for such points of the questionnaire. These interpretations were promptly incorporated into a second edition of the field manual developed for the enumerators. Appendix D Data Validation Procedures In addition to the obvious error of estimate, there are several potential sources of error in a sample survey of the type described here, simply because a random sample was taken instead of a complete enumeration of the population of farms.' Among these sources of error are: failure to measure some of the units in the chosen sample; errors of measurement, because farmers' answers were biased or because they could not remember or did not have certain information; errors of measurement because enumerators were biased, or were unable to translate the questionnaire into terms meaningful to farmers; errors introduced in editing, coding, and tabulating the results. The survey team was well aware of these sources of error and sought every means to control them with the time and resources available. In areas where transport infrastructure is primitive, it was sometimes impossible to include all the units in the sample as originally designed- some of the farms surveyed could be reached only by burro or canoe! The drawing of a backup sample from the identical population minimized the consequences of this problem. Table B-3 shows the degree to which the preselected sample was realized. Biases on the part of respondents are in general very difficult to measure, and the state of the art is far from precise. At least enumerators who were well versed in the agricultural practices of the area (particularly the exten- sion agents) could immediately recognize answers that were obviously incorrect. In addition, the members of each coordinated team jointly reviewed their results daily, which assisted in identifying nonsystematical- ly biased responses. The survey was not advertised to the farmers as having any specific objective (such as property valuation for tax purposes, or particular de- velopment projects), and they generally had no obvious reason to bias their 1. Vinh Le-Si assisted in drafting this section. 253 254 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL answers. An exception, though, arose on the subject of sharecroppers: Recently enacted labor and land tenure laws had placed restrictions on farms with sharecropper employment. Several obvious attempts to "hide" sharecroppers-for example, by defining them as renters-or to give in- accurate information were discovered. Answers that required a subjective evaluation by the farmers, however, were a source of problems. In particular, the questions on value of land and structures proved very difficult, partly because the market for land is very weak in many areas. The inconsistency in these subjective answers shows up in the tabulated results. For one item, farm size, an alternative check was possible. In conjunction with an overlapping partial soil survey, precise measurements of the farms were taken and showed that virtually all farmers had accurate information on the size of their farms. The use of complementary enumeration teams, composed of individuals drawn from several institutions and different backgrounds, helped to con- trol enumerator bias, at least in the aggregate. Again, the dailv reviews in the field, in which the entire team was usually involved, made it possible to spot bias on the part of individual enumerators. Because the enumerator is identified on each questionnaire, it would be possible to check for biases by, for example, cross-sectional analysis of the questionnaires enumerated bv extension agents and SUDENE analysts. Time constraints have not permitted such checks as of the time of writing. As mentioned above, the cleaning process began in the field with daily review of the incoming questionnaires by the enumeration team supervisor and members of the Recife-based SUDENE/World Bank coordinating team. As quickly as possible, the questionnaires were returned to the Recife office where they were reviewed by another SUDENE team before being key- punched. This last review was primarily concerned with verifying the codes for crops, equipment, type of farm management, and so forth, and checking calculations made in the field. The questionnaires were then keypunched and loaded onto magnetic tape. At this point it was possible to use the computer to help detect certain types of remaining errors. Two programs were developed, Critica and Consistencia, and each questionnaire was processed with both programs twice. Critica checked each card individually, searching for detectable keypunch error, incorrect or missing codes, missing information, and within-card consistency. Some examples of the latter checks are: that the areas shown under sources and under uses of farmland (from Table 2A of questionnaire 1) are equal; that the volume of output distributed equals the DATA VALIDATION PROCEDURES 255 volume produced (from questionnaire Table 6); that the total annual labor remuneration equals the number of days worked times the daily wage (or number of weeks times the weekly wage), plus remuneration in kind (from questionnaire Table 9). The first processing with Critica revealed errors, most of them in keypunching, in about 30 percent of the cards. After correcting them, a second pass was made to verify the corrections before proceeding to Consistencia. The objective of Consistencia was to reveal errors in consistency among cards and tables. TIhe questionnaire was designed to permit several ways of asking certain questions and hence to permit cross-checks among answers. Some examples of Consistencia checks are: that the area of crops and pasture enumerated in questionnaire Table 5 matches the area given in Table 2A; that the consumption, sales, and other types of distribution of all outputs match the production and purchases; that a new crop (less than one year of age) identified in Table 5 should have associated labor inputs for soil preparation and planting shown in Table 11 of the questionnaire; that certain inputs shown in questionnaire Table 12 (for example, gasoline and oil) should match certain types of eqluipment (for example, mechanized equipment); that the labor drawn from various sources, shown in question- naire Tables 10, 11, and 15, should be consistent with the census of labor in Table 9; and that the sharecropping information (land in Table 2A, produc- tion in Table 5, and the sharecropper census in Table 14) be consistent across tables. Consistencia revealed actual or possible errors in another 25 percent of the questionnaires, each of which was resolved by the Recife reviewing team, and, again, a second pass was made to check corrections. The large number of errors encountered bv Critica and Consistencia perhaps reflected difficulties on the part of the enumerators in fully under- standing the spirit of the consistency requirements of the questionnaire and also, perhaps, their difficulties with the precoding svstem. Undoubtedly, the training period was insufficient, and some conceptual difficulties re- mained with the enumerators in the field. Many of the errors in consistency could have been avoided if the training had been more comprehensive and if the enumerators had been informed of the actual checks Critica and Consis- tencia would later make. Such prior warning, however, would have de- stroyed our ability to make independent tests of the data and would probably have yielded poorer data quality (if the enumerator were more concerned with the various consistency requirements, his effort to obtain accuracy on specific answers might suffer). The data cleaning process took about nine months (the last data were collected in April, and Consistencia corrections were completed in Novem- 256 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL ber). The reviewing team comprised twenty people on the average (many of them returning field supervisors), and about fifty-five hours of SUDENE'S IBM 360/30 computer time was required. During the final stages of testing with Critica and Consistencia, the researchers began examining statistics on items such as the distribution of prices and yields, and some technical coefficients of production for use in preliminary linear programming work. We found that certain types of errors still remained, particularly those concerned with measurement units. The unit of production, for example, for most crops was kilograms but was metric tons for sugarcane. Often the enumerator mistakenly used kilograms for sugar, and if he did this consistently throughout the question- naire the error remained undetected. Similar errors often encountered concerned eggs, which should have been recorded in units of 1,000 eggs but were often given in units of eggs; forage inputs (kilograms instead of tons); and electricity (only the annual cost was asked, but watts were sometimes given as well, resulting in enormous numbers when watts and annual cost were multiplied). Therefore, still another round of testing was needed. We called this Supertest, because it involved interfarm comparisons, whereas the pre- vious tests were concerned with the questionnaires on the basis of indi- vidual farms. The strategy followed in Supertest was to calculate the distributions of various key variables and search for outliers from these distributions.2 These key variables were prices, yields, gross output, gross labor input, purchased inputs for livestock, and other purchased inputs. For those variables that are not likely to be highly correlated with farm size (prices, yields, ratio of livestock inputs to herd value, and so on) we simply calculated the means and standard deviations on a statewide basis and flagged those farms for which the variable was not within one standard deviation of the mean. 'rhe output from the Supertest computer program displayed these flagged farms with the corresponding values of the variables considered, enabling an economist to examine the results at a glance and decide whether further examination of the questionnaire itself was warranted. A low level of output per hectare, for example, was easily explicable if labor per hectare was also low. If there was no such obvious reason, the questionnaire was 2. For a technical discussion of the strategy of detecting errors, see Philip Musgrove, "Detecting Errors in Economic Survey Data: Multivariate vs. Univariate Procedures," Annals of Economic and Social Measurement, vol. 3, no. 2 (1974), pp. 333-45; and Frank E. Grubbs, "Procedures for Detecting Outlying Observations in Samples," Tecbnometrics, vol. Ii, no. I (February 1969), pp. 1-21. DATA VALIDATION PROCEDURES 257 again reviewed. In some cases, no error was detected, as the values of other variables provided an explanation. Low values for both output per hectare and output per man, for example, accompanied by a high input of labor per hectare may be perfectly plausible if the agricultural year surveyed on a particular farm was devoted largely to the establishment of new long-cycle crops; high output ratios accompanied by low labor inputs could be the result of drawing down livestock inventories during the year. Supertest revealed that, once these apparent anomalies had been accounted for, about 7 percent of the questionnaires contained errors. Virtually all these errors were due either to the use of incorrect units for recording physical products, prices, or labor use or to a misplacing of the coded columns of the questionnaire.' 3. That is, if the number ]Ois to be placed in the IBM card field comprising columns 2-32, a I should be placed in column 31. If it is misplaced or mispunched in column 30, the computer will read 100. Bibliography The word "processed" describes works that are reproduced from typescript by mimeograph, xerography, or similar means; such works may not be cataloged or commonly available through libraries, or may be subject to restricted circulation. "Agrarian Reform: A Time-worn Subject Returns to the Headlines." Veja, Decem- ber 7, 1977. Ahluwalia, Montek S., John H. Dulov, Graham Pyatt, and T. N. Srinivasan. "Who Benefits from Economic Growth? An Examination of Fields' Reexamina- tion for Brazil." American Economic Review., vol. 70, no. I (March 1980), pp. 242-45. Andrade, Manuel Correia de. A terra e o homen no NVordeste. Sao Paulo: Editora Brasiliense, 1964. Baer, Werner. 'The Brazilian Growth and Development Experience: 1964-1975." In Brazil in the Seventies. Edited bv Riordan Roett. Washington, D.C.: American Enterprise Institute for Public Policy Research, 1976. Baer, Werner, and Isaac Kerstenetzky. "The Brazilian Economy." In Brazil in the Sixties. Edited by Riordan Roett. Nashville, Tenn.: Vanderbilt Universitv Press, 1972. Berry, R. Albert, and William R. Cline. Agrarian Structure and Productivity in Developing Countries. Baltimore, Md: Johns Hopkins University Press, 1979. Bolch, B. W., and C. J. Huang. Multivariate Statistical Methods for Business and Economics. Englewood Cliffs, N.J.: Prentice-Hall International, 1974. "Brazil's Arid Northeast Skeptical as Help Is Promised Once Again." New York Times, September 21, 1979, p. A13. "Brazil's Outcast Shanty Dwellers Offered a Helping Hand at Last." New York Times, October 22, 1972, p. A4. Britton, D. K., and Berkeley Hill. Size and Efficiency in Farming. Lexington, Mass.: Saxon House Studies/Lexington Books, 1976. Buckwell, A. E., and P. B. R. Hazell. "Implications for Aggregation Bias for the Construction of Static and Dynamic Linear Programming Supply Models." Journal of Agricultural Economics, vol. 23, no. 2 (1972), pp. 119-34. Candler, Wilfred V. "A Demonstration Model for the Dairy Industry of La Laguna." In The Book of CHAC: Programming Studies for Mexican Agriculture. Edited by Roger D. Norton and Leopoldo Solis MI. Baltimore, Md.: Johns Hopkins University Press, forthcoming. 259 260 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Castro de Rezende, Gervasio. "Plantation Systems, Land Tenure, and Labor Supply: An Historical Analysis of the Brazilian Case with a Contemporary Study of the Cacao Regions of Bahia, Brazil." Ph.D. dissertation. Madison: University of Wisconsin, 1976. C(astro, Josue de. Death in the Nortbeast. New York: Random House, 1969. Cavalcanti de Albuquerque, Roberto, and Clovis de Vasconcelos Cavalcanti. "Re- gional Development Policies in Brazil: Precedents and Prospects." Seminar on Regional Development/Brazil, European Economic Community. Brasilia, Octo- ber 17, 1972. Cehelsky, Marta. Land Reform in Brazil: The Management of Social Change. Boulder, Colo.: Westview Press, 1979. Chenery, Hlollis B. "Development Policies for Southern Italy." Quarterlyjournal of Economics, vol. 76 (November 1962), pp. 515-47. Chow, G. S. "Tests of Equality between Sets of Coefficients in Two Linear Regressions." Econometrica, vol. 28, no. 3 (uly 1960), pp. 591-605. Cirne Lima, Luis Fernando. Jornal do Brasil, November 20, 1977. Cline, William R. Economic Consequences of a Land Reform in Brazil. Amsterdam: North-Holland Publishing Company, 1970. Cochran, William G. Sampling Techniques. New York: John Wiley and Sons, 1963. Comite Interamericano de Desarrollo (CIDA). Land Tenure Conditions and Socioeconomic Development of the Agricultural Sector, Brazil. Washington, D.C.: Pan American Union, 1964. Conjuntura economica (Rio de Janeiro), vol. 31, no. 7 Uuly 1977). D)ay, Richard H. "On Aggregating Linear Programming Models of Production." Journal of Farm Economics, vol. 45 (November 1963), pp. 797-813. D)e Azevedo, Fernando. Brazilian Culture: An Introduction to the Study of Culture in Brazil. Translated by William Rex Crawford. New York: Macmillan Company, 1950. [)illon, John L., and Pasquale L. Scandizzo. "Risk Attitudes of Subsistence Farm- ers in Northeast Brazil: A Sampling Approach." American Journal of Agricultural Economics, vol. 6, no. 3 (August 1978), pp. 425-35. [uloy, John H., and Roger D. Norton. "CHAC: A Programming Model of Mexican Agriculture." In Multi-level Planning: Case Studies in Mexico. Edited by Louis M . Goreux and Alan S. Manne. Amsterdam: North-Holland Publishing Company, 1973. . "Prices and Incomes in Linear Programming Models." AmericanJournal of Agricultural Economics, vol. 57 (November 1975), pp. 591-600. Eckstein, Shlomo, Gordon Donald, Douglas Horton, and Thomas Carroll. Land Reform in Latin America: Bolivia, Chile, Mexico, Peru, and Venezuela. World Bank Staff Working Paper no. 275. Washington, D.C.: World Bank, April 1978. Egbert, Alvin C., and Hyung M. Kim. "Analysis of Aggregation Errors in Linear Programming Planning Models." American Journal of Agricultural Economics, vol. 57, no. 2 (May 1975), pp. 292-301. BIBLIOGRAPHY 261 A Development Modelfor the Agricultural Sector of Portugal. World Bank Staff Occasional Paper no. 20. Baltimore, Md.: Johns Hopkins University Press, 197 5. El-Tobgy, H. A. Contemporary Egyptian Agriculture. Beirut: Ford Foundation, 1974. Estudo Nacional da Despesa Familiar. Consumo alimentarantropometria, dadosprelimi- nares. Regiao V. Rio de Janeiro: FIBGE, 1977. - Despesas das familias, dados preliminares. Regiao V. Rio de Janeiro: FIBGE, 1978. Feder, Gershon, and P. L. Scandizzo. "A Two-Region Multisectoral Model for Brazil." Washington, D.C.: World Bank, Development Research Center, 1977. Processed. Fields, Gary. "Who Benefits from Economic Development? A Re-examination of Brazilian Growth in the 1960s." American Economic Review, vol. 67 (September 1977), pp. 570-82. Fishlow, Albert. "Brazilian Size Distribution of Income." American Economic Re- view, vol. 62, no. 2 (May 1972), pp. 391-402. Forman, Shepard. The Brazilian Peasantry. New York: Columbia Universitv Press, 1975. Freyre, Gilberto. The Masters and the Slaves: A Study in the Development of Brazilian Civilization. New York: Alfred Knopf, 1946; abridged ed., 1964. Fundacao Getulio Vargas. Precos pagos pelos agricultores. Rio de Janeiro, 1972. Furtado, Celso. The Economic Growth of Brazil: A Surveyfrom Colonial to Modern Times. Translated by Ricardo W. de Aguiar and Eric Charles Drysdale. Berkeley: University of California Press, 1963. Goodman, David. "Industrial Development in the Brazilian Northeast: An Interim Assessment of the Tax Credit Scheme of Article 34/18." In Brazil in the Sixties. Edited by Riordan Roett. Nashville, Tenn.: Vanderbilt University Press, 1972. Goodman, David, and Roberto Cavalcanti. Incentivos a industrializacao e desenvol- vimento do Nordeste. Rio de Janeiro: IPEA/INPES, 1974. Goreux, Louis M., and Alan S. Manne, eds. Multi-level Planning: Case Studies in Mexico. Amsterdam: North-Holland Publishing Company, 1973. Grubbs, Frank E. "Procedures for Detecting Outlying Observations in Samples." Technometrics, vol. 11, no. I (February 1969), pp. 1-21. Grupo de Trabalho para o Desenvolvimento do Nordeste (GTDN). Uma politica de desenvolvimento para o Nordeste. Rio de Janeiro: Conselho de Desenvolvimento, 1959. - Uma politica de desenvolvimento economico para o Nordeste, 2d ed. Recife: SUDENE, 1967. Hadley, George. Non-Linear and Dynamic Programming. Reading, Mass.: Addison- Wesley Publishing Co., 1964. Hall, Anthony L. Drought and Irrigation in North-East Brazil. Cambridge, England: Cambridge University Press, 1978. Hansen, Morris, William Hurwitz, and William G. Meadow. Sample Survey Methods and Theory. Vol. 1. New York: John Wiley and Sons, 1953. 262 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Harris, John R., and Michael P. Todaro. "Migration, Unemployment, and De- velopment: A Two-Sector Analysis." American Economic Review, vol. 60, no. I (March 1970), pp. 126-42. Havighurst, Robert J., and J. Roberto Moreira. Society and Fducation in Brazil. Pittsburgh, Pa.: University of Pittsburgh Press, 1965. Hazell, Peter B. R. "Farm Planning under Uncertainty." A merican Journal of Agri- cultural Economics, vol. 53 (February 1971), pp. 53-62. Hazell, Peter B. R., and P. L. Scandizzo. "Competitive Demand Structures under Risk in Linear Programming Models." American Journal of Agricultural Economics, vol. 56 (May 1974), pp. 235-44. Heady, Earl O., and John L. Dillon. Agricultural Production Functions. Ames: Iowa State University Press, 1960. Hirschman, Alfred 0. Journeys toward Progress: Studies of Economic Policy-Afaking in Latin America. New York: TIwentieth Century Fund, 1963. Houthakker, Hendrik S., and Lester D. Taylor. Consumer Demand in the United States: Analyses and Projections. Cambridge, Mass.: Harvard University Press, 1970. Instituto Brasileiro de Geografia e Estatistica (IBGE). Anuario estatistico do Brasil. Rio deJaneiro, 1975, 1976. . Censo demografico VIII: Recenseamentogeral, 1970. Vol. 1, nos. 5-10. Rio de Janeiro, 1972. . Dados preliminaresgerais do censo agropecuario 1V1II: Recenseamento geral, 1970. Rio de Janeiro, 1972. Institute of Economic and Social Planning (IPEI). "Metodologia da amostragem por areas geograficas de unidades produtoras." Sao Luis, Maranhao, 1974. Pro- cessed. "Pesquisa por areas geograficas de unidades produtoras." Sao Luis, Maranhao, 1974. Processed. International Trade Centre. Cassava: Fxport Potential and Mlarket Requirement. Ge- neva: UNCTAD/GArr, 1977. Johnson, Allen W. Sharecroppers of the Sertao. Stanford, Calif.: Stanford University Press, 1971. Katzman, Martin T. Cities and Frontiers in Brazil: Regional Dimensions of Economic Development. Cambridge, Mass.: Harvard University Press, 1977. Kutcher, Gary P. "Agricultural Planning at the Regional Level: A Programming Study of Mexico's Pacific Northwest." Ph.D. dissertation. College Park: Uni- versity of Maryland, 1972. Kutcher, Gary P., and Pasquale L. Scandizzo. "Northeast Brazil Regional Studies Project: Sampling Procedures." Washington, D.C.: World Bank, Development Research Center, 1974. Processed. "A Partial Analysis of Sharetenancy Relationships in Northeast Brazil." Journal of Development Economics, vol. 3 (December 1976), pp. 343-54. BIBLIOGRAPHY 263 Lacerda de Melo, M. "Espacos geograficos e politica espacial: 0 caso do Nordeste." Boletim economico, vol. 5, no. 2 (July 1969/December 1971), pp. 7-139. Langoni, Carlos Geraldo. Distribuicao da renda e desenvolvimento economico do Brasil. Rio de Janeiro: Fditora Expressao e Cultura, 1973. Leff, Nathaniel H. "Economic Development and Regional Inequality: Origins of the Brazilian Case." Quarterly Journal of Fconomics, vol. 86, no. 2 (May 1972), pp. 243-62. Leuthold, Raymond M. "On the Use of Theil's Inequality Coefficients." American Journal of Agricultural Economics, vol. 57, no. 2 (May 1975), pp. 344-46. Ludwig, Armin K., and flarry W. Taylor. Brazil'sNewAgrarian Reform. New York: Praeger, 1969. Mellor, John W. The Economics of Agricultural Development. Ithaca, N.Y.: Cornell University Press, 1966. Miller Paiva, Ruy. "Modernizacao e dualismo tecnologico na agricultura. " Pesquisa e planejamento economico, vol. 1, no. 2 (December 1971), pp. 171-234. Miller Paiva, Ruy, and George F. Patrick. Acesso a terra no Nordeste. Colecao Relatorios de Pesquisa. Rio de Janeiro: IPEA/INPES. Processed. Miller, T. A. "Sufficient Conditions for Exact Aggregation in Linear Programming Models." Agricultural Economics Research, vol. 18 (1966), pp. 52-57. Musgrove, Philip. "Detecting Errors in Economics Survey Data: Multivariate vs. Univariate Procedures." Annals of Economic and Social Measurement, vol. 3, no. 2 (1974), pp. 333-45. Nakajima, Chihiro. "Subsistence and Commercial Farms: Some Theoretical Mod- els of Subjective Equilibrium." In Subsistence Agriculture and Economic Develop- ment. Edited by C. R. Wharton, Jr. Chicago: Aldine, 1970. Nicholls, William H. "Paiva e o dualismo tecnologico na agricultura: um comen- tario." Pesquisa eplanejamento economico, vol. 3, no. I (March 1973), pp. 15-50. - "The Brazilian Agricultural Economy: Recent Performance and Policy. " In Brazil in the Sixties. Edited by Riordan Roett. Nashville, Tenn.: Vanderbilt University Press, 1972. Nugent, Jeffrey B. "Linear Programming Models for National Planning: Demon- stration of a Testing Procedure." Econometrica, vol. 38, no. 6 (November 1970), pp. 831-55. "O feijao e o sonho," Veja, March 2, 1977. Page, Joseph A. The Revolution That Never Was: Northeast Brazil, 1955-64. New York: Grossman Publishers, 1972. Paris, Quirino, and G. C. Rausser. "Sufficient Conditions for Aggregation of Linear Programming Models." American Journal of Agricultural Economics, November 1973, pp. 659-66. Patrick, George F. "Efeitos de programas alternativos do governo sobre a agricul- tura do Nordeste." Pesquisa eplanejamento economico, no. I (February 1974), pp. 43-82. 264 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Pfeffermann, Guy P., and Richard C. Webb. The Distribution of Income in Brazil. World Bank Staff Working Paper no. 356. Washington, D.C., September 1979. Reboucas, D. E. "Interregional Effects of Economic Policies: Multisectoral General Equilibrium Estimates for Brasil." Ph.D. dissertation. Cambridge, Mass.: Har- vard University, 1974. Robock, Stefan H. Brazil's Developing Northeast: A Study of Regional Planning and Foreign Aid. Washington, D.C.: Brookings Institution, 1963. Roett, Riordan, ed. Brazil in the Sixties. Nashville, Tenn.: Vanderbilt University Press, 1972. Scandizzo, P. L., and T. Barbosa. "Substituicao e produtividade de fatores na agricultura Nordestina." Pesquisa eplanejamento economico, vol. 7, no. 2 (August 1977), pp. 367-404. Schuh, G. Edward. The Agricultural Development of Brazil. New York: Praeger, 1970. . "Modernizacao e dualismo tecnologico na agricultura: Alguns comen- tarios." Pesquisa e planejamento economico, vol. 3, no. I (March 1973), pp. 17-94. Selowsky, Marcelo. Who Benefits from Government Expenditure? A Case Study of Colombia. New York: Oxford University Press, 1979. Stigler, George J. The Theory of Price. Rev. ed. New York: Macmillan Company, 1952. SUDENE. "The Brazilian Northeast and Its Institutions for Economic Develop- ment." Recife, 1961. - Coeficientes tecnicos do Nordeste. Recife, 1976. A economia agricola do Nordeste: Diagnosticoparcial eperspectivas. Recife, 1976. UJnited Nations, Food and Agriculture Organization. .4gricultural Commodity Projec- tions, 1970-1980. Vol. 2. Rome, 1971. . Coffee in Latin America. Rome, 1960. Webb, Kempton E. The Changing Face of ANortheast Brazil. New York: Columbia University Press, 1974. World Bank. Land Reform. Sector Policy Paper. Washington, D.C., 1974. "Rural Development Issues and Options in Northeast Brazil." Report no. 665a-BR. Washington, D.C.,June 23, 1975. A restricted-circulation document. Rural Development. Sector Policy Paper. Washington, D.C., February 1975. WorldDevelopment Report, 1979. New York: Oxford University Press, 1979. Index Agrarian reform, development and absence Capital: agricultural, 53-57; in Northeast, of, 191-94. See also Land reform 23-24; constraints at farm level, 166; farm Agreste, 153; agricultural development and, models and, 107, 144; farm size and. 84-87 7;characteristicsof, 35;familylaborin, 91; Cattle: beef production and, 60, 168; breed- farm size in, 49, 53; input costs in, 65; as ing of, 7-8; in farm models, 111-12; as physiographic zone, 22 income source, 54; inputs and, 65. See also Agricultural Planning Commissions (CEPA), Livestock 249 Cavalcanti, Clovis de Vasconcelos, 15, 17 Agricultural production: capital and, 55; ex- Cavalcanti de Albuquerque, Roberto, 15, 17 change rates and, 87-88; farm models and, Chenery, Hollis B., 63 n.10 1()5, 145; farm size and, 97-100: input Climate, 21; risk and, 110-11, 178, 179; set- costs and, 63-67; model-simulated levels tlement of Northeast and, 9. See also of, 152-55, 158; model overestimation of, Drought 162, 163; model pattern of, 103; in North- Cline, William R., 197, 205, 206, 208 east, 24-25. See also Productivity Cacao, 59, 62, 87, 92, 104, 216, 217; farm Agricultural sector: characteristics of, 89-94; incomes and, 69. See also Crops early history of, 5-10; economic projection Cochran, William G., 245 for, 170-73; family expenditures and, 74- Coffee, 6, 87 77; as focus of sUDENFJWorld Bank survey, Colonization: of Brazil, 5-7; land reform law 4; global expenditures and, 79-81; mone- and, 196; of western and northern Brazil, tary expenditures and, 77-79; output and, 14-15 60-63; production costs and, 63-67. See Commercial farms. 95; farm models and, also Farmers; Farm size; Model of North- 122; market dependency and, 89-94. See eastern agricultural sector also Family farms Article 34/18. 16 Consistencia (data evaluation program), 254- 56 Bahia. 14; zone delineation and, 31, 34 Consumers. 173, 176, 177, 187; behavioral Bank of the Northeast (BNB), 16; creation of, changes and, 170; demand intervention 13 and, 189-90, 191; land reform and. 203; Beans, 58, 117, 118, 178, 217, 218; demand wage subsidies and, 182 intervention and, 189; farm family con- Corn, 58, 117, 118, 178, 217, 218; farm fam- sumption and, 109. See also Crops ily consumption and, 109. See also Crops Brazil: early agricultural development in, 5- Costs, 163, 169; agricultural inputs and, 63- 10; growth of, 3-4. Se also Model of North- 67; DRc coefficients and, 167-68; farm eastern agricultural sector; Zones (North- models and, 107; land reform, 206-08; of eastern) production, 25 Brazilian Agricultural Research Company Cotton, 13, 58, 59, 62, 88, 178, 193, 195, (EMBRAPA), 19, 176 216, 217, 218; farm models and, 108, 117- Brazilian Institute for Agrarian Reform 18; model-simulated production of, 153, (IBRA), 15 175; in the Sertao, 8, 113. See also Crops Brazilian Land Statute of 1964, 29, 196-97, Credit, 213; constraints at farm level and, 210 166; farm models and, 107 Buckwell, A. E., 123, 125 Critica (data evaluation program), 254-56 265 266 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Crops: agricultural production analysis and, 163-64; model performance and, 156-59; 24; exports and, 62, 117-18, 167, 168, 173, model simulation and, 29; nonfarm, 41; 188, 202; farm models and, 105, 108, 117- policies and, 179-83; sharecropper, 23, 18, 145, 147-48; long-cycle, 59, 64, 108, 172, 174-75, 177; subsidization of, 28; 153, 165-66, 171, 172, 173; medium- technical progressand, 177. SeealsoUnder- cycle, 108; model-simulated production employment; Unemployment of, 153-54, 155; output from, 60-63, 173, Fnumerators, training of, 249-52 179, 191, 202; projections for, 175-77; risk Equipment: farm type and, 95; land reform and, 177-79; short-cycle, 58, 118. See also and, 206, 207; maintenance of, 64-65; Food crops; Tree crops; Yields; names of value of, 54-55 specific crops Estates, 88; development of, 6, 7, 8; develop- Currency: devaluation of, 214; overvaluation ment strategy and, 193-94; land reform of, 168 and, 28-29, 195, 198-99, 207, 211; model and, 28, 104, 127, 135; size of, 27; unused Day, Richard II., 121, 122, 123 land on, 16-17; wage subsidies and, 182 Debt, foreign, 213 Estatuto da terra, Law no. 4504 (land reform, Demand, 170, 208; constraints at sectoral 1964), 29, 196-97, 210 level and, 167, 168; economic projection Exchange rate, 14; exports and, 87-88 and, 171; for labor, 116; population Expenditures: agricultural input, 25, 65; growth and, 172-73; price elasticities of, clothing, 80; farm models and, 107408; 120, 160-62; programming model and, food, 80; global, 79-8 1; monetary, 77-79; 118-20, 141, 149; promotion of, 183-91 national study of family, 73-81 Development, 101; in Agreste, 35; Furtado Exports, 24, 213; constraints at sectoral level report and, 13-19; strategies for, 191-94. and, 167, 168; crop, 62, 117-18, 167, 168, See also Policy 173, 188, 202; exchange rates and, 87-88; Development Program for Integrated Areas Furtado report and, 13 in the Northeast (POLONORDESTE), 18, 19 Extension services, 207, 208 Diet (rural), 70-71. See also Nutrition Domestic resource cost (DRC) coefficients, Family farms, 202; demand promotion and, 167-68 191; efficiency of labor use on, 100; farm Drought, 17, 31; government responses to, type and, 95, 97; income projection and, 10- 13; regular occurrences of, 9; relief 177, 193; labor on, 28, 77, 105; market work for, 19; riskand, 110-1 1, 178, 179; in dependency and, 89-94; model typology semiarid Sertao, 33. See also Climate and, 104, 127; size of, 27. See also Farm Duloy, John 11., 118, 149 families Dutch occupation of Brazil, 5 Farmers, 103, 187;demandinterventionand, 189, 191; income of, 67-68, 71, 173-75; Ecology, 21, 87 projected income growth of, 173-75, 182; Education, 213-14 tenant, 7; work days of, 41. See also Family Efficiency: demand and, 184; in factor alloca- farms; Farm families; Sharecroppers tion, 100; farm size and, 92, 97-100; tech- Farm families: expenditures of, 74-81; farm nical, 96, 97 level constraints and, 164, 165, 167; farm Egbert, Alvin C., 151 models and, 105, 123; food consumption Emigration, 14, 15, 175 and, 109-10, 117; income and, 173-74, Employment, 103, 172, 191; access to land 182;laboruseand, 100, 156, 158, 164;1and and, 23; base case solution and, 173, 174- reform and, 29, 199, 201-02; projected rise 75; development and, 192-94; farm type in labor of, 177; reservation wage and, and, 93; Furtado report and, 15, 16; land 159-60; seasonality and, 93; workers in, reform and, 195, 199-200, 203-205, 206, 39. See also Family farms 208, 210, 211; model overestimation of, Farm-level programming, 223 INDEX 267 Farm models: aggregation of, 120-37; char- Ilazell, P. B. R., 110, 118, 123, 125, 147 acteristics of, 109-16; constraints on, 164- 1 louseholds: access to land of, 69-70; average 67; core matrix of, 105-09; land reform farm, 39; expenditures of, 74-77; global and, 197-202; linkages among, 116-18; expenditures and, 79-81; monetary ex- sample selection of, 229-30, 233-39 penditures and, 77-79; rural, 41 Farm types: aggregation of, 120-37; farm size Hlouthakker, Hendrik S., 170 and, 26-27; market dependency of family and nonfamily, 89-94; model and, 27-28, 29; model typology and, 104, 122-35;mul- illiteracy, 206, 213; rates of, 4 tiple classification of, 94-97; zones and, Imports of food, 13-14 103 Import substitution, 88, 213 Farm size: analvsis of, 23; expenditures and, Income, 77, 103, 189, 191; from agriculture, 65; farm models and, 121, 122-23; farm 25-26, 63, 67-73; base case solution and, performance and, 26-27; farm population 173-75; from cattle, 54; development and, and, 245; income and, 71-73; labor use 192-93; elasticities and demand con- and, 82-89,218; land analysis and, 41-47, straints and, 188; expenditures and, 65; 53; land reform and, 195, 197-98; market factors constraining, 102; landowner and dependency and, 89-94; model typology sharecropper, 115; landownership and, 41; and, 104, 127-35; multiple classification land reform and, 195, 201-02, 205, 208, of, 94-97, 125-27; policy and, 2 14-15; re- 209-10, 211; in Northeast, 3-4, 13, 212; turns to scale and, 97-100; resource use projection for family farm, 177, rural, 61, and, 125; survey sample and, 247, 254 70-71p wage smibsidies and, 182 Favelas (urban squatter settlements), 216, Industrial development, 215; Furtado report and, 14, 15-16 Fazendas, 153, 193, 206; farm model and, Inflation, 14; land values and, 88 115, 122, 127, 165; farm size and. 27; Infrastructure: farm models and, 118; invest- model farm typology and, 104; wage sub- ment in, 214; SUDENE and socioeconomic, sidies and, 182 18 Federal University of Pernambuco, 223, 249 Inputs, 153; agricultural, 25; farm size and Fieldwork (survey), 249-52 use of, 97, 100; livestock and, 112; produc- Food crops: economic projections and, 170; tion costs and, 63-67 farm models and, 117-18; increasing, 217; Institute of Economic and Social Planning output of, 173. See also Crops (IPEI/Maranhao), 240 Food consumption, 173, 202, 217; of farm Instituto Brasileiro Geografia e Estatistica families, 109-10, 117 (IBGE), 25; definition of zones and, 30, 31; Fruits, 59, 118 sampling procedures and, 229 Furtado, Celso, 8; report of, 13-19 Interest rates, 214 Government intervention: colonization Interviews (survey), 223, 224, 236. See also efforts and, 14-15; droughts and, 10-12; Questionnaire (survey) industrial inicentives and, 15-16; land re- Investment, 13, 55; agricultural, 214; costs indutrll ncentvesand 15-6; andre- of, 64; tree crops and, 108; land reform form and, 16-19; Northeast regional de- and, 206 velopment and, 13-14 aId See Institute of Economic and Social Grazing land, 179, 216, 217; absorption of Planning Econom slash-and-burn farms and, 49; livestock Planning (ipEi,Maranhao) and, 112-13. See also Land Irrigation, 11-12, 171, 210; farm type and, Growth rate, 3, 212-13; model and, 102 95; in Middle-North, 33 lladley, George, 145 n.24 Kim, Hyung M., 151 flall, Anthony L., 11 Kutcher, Gary P., 151 268 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Labor force: agricultural households and, 77; Livestock: beef production and, 59-60; con- definition of agricultural, 36-38; develop- straints at farm level and, 166; in farm ment and, 193; farm-level constraints models, 111-13, 177, 144; inputs and, 25, (model) and, 164; farm models and, 105, 65, 112; land reform and, 202, 203, 207; 107, 116-17, 122, 143; farm size and, 26, model performance and, 154-55; output 82-89, 100, 218; market dependency and, and, 60, 63, 179, 217; projections for, 176, 89-94; in Northeast, 22-23; plantation, 6; 177; risk and, 178, 179; role of Northeast policies and, 170, 215; sUDENE/World and, 216, 217; value of, 54. See also Cattle Bank survey and, 38-41 Locational differences in land, 50, 53, 87 Labor markets, linkages among farm models Ludwig, Armin K., 196 and, 116-17 ILabor utilization, model performance and, 156-59 Management: decentralization of, 97; land Land: access to, 23, 69-70; area of farm, 23; reform and, 206 projected availability of, 175; constraints Manioc, 58, 118, 217, 218; demand interven- at farm level (model) and, 164-65, 167; tion and, 189; farm familv consumption distributionof, 41-47,53;estateuseof, 28; and, 109. See also Crops farm models and, 107, 142; farm size and Maranhao, 14, 15, 77, 239, 240; zone delinea- use of, 82-89, 100; output and, 61; own- tion and, 31 ership percentages and, 22; quality of, 50- Market dependency, 89-94 53; rental, 195; tvpes in model, 105; under- Migration: development policy and, 215-16, use of, 179; use ofless productive, 171. See 217, 218; earlv inland, 7, 8; slash-and-burn also Grazing land agriculture and, 38 Landowners, 158; definition of "farm" and, Mining, 7-8, 9 246; emigration and, 15; employment and, Model of Northeastern agricultural sector: 23; family reservation wage and, 159-60; activities (variables) in, 138-39; algebraic farm size and, 23; income and, 41, 69; land statement of, 137-44; bounds of, 144; com- reform and, 195, 206; percentage of land modity balances in, 140-41; consumption owned by, 218; percentages of, 22; share- levels in, 141; core of, 105-09; demand croppers and, 115, 116 and, 118-20, 141; deviation constraints Land reform: conclusions concerning, 210- (for risk) in, 142; differences in, from other Ii; cost estimate of, 206-08; development agricultural models, 27-28; facets of strategies and, 208-10; estates and, 29-30, Northeastern agriculture and, 28-30; farm 195, 198-99, 207, 211; Furtado report and, model linkages and, 116-18; farm type 16-19; legislative history of, 196-97; mod- aggregation and, 120-37; indexes in, 137- ule farm defined and, 197-98; need for, 38; individual farm models and, 109-16; 216-18; overview of, 195-96; performance labor constraints in, 143; land constraints of module farm and, 199-202; potential in, 142; land reform simulation in, 193- beneficiaries of, 205-06; potential effect of, 211; livestock maintenance balance in, 144; 202-05; productivity and, 88; in sugar objective function in, 140; overview of, zone, 14 n.34. See also Agrarian reform 104; parameter estimation and testing and, Land values, 254; analysis of, 47-50, 53; er- 145-49; parameters/data in, 139-40; ques- ratic nature of, 87; inflation and, 88; land tionnaire design and, 220-27; reason for reform and, 207 constructing, 102-03; research team and, Langoni, Carlos Geraldo, 3 219; sharecroppers in, 141, 143-44; valida- Law no. 4504 (land reform, 1964), 29, 196- tion procedures for, 150-51; working 97, 210 capital in, 144 Leuthold, Raymond M., 151 n.3 Modern farms, 123, 127. See also Farm mod- Linear discriminant functions, 135-37 els; Farm types INDEX 269 Municipios, selection of, 229, 230-33, 241- Pessao, Epitacio, 11 45 Piaui, 16, 47; zone delineation and, 31 Plantations, 27; development of, 6, 7; model Nakajima, Chihiro, 90, 91, 122 and, 104, 122, 125, 127, 166 National Agricultural Extension and Credit Policy: changes in, to benefit agriculture, Agency (ANCAR), 239, 240, 249 214-15; demand promotion and, 183-91; National Institue of Colonization and Agra- educational, 213-14; labor and, 215; mi- rianReform(INcRA), 29;modulefarmand, grants and, 215-16, 217, 218; program- 197-98, 200, 201, 205, 210, 246 ming model and, 102; for rural Northeast, National Study of Familv Expenditures 216-18; simulation of options in, 169-70. (ENDEF), 25, 73-81 See also Development National Department of Works against the Population: in Center-South, 3; droughts Droughts (DNOCS), 10, 11-12, 210 and, 9; economic projection and, 172; ex- Northeast Development Superintendency penditures of rural, 77; farm, 245-48; land (StDENE), 13; creation of, 14, 15; drought reform and, 17; of Northeast, 3-4, 36; ru- relief and, 19; field tests and, 223, 249; ral, 60 programs of, 17-18; 34/18 law and, 16. See Portuguese occupation of Brazil, 5 also SUDENE/World Bank survey Poverty: causes of, 23, 25-26; droughts and, Norton, R. D., 118, 149 11; in rural Northeast, 4, 67, 70, 77, 215 Nugent, Jeffrey B., 150 Prices: agricultural, 120, 160-63, 186; base Nutrition, 73. See also Diet case solution and, 173; demand interven- Nuts, 59 tion and, 188-91; land reform and, 203; land shadow, 165; market (model), 117, Occupants. See Squatters (occupants) 118, 152-55; short-term capital shadow, Oil palm, 216, 217 166; wage subsidies and, 182 Outputs, 103, 163; agricultural, 60-63; base Price support programs, 28-29, 178, 186 case solution and, 173; changes affecting Producers, 170-72 (model), 171; farm size and, 8 3-84; hous- Production. See Agricultural production ing and, 55; land reform and, 202, 203, Productivity, 30; farm size and, 82-89; land 206; model simulation of, 29. See also reform and, 202, 210. SeealsoAgricultural Crops, output from; Livestock, output and production Owner-operator (defined), 39. See also Land- Product markets, linkages among farm mod- owners els and, 117-18 Profit-maximizing behavior, 100-01 Parameters: data and, 139-40; estimation and Program for Land Redistribution and Stimu- testing of, 145-49 lation of Agro-Industry in the Northeast Paris, Quirino, 121 (PROTERRA), 18, 19, 210 Pasture land. See Grazing land Project options, simulation of, 169-70 Permanent workers, 103, 164, 172, 177, 191; defined. 39; employment and, 23; employ- Questionnaire (survey): data validation and ment and utilization of (model), 156, 158; 253-57, (sign of, vaidwon and expenditures and, 75, 77; family reserva- 253-57; design of, 220-27; fieldwork and, tion wage and, 159; farm models and, 107, 116; income of, 68; land reform and, 200, 205; percentages of. 22; wage subsidies for, Rausser, G. C., 121 179-83; work days of, 41. See also Share- Renters, 38; defined, 39 croppers; Temporary workers; Workers Research, 19, 176; policy and, 214-15 Pernambuco, farm selection example and, Resource use patterns, farm tvpes and, 123- 236, 237 25 270 THE AGRICULTURAL ECONOMY OF NORTHEAST BRAZIL Revenue deviation (model), parameters and, Structures (farm), 254; land reform and, 206, 147-48 207; value of, 54, 55 Rice, 58, 118, 217; farm family consumption Subsidies, 215; agricultural investment and, and, 109 214; employment and, 28; on wages, 179- Risk: deviation constraints for, 142; farm 82 models and, 110-11; influence of, 170, Subsistence sector, marketing and, 89-94 177-79; parameter testing and evaluation SUDENE/World Bank survey: as analysis of and, 147, 148; sharecropping and, 116 existing situation, 101; approach followed Robock, Stefan 11., il, 12 in, 19-20; family expenditures and, 7 3-74; Rural sector: diet and, 70-71; early social focus of, 4; income data and, 67; labor structure of, 6; family expenditures and, forceand, 38-41;landanalvsisand,41-53; 74-77; global expenditures and, 79-81; summary and conclusions of, 20-30; zonal growth rate and, 2 12-13; monetary ex- delineation and, 30-35. See also Northeast penditures and, 77-79; output in, 60-61; Development Superintendency (SUDENE); policy solutions and, 213-16; potential role World Bank of, 216-18; poverty in Northeast, 4, 67, Sugarcane crop, 13, 34, 59, 62, 87, 92, 104, 70, 77, 215 118, 216, 218; historical aspects of, 5-7, 8; land reform and, 14 n.34; production of Sampling procedures (survey), 228-48 (model), 153. See also Crops Sao Francisco Valley Commission (CVSF), 12 Scandizzo, P. L.., 110, 118, 147 Taxes, 187; land reform and, 196 Self-employed category, 75-77 Taylor, Hlarry W., 196 Sergipe, 47 Taylor, Lester D., 170 Sertao, 104, 153, 159, 193; agricultural de- Technology, 170, 208; crop yields and, 175- velopment and, 7; characteristics of, 33- 77; farm size and,96; model and, 28; policy 34; crop output and, 182-83;cvsF and, 12; and, 214-15; primitiveness of, 59 drought and, 9; farm models and, 107; Temporary workers, 103, 164, 167, 172, 175; farm size in, 49; industrialization in, 15; defined, 39; employment and (model), input costs in, 65; introduction of cotton 156, 157, 158; expenditures and, 75; farm to, 8; as physiographic zone, 21; share- models and, 107, 116; farm type and, 93; cropping and, 113-15 income of, 68, 70; land reform and, 199; Sharecroppers, 103, 164, 214, 215, 216; de- percentages of, 22; poverty of, 23; work fined, 39; early agricultural development days of, 41. See also Permanent workers; and, 6; employment of, 23, 28, 156, 158, Sharecroppers; Workers 172, 174-75; family reservation wage and, Testing procedures: for model, 15(05 1; para- 159-60; farm models and, 113-16, 117, meter, 145-49 127, 141, 143-44; farm type and, 95; in- 34/18 program, 16 come of, 68, 69; land reform and, 195, 205, Traditional farms, 123. See also Farm models; 206, 209-10; percentages of, 22, 38; as Farm types renters, 38; survey sample and, 239, 240; Transitional farms, 123, 125, 127. See also wage subsidies and, 182-8 3; work days of, Farm models; Farm types 41 Tree crops, 59, 171, 172; farm models and, Sisal, 59, 62 108; long-cycle, 165-66, 173. SeealsoCrops Skill training, 213 Trindade, Jose Augusto, 11 Slash-and-burn agriculture, 7,49; in Middle- North, 33; population engaged in, 38; in Underemployment, 16, 215; land reform West, 31 and, 17, 195; model performanceand, 157, Squatters (occupants), 38; defined, 39; settle- 158; seasonality and, 23; underuse of land ments of, 216, 217; survey sample and, and, 179. See also Employment; Unem- 239, 240 ployment INDEX 271 Unemployment, 17, 23, 93, 156, 157, 172. Yields: economic projection and, 171-72; See also Employment; Underemployment effects of improvements in, 175-77; land Urban sector, 87, 217; underemployment in, quality and, 52-5 3; parameter testing and 16 evaluation and, 147-48; policy and, 214- Usinas (sugar mills). 6, 44. See also Sugarcane 15; statistics for, 60. See also Crops crop Zona da Mata (forest zone), 34. 44. 59. 91, Validation: of model, 150-51; of data, 253- 153 57 Zones (Northeastern): capital per hectare and, 84; delineation of, 30-35; described, Wages: constraints at sectoral level and, 167; 21-22; equipment in, 55; farm size and, family reservation, 159-60; farm models 95-96; farm typology and, 127-3 5; income and, 105-07; input costs and, 64; subsidies and, 71; input costs and, 64-65; land re- on, 179-82 form and, 198, 199-201, 205; land values W'orkers, hired, 63, 68, 104, 191. See also and, 47-50; locational hypothesis and, 87; Permanent workers; Sharecroppers; Tem- market dependency and, 91-92; program- porary workers ming model and, 27, 103, 104, 116, 117, Working Group for the Development of the 118, 122, 157; structures in, 54 Northeast (GTDN), 13, 14, 15 World Bank, 176; field tests and, 223, 249. See also SUDENE/World Bank survey The full range of World Bank publications, both free and for sale, is described in the Catalog of WVorld Bank Publications; the continuing research program is outlined in WVorld Bank Research Program: Ab- stracts of Current Studies. Both booklets are updated annually; the most recent edition of each is available without charge from the Publications Unit, World Bank, 1818 H Street, N.W., Washing- ton, D.C. 20433, U.S.A. Gary P. Kutcher is an economist with the Development Research Center of the World Bank and is currently serving on the WVorld Development Report core team. Pasquale L. Scandizzo is a senior economist, formerly with the Development Research Center and now with the Agriculture and Rural Development Department of the World Bank. 0-8018-2581-4