M [[[[[[[ llEllll1/ll2ljl3llllL7/flllllllflllllslllEll/"ll!!! This is to certify that the dissertation entitled AN ECONOMETRIC SIMULATION STUDY OF THE EFFECT OF EXCHANGE RATE OVERVALUATION 0N BRAZILIAN AGRICULTURE presented by Doraci Heloisa Crocomo has been accepted towards fulfillment of the requirements for Ph.D. degeehlAgricuItural Economics MSU LIBRARIES \. RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. AN ECONOMETRIC SIMULATION STUDY OF THE EFFECT OF EXCHANGE RATE OVERVALUATION ON BRAZILIAN AGRICULTURE By Doraci Heloisa Crocomo A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1982 Q 1/7 NW ABSTRACT AN ECONOMETRIC SIMULATION STUDY OF THE EFFECT OF EXCHANGE RATE OVERVALUATION ON BRAZILIAN AGRICULTURE By Doraci Heloisa Crocomo Previous studies have suggested that overvaluation of the exchange rate has discriminated against Brazilian agriculture. This study was an attempt to measure the effect of this overvaluation on several important agricultural commodities. An econometric simulation model was constructed, where simul- taneous interactions between the consumer and farm levels of the market channels were possible. The econometric model linked supply, demand, trade, and government policy for the major agricultural com- modity markets. The model was then simulated to quantify the effects of the overvaluation of the domestic currency on Brazilian agriculture. The products considered include corn, rice, wheat, soybeans, soymeal, soyoil, beef, hogs, and chickens. For each commodity, func- tional relationships were specified, based on annual time-series data covering the l96l-8O period, for production, domestic disappearance, and stocks. Trade was considered the residual in all cases. The conceptual model focused on the effect of government intervention on domestic producer prices. A set of price relationships was estimated to explain the government intervention, based on economic variables. Doraci Heloisa Crocomo To analyze the impact of overvalued exchange rates on the con- sidered commodities, the model was simulated under two different sets of assumptions. In the first, the model generated a base forecast using the actual exchange rates for the 197l-80 period. In the second simulation, "equilibrium" instead of official exchange rates were used. The changes in the endogenous variables were attributed to the removal of the overvaluation, which averaged 27 percent a year for the simulated period (l97l-80). The results showed strong evidence of government intervention in domestic agricultural prices by the maintenance of overvalued exchange rates for corn and soyoil. Some indications of this kind of intervention were found for soybeans and wheat, and no evidence of such intervention was found for soymeal, rice, or beef. However, because of the interdependence between commodities, a removal of the price intervention on corn, wheat, soyoil, and soybeans would affect all Brazilian agricultural commodities. To my mother, Luiza. ACKNOWLEDGMENTS I wish to express my appreciation to Dr. Stanley Thompson, who served as my major professor, for providing excellent counsel throughout my graduate program and constructive criticism of this study. I also wish to express my deepest gratitude to Dr. Donald Mitchell, my thesis advisor, for his ideas, encouragement, and guid— ance. Without his help, I would not have been able to find solutions to the more difficult problems. Thanks is also extended to Dr. Vernon Sorenson and Dr. Darrell Fienup of my thesis committee. My apprecia— tion to Larry J. Connor and Dr. Normal P. Obst, who served on my guidance committee. I shall always be indebted to Dr. Eliseu Alves, president of the Empresa Brasileira de Pesquisa Agropecuaria (EMBRAPA) of the Ministry of Agriculture in Brazil, for providing the opportunity and financial support for my graduate program. Dr. Eliseu has been, above all, a warm friend. His encouragement and trust in me made it possible for this research to come to a happy ending. I wish to thank Dr. Edward Schuh, who helped me choose the thesis subject and gave me helpful suggestions. I also wish to thank Dr. Mauro R. Lopes, who was always willing to use his valuable time to discuss the subject with me and to make important suggestions. Appreciation is extended to Dr. Ignez V. Lopes, who made very useful information available to me. I am thankful to Chris Wolf and Celso Crocomo for their assistance in completing the computer work. Special thanks are due to Mrs. Sue Cooley for her careful editing and expert typing job. She worked without time limits to make it possible to finish this thesis. My gratitude is extended to my family in Brazil, for their encouragement and moral support, especially to my mother for her faith in me. She was the inspiration behind my desire to accomplish this goal. Finally, I am sincerely grateful to my husband, Celso, and my daughters, Roberta and Marina, for their love and understand- ing throughout the good and bad times. TABLE OF CONTENTS Page LIST OF TABLES ......................... vi LIST OF FIGURES ......................... viii Chapter I. INTRODUCTION ...................... l Background ...................... l The Problem ...................... 5 Objectives of the Study ................ 9 Organization of the Dissertation ........... l0 II. THEORETICAL CONSIDERATIONS ............... l2 Large-Country Case .................. l3 Small-Country Case .................. l8 III. OVERVIEW OF THE MODELING APPROACH ............ 21 Introduction ..................... 2l Simulation Model ................... 23 Estimation Procedures ................. 26 Validation ...................... 30 IV. SUPPLY ......................... 33 Introduction ..................... 33 Crop and Meat Production ............... 38 Crop Production ................... 41 Meat Production ................... 56 Validation ...................... 66 V. DEMAND ......................... 72 Introduction ..................... 72 Consumer Demand .................... 76 Soyoil ....................... 76 Rice ........................ 78 Wheat ........................ 8l Meat ........................ 83 Crush Demand (Soybeans) ............... Feed Demand ..................... Soymeal ...................... Corn ....................... PRICES, STOCKS, AND NET TRADE ............. Price Determination ................. Price Relationships: Soybeans, Soymeal, Soyoil, Rice, Corn, and Beef .............. Support Price of Wheat .............. Producer Prices of Hogs and Chickens ....... Retail Prices ................... Carryover Stocks .................. Net Trade ...................... Validation ..................... ANALYSIS OF THE MODEL ................. Introduction .................... ”Equilibrium" Exchange Rates ............ Simulation Analysis ................. Comparison Between the Baseline and Alternative Scenarios ..................... SUMMARY AND CONCLUSIONS ................ Summary ....................... Conclusions ..................... Suggestions for Further Research .......... APPENDICES .......................... VARIABLE DEFINITION AND SOURCES ............ HISTORICAL DATA .................... COPY OF THE FORTRAN SUBROUTINES USED TO SOLVE FOR THE ENDOGENOUS VARIABLES IN THE MODEL ........ BIBLIOGRAPHY ......................... vi TDD 108 109 llO ll4 120 120 132 132 I35 l37 I47 I47 l49 152 l54 T55 l60 168 173 Table l.l 1.2 LIST OF TABLES Brazil's Total Exports, Total Imports, and the Import of Petroleum as a Share of Total Imports ........ Value of Brazilian Exports and Share of the Major Sectors ........................ Brazilian Land Distribution Among the Regions and Agricultural Land Use ................. Elasticities of Hectarage Response ............ Parameter Estimates for Constrained Second-Degree Polynomial Lag Model. Dependent Variable: TPB ..... Simple and Cumulative Price Elasticities of Beef Slaughter ....................... Comparison of Cumulative Elasticities of Beef Supply in Three Studies ..................... U-Statistics for Area Equations ............. U-Statistics for Total-Production Equations ....... Average Elasticities and Cross-Elasticities of Demand at the Retail Level for Beef, Pork, and Broilers . . U-Statistics for Demand Equations ............ Parameter Estimates for the Soybean-Inventory Equations. Dependent Variable: ESSB ................ Parameter Estimates for Soyoil—Inventory Equations. Dependent Variable: ESSO ................ Parameter Estimates for the Soymeal-Inventory Equations. Dependent Variable: ESSM ................ U-Statistics for Producer-Price Equations ........ U—Statistics for Retail-Price Equations ......... Page 34 55 58 6O 60 7T 7T 86 93 ll6 ll7 ll8 U-Statistics for Stock Equations ............. Current Values of Brazilian Official and "Equilibrium" Exchange Rates and Degree of Overvaluation, 1971-80 Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Corn . Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Rice . . . Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Wheat Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Soybeans . Rate of Change Between the Results of the Baseline and Alternative Scenarios for the l97l-80 Period: Soymeal Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Soyoil . . Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Beef . . . Rate of Change Between the Results of the Baseline and Alternative Scenarios for the 1971-80 Period: Hogs . . . .10 Rate of Change Between the Results of the Baseline and .11 Alternative Scenarios for the 1971-80 Period: Chickens . Rate of Change Between the Average Results of the Baseline and Alternative Scenarios for the 1971-80 Period: All Nine Commodities .............. .12 Net Trade Average Results of the Simulated Baseline and Alternative Scenarios for the 1971-80 Period: All Nine Commodities .................. viii Page 128 135 138 138 139 139 140 140 141 141 142 142 146 LIST OF FIGURES Effect of Overvaluation of Exportable Products, Large-Country Case ........... . ....... Effect of Overvaluation on Exportable Products, Small-Country Case ................... Aggregate Simulation Model of the Brazilian Agricultural Sector ...................... . . . Map of Brazil: Geographical Distribution of the States . . Map of Brazil: Geographical Distribution of the Regions Actual and Simulated Values for Area: Corn, Rice, Wheat, and Soybeans, 1971-80 ................. Actual and Simulated Values for Production: Corn, Rice, and Wheat, 1971-80 ................... Actual and Simulated Values for Production: Soybeans, Soymeal, and Soyoil, 1971-80 .............. Actual and Simulated Values for Meat Production: Beef, Hogs, and Chickens, 1971-80 .............. Actual and Simulated Values for Demand: Soyoil, Rice, and Wheat Flour, 1971-80 ................ Actual and Simulated Values for Meat Demand: Beef, Hogs, and Chickens, 1971-80 .............. Actual and Simulated Values: Soybean Crushing Demand, and Feed Demand for Soymeal and Corn, 1971-80 ..... Actual and Simulated Values for Prices: Corn, Rice, and Wheat, 1971-80 ................... Actual and Simulated Values for Prices: Soybeans, Soymeal, and Soyoil, 1971—80 .............. Actual and Simulated Values for Prices: Beef Cattle, Hogs, and Chickens, 1971-80 .............. ix Page 14 19 24 35 36 67 68 69 7O 94 95 96 121 122 6.4 6.5 6.6 6.7 6.8 6.9 7.1 Actual and Simulated Values for Retail Prices: Rice and Wheat Flour, 1971-80 ................ Actual and Simulated Values for Meat Retail Prices: Beef, Hogs, and Chickens, 1971-80 ........... Actual and Simulated Values for Ending Stocks: Soybeans, Soymeal, and Soyoil, l971-80 .............. Actual and Simulated Values for Net trade: Wheat Imports, Corn, and Rice, 1971-80 p ................ Actual and Simulated Values for Net Trade: Soybeans, Soymeal, and Soyoil, 1971-80 .............. Actual and Simulated Values for Meat Net Trade: Beef, Hogs, and Chickens, 1971-80 .............. Official and "Equilibrium" Exchange Rates: Base, January 1980 ................... Page 124 126 130 134 CHAPTER I INTRODUCTION Background The crisis that was started in 1973 by OPEC's decision to raise international petroleum prices brought a variety of consequences for the oil-importing countries. In the Brazilian case, the decision was particularly harmful because the country produces only about 20 percent of its petroleum needs. The value of total imports more than doubled in 1974 in comparison to the previous year. In the same period, exports that had been experiencing a steady increase, mainly due to the government's effort to promote expansion and diversifica- tion, grew only 12.8 percent-—that is, from US$6,199 million to US$7,951 million, as shown in Table 1.1. The situation has not changed much since 1974, despite the country's effort to curtail imports and to increase exports. The share of petroleum imports in relation to the total import bill has increased over time; it now constitutes almost one-half of the total (Table 1.1). The main problem is that means of transportation depend almost exclusively on petroleum, and the dependency is not expected to decrease in the near future. The second largest item on the import list is wheat, and again Brazil depends heavily on imports to supply its internal demand. Furthermore, the importing of capital goods cannot be cut severely without harming the development process. 1 Because not much can be done to curtail imports, the best way to improve the balance of trade is by increasing exports. Table 1.1.--Brazil's total exports, total imports, and the import of petroleum as a share of total imports (05$ million FOB). Year Exports Imports Pfiéggligm Share (%) (1) (2) (3) (3/2) 1970 2,740 2,508 236 9 1971 2,904 3,245 327 10 1972 3,991 4,235 409 10 1973 6,199 6,192 711 11 1974 7,951 12,642 2,840 24 1975 8,670 12,211 2,875 24 1976 10,128 12,347 3,613 29 1977 12,120 12,023 3,814 32 1978 12,659 13,639 4,196 31 1979 15,244 17,961 6,403 36 1980 20,132 22,960 9,405 41 SOURCE: Conjuntura Econ6mica, various issues. The balance-of—payments problem faced by Brazil is accompanied by a very high internal rate of inflation with consequent deteriora— tion of the income distribution in a country in which the annual rate of population growth is still around 2.5 percent. It an attempt to alleviate these problems, the government that came to power in 1979 has selected the agricultural sector as its top priority. in policy in favor of this sector is believed to be underway. For the first time in Brazil's history, a change After World War 11, Brazil experienced a very high rate of growth as a result of the import-substitution industrialization policy that was adopted. This policy was guided by the common belief that this was the only way under-developed countries could achieve economic develop- ment (Prebisch, 1949). By 1963, policies to promote the expansion and diversification of exports as a source of economic growth were adopted, and the manufacturing sector was the one chosen to benefit from such policies. The demand placed on the agricultural sector now is expected to be very heavy, due to the need for exports for foreign-exchange earnings and the needs of the domestic market. The supply of food for the domestic market will have to increase by more than the present expansion rate in order to have abundant and cheap food. Although the historical contribution of agricultural products to the export sector has been decreasing compared to that of manufactured products, it is still responsible for a large share of the export market, as shown in Table 1.2. Agriculture is believed to be the best way to increase exports because of certain characteristics of the sector. First, the world demand for agricultural products is expected to continue to grow at least at the same rate as the population will grow. Second, most of the oil-importing countries are experiencing various degrees of recession, making it hard for the developing nations to sell their manufactured products. Finally, Brazil is only a mar- ginal exporter for the majority of its agricultural products, so the increase in exports can be achieved without influencing international prices. Therefore, no other sector has the opportunity for advance- ment like the agricultural sector does. Table 1.2.--Value of Brazilian exports and share of the major sectors (US$ million FOB). Year ETota] Agriculturea Manufactured Semi-Manufactured xports Total % Total % Total % 1970 2,740 2,050 75 420 15 250 9 1971 2,904 1,990 69 580 20 240 8 1972 3,991 2,720 68 910 23 310 8 1973 6,199 4,100 66 1,470 24 480 8 1974 7,951 4,810 60 2,330 29 630 8 1975 8,670 5,030 58 2,580 30 850 10 1976 10,128 6,130 61 2,770 27 840 8 1977 12,120 6,930 57 3,840 32 1,040 9 1978 12,659 5,980 47 5,080 40 1,420 11 1979 15,244 6,510 43 6,680 44 1,890 12 1980 20,132 8,488 42 9,028 45 2,349 12 SOURCE: Conjuntura Econ6mica, various issues. NOTE: During the 19605, the participation of the agricultural sector remained around 80% (Miller Paiva, p. 47). 3The contribution of agricultural products is undervalued because the participation of agricultural raw materials in the semi- manufactured sector is not shown. To increase the supply of agricultural products to both the internal and the external markets, it is necessary to create favor- able conditions for increasing supply. The agricultural sector is known for its ability to respond very quickly to price incentives, so small price increases can induce positive response in production. However, this response must come mainly through an increase in yields. Such an increase is possible, though, because technology is already available for some important crops and depends only on the right incentives to be profitable. Until recently, it has been argued that one of the few gov- ernmental actions that benefited the agricultural export sector, as well as other sectors, was the mini-devaluation policy ("floating- peg") adopted in 1968 (Zockun et a1., 1976, p. 44). This policy helped to decrease the instability of the export activity by reducing the speculation associated with the previous devaluation policy. The mini-devaluation policy coincided with a period of favorable inter- national prices for some agricultural products and had positive effects on the export sector. The Problem Although the agricultural sector has typically been the sector responsible for assisting the development process through the transfer of its surpluses, only recently has the Brazilian government decided to select this sector as the one to receive top priority in its development strategy. Historically, the agricultural sector has been discriminated against by economic policies aimed at protecting other forms of development, such as the industrialization for import sub- stitution or the export promotion of manufactured goods. A series of restrictive policies such as export quotas, multiple exchange rates, overvalued exchange rates, and even more complicated methods of licensing and deposit were some of the ways governmental policies penalized the agricultural sector. At the same time, the industrial sector received different forms of subsidies such as preferential treatment concerning the import of capital goods, raw materials, and other inputs; preferential exchange rates; and tax exemption, fol- lowed by massive public investment to stimulate the sector.1 Besides the economic measures directed toward the export market, agriculture also suffered from domestic intervention such as controlled or fixed prices, retention of stocks, etc. These policies had effects similar to the measures directed to the external market in the sense that their main objective was to discourage exports until the domestic market had been "perfectly” supplied.2 The consequences of these restrictive policies directed toward either the external or the internal markets or both were to drive resources out of the agricultural sector. This resulted in fewer exports and less foreign-exchange-revenue earnings. Valdes (1973) found a similar situation in Chile in a study of the effects of restrictive agricultural policies. Much of the "trade gap" in 1Several exhaustive studies have been conducted concerning the measures adopted by the government during the post-World War II period. They clearly demonstrated that the agricultural sector was being penal— ized. Among them are Veiga (1974), Bergsman (1970), Zockum et a1. (1976), and Lopes and Schuh (1979). 2Leff (1967, 1969) claimed that during most of the post-World War II period and up to 1967, Brazil followed an "exportable surplus” approach to trade, by which a country exports the "surplus" that is "leftover”after the domestic market has been "adequately" supplied, even if internal prices are lower than the world—market prices. agricultural products in that country in the post-World War 11 period was a result of the international commercial policies followed by the Chilean government. Of all the Brazilian policies adopted, the most persistent was the maintenance of an overvalued exchange rate. This measure constituted an implicit export tax on the agricultural sector. There seem to have been several reasons for maintaining this policy, as suggested by Thompson and Schuh (1977): (a) to exploit the country's dominant position in the world coffee market, where it was possible to shift the tax onto the foreign consumer; (b) to keep the domestic price of export products below their opportunity costs, as a way to control domestic inflation; and finally (c) to keep the price of imported goods low to encourage industrialization. Bergsman (1970) stated that during the 1947-67 period "the overvalued exchange rate probably owed its existence at least as much to a desire to keep food prices down as to a desire to industrialize" (p. 152). Several researchers have shown the persistence of the over- valuation in the last three decades. Bergsman (1970) and Bergsman and Malan (1970) estimated the magnitude of this implicit taxation during the 1954—66 period as being between 23 and 27 percent. Bacha et a1. (1971) estimated that during the first years of the 1960s the exchange rate was about 20 to 25 percent overvalued. Schuh (1976) pointed out that after being close to equilibrium in 1970, the exchange rate was overvalued again by 25 percent by 1976. By the end of 1979, the government decided to devalue the Brazilian currency (cruzeiro) by 30 percent, leading it to an equilibrium position and breaking the rule adopted in 1968 to devalue constantly and by small amounts (mini-devaluations). However, after a couple of months the currency again became overvalued because the mini- devaluation policy was not able to compensate for the inflation dif- ferential between Brazil and the United States. In a classic paper, Schuh (1974) argued that an important variable in understanding the agricultural problems faced by the United States during the 19505 was the exchange rate and its role in trade, in the valuation of resources within the U.S. economy, in the distribution of benefits of economic progress between consumers and producers within an economy, and in the way the benefits of technical change are shared between the domestic population and the world at large. (p. 1) In the same paper, Schuh explained why overvaluation tends to have detrimental effects on the agriculture of low-income countries, in which new production technologies are not available. In this situa- tion the effect of overvaluation can be so strong that countries can move from being net exporters to being net importers of agricultural exports. Several studies of the effects of overvaluation and other restrictive trade policies over particular agricultural and livestock commodities were conducted for the Brazilian case. Thompson and Schuh (1977) concluded that during the 1947-70 period, Brazil, which had been only a marginal exporter of corn, could have exported an average of 1.26 million metric tons more of corn each year if the cruzeiro had been devaluated by 20 percent. In effect, the increase in export earnings from corn would have been 475 percent if the exchange rate had been kept near an equilibrium level. Ayer and Schuh (1971) found that for each cruzeiro in consumer suplus gained in the cotton market through export restrictions, the country paid over 2.5 cruzeiros in foregone export earnings. Studying the beef sector in Brazil, Lattimore (1974) concluded that because of the intervention policy, each .45 cruzeiro saved by the Brazilian consumer caused the producer to be taxed one cruzeiro. Based on the evidence that restrictive trade policies have discriminated against the agricultural sector, it appears that if a less-restrictive set of economic policies and, more specifically, a policy of more-near-equilibrium exchange rate had been adopted throughout the years, Brazil could now be a steady exporter of sev- eral agricultural products. The present study is an attempt to measure the effect of these restrictions over a system of several important commodities, in which simultaneous interactions between the consumer and farm levels of the market channels are possible. Objectives of the Study The specific objectives of this study are: 1. To build an econometric model that links the supply, demand, trade, and government sectors in Brazil for the major agri- cultural commodity markets. 2. To integrate this econometric model into a model system that can be used for prediction and policy-analysis simulation, i.e., for testing the operation of stabilization schemes under alternative assumed conditions. 10 3. To quantify the effects of overvaluation of the domestic currency on Brazilian agriculture. The general objective of this study was to increase the under- standing of the markets for the agricultural products considered and to provide instruments for their analysis. More specifically, it was attempted to explain the historical discrimination against the agri- cultural sector through a persistent overvaluation of the Brazilian currency. The products considered in this study included corn, rice, wheat, soybeans, soymeal, soyoil, beef, hogs, and chickens. These products were chosen mainly because of their importance in the domestic economy and because they are interrelated, competing for the same area and/or as substitutes in terms of consumption. No mention is made of other important agricultural products, such as coffee and sugar cane, in order to keep the model manageable and because the interest was mainly concentrated in the grain-livestock sector. Organization of the Dissertation The remainder of the dissertation is organized as follows. Chapter II contains the theoretical considerations used to demon- strate the effects of changes in trade policy. Chapter III includes an overview of the modelling approach. The results of the estimation of the parameters and validation of the estimated model are presented in Chapters IV through VI. Production is discussed in Chapter IV, demand in Chapter V, and net trade and price and stock relationships 11 in Chapter VI. Chapter VII demonstrates the use of the model for policy analysis. The final chapter contains a summary, conclusions, and suggestions for further research. Supporting materials are contained in the appendices. CHAPTER II THEORETICAL CONSIDERATIONS One of the objectives of this study was to quantify the effects of the overvaluation of the domestic currency on Brazilian agriculture. These effects can be shown graphically with the help of a partial-equilibrium framework. The use of partial-equilibrium analysis, although subject to several criticisms, can be very useful for "rough and ready" analyses of the effects of changes in trade policy (Thompson, 1977). The partial-equilibrium approach considers only one commodity in isolation and does not include other markets in the analyses. But, to link the agricultural sector with the other sectors of the economy and still have a comprehensive system, simpli- fying assumptions are required. To illustrate the effects of an overvalued exchange rate in a one-commodity, two-country model, a set of two-dimensional graphs can be used. The more general case, whereby the exporting country has a sufficiently large world-market share to influence the world- market equilibrium price by changing the quantity it exports, will be presented first. Because Brazil can still be considered a small country in world markets for most of its agricultural products, the small-country case will be presented following the more general case. 12 13 Large-Country Case1 Figure 2.1, panel (a) represents the domestic demand and supply curves of a certain homogeneous commodity of an exporting country. When there is no international trade, the equilibrium price and quantity are determined by the intersection of the domestic supply and demand curves. However, if there is free trade, the domestic price at the port of the exporting country is equal to the F08 (free on board) price. Panel (c) of the same figure shows the "domestic" supply and demand curves for the importing country, considered here as the rest of the world (ROW).2 The intersection of the domestic demand and supply curves gives the "domestic" equilibrium price and quantity for the ROW3 before trade begins. The prices are expressed in domestic currency for the exporting country and in dollars for the ROW. It was assumed that the exchange rate was in equilibrium and that transportation costs could be ignored for simplicity. Panels (b) and (b') represent the world-market conditions in the exporting country's domestic currency and in dollars, respectively. An excess-supply curve (ES) can be drawn, panel (b), on the assumption that at any price above OP] the quantity supplied domestically is 1This section is heavily based on Thompson and Schuh (1977). 2The aggregation of all other trading countries into the "rest of the world" constitutes another important limitation of this partial-equilibrium graphical approach. See Thompson (1977, p. 10). 3The curves were drawn on the assumption that before trade, the domestic price in the exporting country is lower than the equi- librium price in the ROW, the only reason why a country can be an exporter. 14 .mmmo zsgcsous mmsmp .mpozeosa m—nmpsoaxm co cowpm=Fm>Lm>o mo pummwm--.F.N mszmwd on o_soz ogo Lo boom A.ov posses orsoz a a H u o or a m o 4. Mair" Molina 1, “.llllluvm . . . - . . m M m n m m 29.3 _. . " om H m m H n . sh... .. . ................. s. u m .................... o 3 Led. - ._ :--.-l.--..--m¢ l-l-.- Na 30m % .mm u If wmsm Any pwxgms upsoz mmsa Amv asucaou mcwusoaxm or d m o m I a < o h H H . n ” .om “ m m m m u m .. ............... H ..... .. om . u H . mm .Iololnovllaooooopooltu ooooooooo . cuckolfln mm IIIIOIIIIIIIollI MQoooallilllll 0. mm m mm f s m a 15 greater than the quantity demanded. Then ES is obtained by measuring horizontally the differences between domestic supply and demand curves at prices above the equilibrium price and shows the quantities a country will export at various prices. The slope of ES is the combined slope of the domestic supply and demand curves of panel (a). The excess-demand curve (ED) for the rest of the world can be obtained in a similar manner. At prices below the "domestic“ equilibrium price 0P2 in panel (c), more is demanded and less is supplied, assuming that "domestic" supply and demand curves are normally shaped. ED is then obtained by measuring horizontally the differences between the "domestic" supply and demand curves for the ROW at any price below the equilibrium price and shows the quantities the ROW is willing to import from the exporting country at various prices. The derivation of this curve is not as straightforward as the ES curve because it has to be considered as the difference between total imports of all coun- tries less the quantities imported from other exporting countries, so it is in fact a net excess-demand curve. The slope of ED is the combined slope of the "domestic" supply and demand curves of all coun— tries aggregated in panel (c). If it is assumed there are no trade restrictions, that the exchange rate is in equilibrium, and that transportation costs can be ignored, a new world-equilibrium price can be obtained at the inter— section of ES and E0 in panels (b) and (b'). This free-trade price will be equal to the domestic equilibrium prices at the ports-~0P3 in the exporting country and OP4 in the ROW. At this price, the l6 exporting country will produce OB and 0A will be domestically demanded, leaving AB = OF to be exported. At the same time, the ROW is producing DC at the price 0P4, consuming OD, and importing CD = OF. The exporting country's foreign revenue from this scale is equal to CMND, which in turn is equal to AKLB. Up to this point, perfect competition was assumed to exist among the markets, and free trade was allowed. Now assume that the exporting country decides to peg its exchange rate to the dollar at a level that overvalues its currency. This shifts the ED curve of the ROW down to E0' in terms of the exporting country currency. The exporting country ES curve shifts up to ES' from the viewpoint of the ROW, in terms of dollars. These shifts are not parallel but percentage shifts. The immediate effect of this policy decision is to lower the price of the commodity to OP5 in terms of the exporting country currency or to increase it to OP6 in terms of dollars. This means that less will be produced and more consumed domestically, leaving less to be exported. 0n the other hand, in the ROW more will be produced and less consumed because the prices now are higher in terms of dollars. The quantity traded now is OH = 0E = IJ, which is smaller than before the overvaluation, meaning that less foreign- exchange revenue will be earned by the exporting country (ITUJ). One way to measure the average effect of a price change as a result of a trade intervention such as an overvaluation is by using the elasticities of ES and ED. These elasticities, in turn, depend 17 directly on the domestic-price elasticities of supply and demand of the respective countries involved.1 Thompson and Schuh (1977, p. 9) showed that an overvaluation of t percent, ceteris paribus, will lower the domestic price in the exporting country by d—.: <5.» . = (—l—) t percent 1.3 e and will raise the domestic price in the ROW by d P fi = (fig—)1; = (fl-8:) t percent n where: n = the price elasticity of excess supply of the exporting country 6 = the price elasticity of excess demand of the ROW (n > 0 > E) The effectiveness of such a policy will depend on the power of the country to influence the international market price. If the export- ing country is large enough and overvalues its currency or imposes any other kind of trade restriction, it can transfer income from the ROW to the exporting country by increasing the world-market price (in dollars). The magnitude of these changes in prices greatly depends on the elasticities of excess demand faced by the exporting country for each different exportable product. The more inelastic the excess-demand curve of the ROW is for the product considered, 1For derivation of the ES and ED elasticities, see Kreinin (1975), Appendix III, p. 428. 18 the more the exporting country stands to gain from such a restrictive trade policy. Small-Country Case This is a particular situation of the more general case, whereby the exporting country faces a horizontal excess-demand curve for its product (a =-»w). The above formulas will then be reduced to: lim 5LT: P a» _m = t percent and lim d P, e+-w jiié=0 This means that a t percent overvaluation, ceteris paribus, will translate into a similar amount of decrease in the exporting country domestic price with no influence at all in the world-market equilibrium price. The effects of an overvalued exchange rate in a small—country case are shown graphically in Figure 2.2. The analysis is similar to the large—country case. Panel (a) illustrates the exporting country's domestic supply and demand curves for a certain homogeneous commodity. Panel (b) represents the world-market conditions faced by the export— ing country. ES is the excess—supply curve for prices above the domestic equilibrium level P]. ED is the ROW's excess-demand curve, when the exchange rate is in equilibrium. The intersection of these two curves in panel (b) gives the domestic free trade price P2. At this price, the exporting country produces OB and demands OA 19 .mmwu Aswczoouprmsm .mpozvosa mpnwpsoaxw co covumz—m>sm>o do gumwmmuu.m.m mszmw. An. boxgos u_soz Am. zgpczoo mcwpsonxm m 8 . o m m. m. .. < o - .ll-111v” 4 low . will” . m . . . . m . r. .9. m . ...... m .......... m.. 8 . N...-2.-... .ENE-.. a mu .fim ..m 20 domestically, leaving AB = CC to be exported. The revenue of this sale is AGHB. If the exporting country overvalues its exchange rate, the ED curve shifts down to ED'. The new domestic equilibrium price will be 0P3 < 0P2. At this lower price, OE < OB will be produced and 00 > 0A will be demanded domestically, leaving DE = OF < DC to be exported. This means that a lower revenue DIJE will be earned. If a country has no power to influence world-market prices, the effects of the overvaluation tend to stay within the domestic economy. Consumers will gain in the short run because the prices in domestic currency will fall. Lower prices will result in undervalua- tion of the agricultural resources, which in time will leave the sec- tor to find better alternative uses. Also, if prices are lower domestically, more will be consumed internally and less produced. The amount available for trade will decrease, with a consequent decrease in foreign-exchange earnings. Imports will also be cheaper in terms of domestic currency, and the government may have to impose restrictive measures to control imports. The country's capacity to import the capital goods and other inputs necessary to its develop- mental process may also be affected. How much foreign exchange will be lost depends on the elasticities and the amount of the overvalua- tion. CHAPTER III OVERVIEW OF THE MODELING APPROACH Introduction In recent years a variety of complete econometric models have been built, having different objectives and using a series of estimation techniques. These models usually provide ways for better understanding the structure and parameters of the behavioral relation— ships underlying commodity markets as well as the relationships between the various markets and the producer and consumer economies (Adams & Behrman, 1976, p. 2). They have been used for purposes of forecast- ing and for simulation under different policy alternatives. Labys (1973, 1975) has an extensive inventory of studies related to commodity-model building. The objective of this study was to build an econometric model for the grain and livestock sectors to be used for prediction and policy simulation. The model includes production components (acreage, yields, slaughtered animals), domestic-demand equations, feed-demand equations derived from the livestock component for soymeal and corn, soybeans crush demand, a system of price-linkage equations relating domestic to world prices, stock equations where applicable, and a set of identities and ”technical" relationships. In all cases, trade was considered the residual after considering domestic availa- bility and domestic disappearance. 21 22 A simulation approach was chosen because this procedure allows the major agricultural products to be analyzed simultaneously. This enables important cross-price effects among commodities to be con- sidered. The simulation analysis has some advantages over the alter- native multiplier analysis to evaluate effects of policy changes (Labys, 1973, p. 199). It permits one to consider, among other things, varying rates of change in an exogenous variable or varying levels ofseveralexogenous variables at once. It was hoped that this procedure would show more accurately the total effects of a change in the exogenous variables than would the use of simple elasticities computed from the structural form. To model commodity markets, it is necessary to know the char- acteristics of each commodity. This means that different conditions of production, transport, and marketing, as well as government inter— vention, must be taken into consideration (Adams & Behrman, 1976, p. 3). To use this kind of detailed information when building large econometric models is not always feasible because of data and resource limitations. Bearing in mind the limitations in constructing any econo- metric model, it was hoped that useful conclusions could be reached with a relatively simple econometric model built from data available mainly in agricultural balance sheets. A complete description of the model is presented in subsequent chapters. 23 Simulation Model The basic conceptual approach to the model for each commodity was similar. Functional relationships were estimated for domestic disappearance (including food and industrial use, feed, and seed), carryover stocks, production, and domestic-price determination. The demand for exports was considered as a residual. The role of government intervention in determining domestic agricultural prices was included as a set of price relationships relating domestic to world prices and exchange rate. These price relationships also included other economic variables assumed to influence the govern- ment's decision to intervene in domestic prices. Following a study by Reynolds, Heady, and Mitchell (1975), a simplified diagram was developed in which the most relevant inter- relationships among the different components of the model are pre- sented. (See Figure 3.1.) The most important components of the economic model of the Brazilian grain-livestock subsystem are pro- ducers, domestic consumers, the government, and foreign consumers. Because of the recursive and simultaneous nature of the system, a change in one of the submodels can affect the whole system. An increase in beef-cattle price, for example, will affect the other livestock products and will even show effects beyond the sector. Direct and indirect effects will be observed in the hog, poultry, corn, and soybean submodels. A change in the price of beef cattle affects the soybean acreage through changes in the area devoted to grazing. A change in the price of soybeans affects the demand for Brazilian soybeans. The domestic soymeal demand is mainly determined 24 .sopomm Fmszppzuwsmm cowprmsm 85p mo Fwnoe covpmfizswm mummmsmm509:. .a mou_em oo_Lm oo.cm >0m anus: ou_¢ 0 0E : owes : —0Ugn-3W — HUQUr—D: W f K — ”up“ w coon>om L\4 # - vs; uoz coon>03 : _ P. a: 3.39... vcmeoo swoon: 3 Woo .L; 2533 >om ocnsoo >om Doug: soon: _ ouox _.o.mx oo_¢ Luscu . l. vcoeon vcmsoc _ _o>0m 30.5.5 8 . uc. . do; .nau gmaeu o.uuou moo_cm .0805nam ocumo>_4 xoOumo>_4 I a: u A .5 .2... Geneva c0506 Mw. oww .31 “0.5 o xcoa .Loom .auOH .muOP co_um_:aom .3933 >£3.58 m3 ou_¢ .0 ou_cm ccou ou_cm .moE>0m Lo__oLm.xLom.woom .oooEAsm :LOu acmeoo cLoU oochuoz. :LOU A co_uaE:mc0u mu_amu com ”EDUC— U\& ou_La xooumo>_4 ..ooox 25 by the demand for broiler feed, and demand for soyoil is largely determined by the demand for cooking oil. The volume of soybeans and products to be exported is determined mainly by the difference between the respective supply and demand. The same soybean price determines the relative profitability of soybeans, which in turn affects the decision to cultivate soybeans and competitive crops the following year, when interaction of the total demand and supply of soybeans and government intervention produces a new market price. The submodels for the other commodities follow a similar pattern. Additional interactions are generated in the system as the continuous effects of changes in other sectors are performed. The interaction and feedback effects among the products considered (corn, rice, wheat, soybeans, soymeal, soyoil, beef, hog meat, and broilers) are represented by the underlying recursive and simultaneous equa- tions of the model. The preceding submodels were simulated simultaneously over the period 1971—80. Actual lagged values were used to force the model to correct itself. Interactions among the submodels were thus per— mitted. Model evaluation over the historical period was analyzed by performance measures, which are presented at the end of this chapter. These measures relate to the model's predictive ability. The perform- ance results are presented in the next three chapters, after a dis- cussion of the estimated equations. Next, the model is simulated over the historical period under two different scenarios. A baseline forecast was obtained by letting 26 the model feed itself, using estimated lagged values instead of actual values. The predictions from this scenario are then compared with those from an alternative scenario designed to allow analysis of the effect of a change in the level of government intervention by using l-'equilibrium" exchange rate. The results of those simula- tions are presented in Chapter VII. The simulations described in this study were carried out using the GSIM' program developed in the Agricultural Economics Department at Michigan State University. GSIM uses the Gauss—Seidel straightforward numerical method to obtain the solution of simul- taneous systems of equations. The full model consists of 48 equations, 37 of which are structural equations. The remaining 11 equations are technical rela- tionships and identities. Estimation of the econometric equations was based on yearly time series data covering the period from 1961 through 1980. It was felt that this was the most representative period to reflect the current structure of production, consumption, and trade in Brazil. All of the monetary variables used in the analysis were deflated by the respective price index with base 1977 = 100. For a detailed description of the variables entering the equa- tions and data sources, see Appendix A. Estimation Procedures The choice of an estimation procedure depends on the char- acteristics of the model under analysis. Using practical considerations 'Based on the General Analytical Simulation Solution Program (GASSP) developed originally at the USDA. 27 of estimating and applying commodity models’iLabys (1973, p. 135) established the choice of criteria based on theoretical and computa- tional complexity, required sample size, equation or system approach, nature of consistency, and sensitivity to specification errors, identification, and multicollinearity. His comparison scheme is a useful device in making the tradeoff between computational ease and desirable properties of estimates. Ideally, where there is simultaneity in a commodity model, one should apply specific methods for use with simultaneous systems. In this study, it was possible to order the supply (production) equa- tions in a dynamic recursive system in such a way that the dependent variables were related to exogenous and predetermined variables. The ordinary-least-squares procedure is then suitable for estimating the parameters of the supply equations because it can provide esti- mates that are unbiased, consistent, and efficient, as well as giving the maximum likelihood for a recursive system of equations (Labys, 1973, p. 135). The two-stage least-squares method seems to be adequate for estimating demand functions where the interdependence among demand and prices extends to supply. This method, however, fails when the number of observations is less than the number of predetermined variables. Most of the time, this was the case in the present study. It is impossible to estimate the reduced-form coefficients in the first stage of the process because the number of unknowns is greater 28 than the number of observations. Kloek and Mennes (1960) proposed a solution to handling this problem, using principal components of predetermined variables. One can replace the matrix of predetermined variables with one of a smaller number of principal components account- ing for a large proportion of the total variance in the original vari- ables. Although it has some merits, this method also includes several disadvantages. The linear combinations of the principal components have no meaningful economic interpretation. Also, the parameter estimates cannot be expected to be consistent as in the case of origi- nal two-stage least-squares procedures (Labys,1973,p. 142). Anotherdis- advantage is the arbitrariness caused by (a) the fact that the computation of principal components needs a certain normalization rule on the variables (such as a unit mean square) and (b) the number of principal components to be used and whether these components should pertain only to the variables excluded from the equation to be esti- mated or to the set of all predetermined variables of the system (Theil, 1971). Besides, as Maddala (1977, p. 194) pointed out, the principal-components method is only a statistical device and of very limited use, and it can easily be misused in econometric work. According to Theil (1971), there are no obvious solutions in a case of an undersized sample. He presented two other approaches but cautioned that further research is needed before the merits of such procedures can be assessed. The number of predetermined vari- ables entering each reduced-form equation is very large for considerably or even moderately sized models, so that methods specifically designed 29 for use with simultaneous systems such as two— or three-stage least- squares procedures are almost never employed in practice (Brundy & Jorgenson, 1971). It is common to assert that identification does not consti— tute a serious problem with agricultural-commodity models because most of these models are overidentified. Usually, the only rule used for identification is the order (necessary) condition. The rank (necessary and sufficient) condition is normally neglected because it requires a profound knowledge of certain aspects of matrix algebra. The subject should not be ignored, however, because the parameters of an equation sometimes are not identifiable even when their number does not exceed that of the exogenous variables (Theil, 1971, p. 449). Besides the difficulties described above, it is not worthwhile to pursue a solution for the simultaneity problem when one does not know whether the system of equations is identified or not. Based on the preceding considerations, it was decided to estimate most equations by the ordinary-least-squares (OLS) method. Exceptions were made in the case of beef production, in which a polynomial-distributed—lag model was used; and when serially cor- related errors were detected, the generalized-least-squares method was used. The decision to use the OLS method may lead to simultaneous equation bias in the coefficients. Given the relatively small number of observations for the present study, and, more specifically, the limited number of years for some of the products considered, such as broilers and soyoil, that were relevant in Brazil, it cannot be assured that the use of a more consistent method of estimation (if possible) 30 would result in a significant change in the coefficients. According to Labys (1973), It should be kept in mind that OLS results in biased parameters estimates where simultaneity is present, but the same is true of estimates from other methods unless the given sample size is relatively large and the model specification is fairly exact. Thus, one must weigh resulting bias against the fact that OLS estimates typically have lower variance than estimates obtained from other methods. (p. 138) Given the preceding discussion, it was expected that the use of OLS may still provide meaningful results. It was decided that it was more profitable to improve the model in terms of specification than in terms of methods of estimation, which can be left for later research in which refinements in the data could be obtained, together with alarger number of observations. Validation Once the model equations are estimated, the full—model evaluation over the historical period is analyzed by different per- formance measures. They are related to the mean square simulation error. The first is Theil's inequality coefficient, defined as T /l 2 (vS - r1)2 U' = v/l Tt='s 2t 1 T a 2 Ttgl (Vt) + Ttél 0") where Y: = simulated or predicted value of Yt Y: = actual value T = number of periods in the simulation 31 The numerator of U1 is the root of the mean square error (MSE). This statistic lies between zero and one. If it is zero, the model predicts perfectly for the historical period. According to Maddala (1977,p. 346),lh does notprovide a good ranking of forecasts; thus a second measure, Theil's U2 statistic, is preferred. It is defined as u = «FWEET' 2 / ii, (4)2 The value for U2 ranges from zero to infinity; again, this measure is zero in the case of perfect forecasts. The mean square error can be decomposed into either one of two sets of components: (1) bias, variance, and covariance and (2) bias, regression, and residual proportions.' Maddala pointed out that the latter decomposition should be used; it is defined as: UM = (S ' a) = bias proportion MSE Ss - r Sa)2 . UR = XTT'TEEFTTTTT = regression proportion 2 2 00 = ' MSE )Sa = disturbance proportion where 'E,‘E = the means of st and at defined as _ s _ a a St ‘ (Yt Yt-l)/Yt—l 'For a detailed discussion of this decomposition, see Maddala (1977, p. 344) and Pindych and Rubinfeld (1981, p. 365). 32 = a _ a a (Yt Yt-1)/Yt-1 $5 = t 3" e variance of the predictions Sa = the variance of the actual values 1 ll (f 3" e correlation between s and a If one considers the regression of actual values on predicted values Ya = S t O'J'B't UM will be equal to zero if the parameter a is zero, and UR will be zero if 8 equals one. The bias proportion UM gives an indication of the systematic error--that is, the extent to which average predicted change deviates from average actual change. Note that UM + UR + UD = l. The ideal distribution of these proportions is UM = UR = O, and UD = 1. The above criteria were used to validate the model. The results are presented in the next three chapters, together with the structure and estimation of the model components. CHAPTER IV SUPPLY Introduction This chapter focuses on the production side of the model. To situate geographically the distribution of production within Brazil, a brief idea of the land use is given below. The succeeding sections are devoted to specification, estimation, and validation of the structural relations of the underlying supplies for each product con- sidered. The extension of the Brazilian territory is 851.2 million hectares (8,511,965 kmz). Of this area, approximately 322.6 million hectares (38.2 percent) was occupied by agricultural establishments in 1975, of which only about 70 percent or 227.8 million hectares was under full exploration, distributed as follows: 38.8 million hectares in arable land and permanent crops, 164.9 million hectares in permanent pasture, and 24.0 million hectares in extractive exploration (Mesquita, 1981). After discounting the urban areas and other facilities, Brazil still has more than 50 percent of its territory that is not being explored economically. Naturally, a large percentage of the unexplored area is not suitable for agricultural activity for a number of reasons not to be discussed here. Mainly because of this availability of land, most of the increase in agricultural production has been achieved through area expansion. This pattern is bound to change in the near 33 34 future for the simple reason that most of the suitable land close to the urban centers has already been brought into exploration. Further occupation of land will be possible only with increased costs in transportation and infrastructure. As can be seen in Table 4.1, the land devoted to agricultural exploration is very irregularly distributed among the regions. Figures 4.1 and 4.2 give an idea of the geographical position of the states and their distribution within the respective regions. Table 4.l.--Brazilian land distribution among the regions and agricul- tural land use. Area $ of1 Agricultural A % . . . ota Use gricultural Region (m""°" ha) Area (million ha) Use (1) (2) (3) (Bl/(l) North 358.1 42.1 29.8 8.4 Northeast 154.9 18.2 79.8 51.7 West-Central 187.9 22.2 93.7 49.8 Southeast 92.5 10.8 72.8 79.3 South 57.8 6.8 46.5 82.8 Total 851.2 100.0 322.6 38.2 Sources: Mesquita (1981) and Fundacao IBGE. The southeastern and southern regions have about 80 percent of their areas already under cultivation, and very little expansion can be expected in the future. The west-central region is the area that can be expanded, at least in the short run. It has the advantage of already having most of the basic infrastructure and of being close to urban centers. But its exploration cannot be accomplished without 35 Noronha (Terri one) . s . _ i _ I- I Bulls/Xx... 7 , I . 3&1 'oIDE JAhERO Fernmdo do MINAS GERAI$ ,Jl uxllu 7.11:.1:v.l ( u , 3 L. .11 L 1 Ci. . . 2. [/7471tuzuv . IT. , .i c I Irv \\..~ 1| .4 I I C , 09:».L,-.-._:,7:.._\.Au_\5.1.1,? .- _\ I i 2 fl1.1.1.35011111/7.ix... T» 1.71 31‘ .z. ,1 I /._.u\l:\:_ . _u\\r .LJCLITQl . in . .2...) , (”Cu ..7_\_\,L:J \rll ; _:_,\.. . sq... ;\_./.I.J1;\,l.»|u.»\\L H.321. “70.1.... ”Six—{Lil .1 1; (EVE-u ,5)», n. . :uu:;__\u \2 . J ., —1 1,, PARANA . I . J11” .rl .ll/1: /\/\\A r \ .V/ I .8953}, .. aid-cervwru l I ll {Ll—l1 _ .s u. I. I/ I so .zsnw;.a .-. 1..- X. 4. G . vr . . _- . .Y... . 1 distribution of the states. ica Geograph Figure 4.l.——Map of Brazil (From Anuario Estatistico do Brazil, Fundacao IBGE, 1980.) 36 Fernando de Noronho (Ten iiério) —-Map of Brazil: Geographical distribution of the regions. Figure 4.2. (From Anuério Estatistico do Brasil, Fundacao IBGE, 1980.) 37 heavy investment in fertilizer and lime to correct soil problems, which are conmon to a large percentage of its area, particularly the Cerrado region. The northern region is the largest in the country, occupying 42.1 percent of the total territory, most of which is cov- ered by the dense vegetation of the Amazon Forest. This region has the disadvantage of being too far from the consumption centers. 0n the other hand, the controversy raised about the devastation of its natural resources, along with problems of soil and climate inadequate for crop cultivation, have prevented a rational exploration of the area. These problems make the region a difficult alternative of fron- tier expansion, at least in the short run. The northeastern region, covering 18.2 percent of the territorial land, is one of the poorest in the country for agricultural exploration. The main problem is related to the unavailability of water. More than half of its area receives only from 250 to 800 mm of rainfall, distributed irregularly during four to six months of the year (Alves, 1981). There is a pos- sibility of expanding agriculture, but only with special irrigation techniques and research in crop varieties well adapted to the dry climate of the region. What emerges is mainly the fact that an increase in agricultu- ral production can still be achieved through an expansion of cultivated area, although with increased production costs. But further growth in production must be attained through an increase in the productivity of agricultural commodities as well. This is the main task of the agri— cultural research coordinated by EMBRAPA, the Brazilian Agriculture Research Corporation; the positive results already obtained by this 38 institution indicate that Brazil has a great potential for further expansion of agricultural production. The agricultural products con- sidered in this study are largely concentrated in the southern, southeastern, and west-central regions. Crop and Meat Production A perfectly competitive market was assumed in this study in both input and output markets. The farmers try to maximize expected profits when making decisions about the amount of land to devote to various crops. To estimate agricultural supply, the Nerlove model, hypothesizing farmers' reactions in terms of price expectations and partial area adjustments, was adopted. Planted area was chosen as an indicator of intended supply instead of quantity produced because it was believed the latter is subject to variations outside the farmers' control, such as weather, pests, and diseases. For a review of argu- ments favoring the approach adopted in this research, i.e., that area shows intended supply better than does quantity, see Nerlove (1958) and Gemmil (1978). The latter showed that this approach is superior in terms of efficiency of forecasting quantity. It was hypothesized that farmers respond to expected crop yield as well as to expected prices (Spriggs, 1981). A linear-trend yield was used to represent expected yield. The explanatory price variable was then represented by a new variable called ”expected return per hectare." This expected return was given by the mul- tiplication of the lagged price by the expected crop yield. The prices from the preceding year were considered as proxies for the 39 expected prices because the announcements of minimum prices during the period under consideration were made too late each year to affect the planting decisions. For this reason and because they had usually been set below prevailing market prices, the minimum prices were not considered as affecting the planting decisions. In the case of wheat, the established support price was used because the govern- ment is the only buyer of the entire crop, and this support price is set in advance of planting time. It is generally believed that this is a more correct way of representing the factors determining planting decisions, and these factors are vitally important to farm—policy decision makers. The expected returns of the major crops that compete for area with the considered crop were included in the hectarage-response func- tion. Another explanatory variable was the lagged-crop hectarage. It was assumed, therefore, that farmers do not react immediately to changes in the factors that affect planting decisions because of the cost of change and inertia or because of the asset-fixity problem. It seemed reasonable to believe that farmers would not change their activities in the short run. All of the four crops considered here occupied large areas in which there were few alternatives; the assump- tion, therefore, was that a large proportion of the hectarage devoted to these crops would be continued over the next year. A linear trend was introduced to account for changes in tech- nology. It was expected that this variable would pick up the influence of omitted variables that were highly correlated with time and then eliminate at least part of the bias to which the coefficient of 40 adjustment derived from the coefficient of the lagged dependent variable is susceptible (Thompson & Schuh, 1977). Four types of equations were used to forecast crop produc— tion: YIELD = F] (TREND) ERETURN = LPRICE * YIELD AREA = F2 (LAREA, ERETURN, TREND) PRODUCTION = AREA * YIELD where the letters L and E at the beginning of the words stand for lagged and expected values, respectively. Obviously, there were expected return variables for the crop under consideration and for the major competing crop(s). However, because of the aggregate nature of the model, only the major competing crops were considered in each case. The use of a lagged dependent variable on the right-hand side of the equations raised the possibility of serial correlation. The use of the Durbin-Watson statistic in this specification is known to be inappropriate. This test is biased toward 2.0 and therefore favors accepting the null hypothesis that there is no serial corre— lation. Maddala (1977, p. 372) stated that the test can be used only when it leads to the rejection of the null hypothesis. An alternative test, the statistic h, developed by Durbin (1970) was used. A great advantage of using this statistic is that it can be computed from the ordinary—least—squares residuals and then tested as a standard normal deviate; thus if h > 1.645, one would reject the hypothesis that there is no serial correlation at the 5 percent level (Johnston, 1972, p.313). 41 In the following sections, the results of the estimated equations for each product are presented and discussed. The values in parentheses are the standard error of the estimates.' Unless otherwise specified, the period covered in the analysis was 1961-1980. Crops like soybeans experienced a very rapid rate of growth during this period, even more significantly after 1970. This unusual growth, together with other reasons, completely changed the broiler industry in the last decade. Hog production also suffered profound modifica- tions in this period, mainly because of the African Swine Fever prob- lem. The development of the mixed-feed industry is closely related to the growth of these two livestock products. But the hog industry is not as efficient as the broiler industry, one reason being its low 4:1 conversion rate compared to 2.5:1 for the broiler industry. Crop Production Corn,--Corn production takes place largely on small- and medium-size farms. Corn is the most widely planted crop in Brazil, accounting for about 25 percent of the total area cropped. Although corn is grown throughout Brazil, production is mainly concentrated in the southern and southeastern regions. There has been a rapid expansion in corn production in recent years. The 44 percent increase in total production in the decade from 1971 to 1980 was achieved through a sig— nificant expansion in area cropped (8.4 percent); yields have increased 'The t-ratio is defined here as the absolute value of the esti— mated coefficient considered, divided by its standard error. The t-ratio is then compared with the student's t-statistic. One asterisk over the parameter indicates that it is statistically significant at the 10 percent level, two asterisks at the 5 percent level, and three asterisks at the 1 percent level. 42 from 1.4 to 1.7 MT/ha. (In the 19605, the average productivity was around 1.3 MT/ha.) The increased yield in this period was a result of the adoption of hybrid varieties and of fertilization in southern Brazil. Despite these low figures, mainly as compared with the United States, Brazil is the world's third largest corn producer, coming after the United States and the People's Republic of China. It appears that Brazil has a potential to become a major force in the world corn trade (Thompson & Garcia, 1978). One of the objectives of the Brazilian Agricultural Research Corporation (EMBRAPA) is to generate yield- increasing technological change, and corn is one of the products studied through the National Center for Corn and Sorghum Research. The impressive yield increases attained by this Center indicate that, in fact, Brazil has the potential for expanding its corn production. In an aggregate view, the principal crop that competes with corn in Brazil is rice. Therefore, the expected return per hectare of rice was expected to have a negative sign. The equations used to forecast corn production are presented below with the R2, Durbin- Watson statistic, and h-statistic from the ordinary-least-squares estimates. The values in parentheses are the standard errors of the estimated coefficients. AC = 842069 + .7443** LAC + 461.24 ERC — 16.15 ERR (.2644) (334.54) (58.95) + 37246.6 TREND (72297.6) R2 = .95 0.11. = 2.41 43 AC = 322337 + .8772*** LAC + 572.34** ERC - 23.69 ERR (.0564) (249.82) (55.78) R = .95 D.W. = 2.49 h = -1.37 YC = 280.59 + l6.00*** TREND (4.39) 2 R = .42 D.W. = 2.05 TPC = AC * YC where: AC = corn hectarage (hectares) LAC = AC lagged one year ERC - expected return of corn per hectare (Cr$l,000/ha) ERR = expected return of rice per hectare (Cr$l,000/ha) TREND = equals 61 for 1961 and increases by one each year YC = yield for corn (kg/ha) TPC = total corn production (MT) In the initial estimation, the area function provided a good fit of the data with high R2 and the expected signs. However, only the lagged dependent variable showed a coefficient significantly different from zero. A possible explanation for these results can be found in the high correlation between LAC and TREND (.96). Although high correlation is only a necessary condition for the presence of multicollinearity, it implies a potential problem. If this is the case, the precision of estimation falls so that it becomes very difficult, if not impossible, to determine the relative influence of the various independent variables (Johnston, 1972, p. 160). Deleting TREND from 44 the area equation improved the results, which appear in the second equation. The R2 remained the same, and all the coefficients had higher t-ratios. The result of the h-statistics computed from the least-squares residuals led to the rejection of the hypothesis that there is serial correlation in the estimated equation. The yield equation is a linear trend, and total production of corn is the product of corn hectarage and yield. Rige.~—Rice is a staple food that typically provides about 25 percent of the caloric and 15 percent of the protein intake for the ab». Brazilian population. The traditional rice-growing areas have been the states of Rio Grande do Sul, Parana, Minas Gerais, Goias, and $50 Paulo. In recent years there has been a rapid expansion of the areas in the West-Central region. The increase in total production in the last decade (30 percent) has been primarily a result of a notable area expansion; yields have remained around 1.45 MT/ha. Irrigated rice cultivation is concentrated in the state of Rio Grande do Sul; the yield has been about 3.8 MT/ha compared to 1.3 MT/ha obtained in the upper-land rice. The ideal would be to have irri- gated and dry-land hectarage and yield equations. Unfortunately, this was not possible because of the unavailability of data. However, only about 10 percent of the area under rice cultivation is irrigated. Total rice production is dominated by dry—land production, so it was hoped that the equilibrium model would not be severely affected by the fact that only aggregate functions were estimated to forecast rice production. 45 At the aggregate level, soybean was considered the major crop competing with rice, and then the expected return of soybean per hectare was included as an explanatory variable. The results of the equations used to forecast rice production are as follows: AR = 4720390 + .5042** LAR + 176.67** ERR - 86.61 ERSB (.2169) (60.32) (134.97) + 92228.5** TREND (36692.1) R2 = .89 0-w. = 2.06 h = -.53 YR = 2206.92 — 10.03*** TREND (3.15) R2 = .36 0-w. = 2.00 TPR = AR * YR where: AR = rice hectarage (hectares) LAR = AR lagged one year ERR = expected return of rice per hectare (Cr$l,000/ha) ERSB = expected return of soybean per hectare (Cr$l,000/ha) TREND = equals 61 for 1961 and increases by one each year YR = yield for rice (kg/ha) TPR = total rice production (MT) The equation for rice hectarage provided a relatively good fit of the data. All signs were as expected, and all of the coefficients with the exception of the one for ERSB had relatively high t—ratios. The h-statistic showed a low value, leading to the conclusion that there was no serial correlation. 45 In the yield equation, a negative sign for trend may appear unacceptable, but in this case it can be explained. Rice production has been diverted from the more fertile soils of the southern states to the Cerrado region. It is usually used as the first crop in the opening of new areas where the less-fertile soils and unstable cli- matic conditions have been negatively influencing the yields. This is compensated for, in part, by the expressive increase in irri— gated rice yields over the time period, although the irrigated area under cultivation has remained almost constant. On the aggregate level, these two opposite effects result in a nearly insignificant decrease in rice yield. Soybeans.--Soybean production increased sharply in the 19605 and 19705. Both area and yield contributed to this phenomenon. The planted area increased fivefold from 1971 to 1980. In the same period, the average yield rose 43 percent, from 1.2 MT/ha to 1.7 MT/ha, allowing total production to increase more than sevenfold. Brazil produced 2.1 million metric tons of soybeans in 1971 and 15.2 million metric tons in 1980. There was, therefore, unforeseen growth in the production, use, and export of soybeans, soymeal, and soyoil in the last decade. Several factors contributed to this dramatic growth in soy- bean production (Williams, 1977). Initially in the South, more pre- cisely in Rio Grande do Sul, the relatively high support price for wheat and subsidies for purchasing machinery and fertilizer increased the wheat hectarage and consequently that of soybeans because of the practice of double-cropping wheat and soybeans in the state. Besides, 47 the same machinery could be used for both wheat and soybeans, and there was no necessity to use fertilizer in the case of soybeans, which depend only on the carryover effects of the fertilizer used in the wheat crop. By the end of the 19605, soybean cultivation spread throughout several other states and continued to expand during the 19705 as a single crop. The factors responsible for this rapid growth were the infrastructure provided by the government, relatively favorable prices, and availability of credit. Also, soybeans are gar— relatively cheap to cultivate because they are nitrogen-fixing legumes; u- this is very important because Brazil produces only a small fraction of its nitrogen fertilizer. At the same time soybean production increased in Brazil, the country also experienced a very rapid growth in the production of broilers. Because soybean meal is the major protein meal used in compound feeds in the broiler industry, this was another factor that contributed to the expansion of soybean production. Brazil is now the second largest producer and crusher of soybeans in the world. It shares with the United States the leading position in terms of exports of soymeal and soyoil. The potential exists for further expansion of soybean produc- tion. This has become more evident because of the positive results achieved by the National Center for Soybean Research in terms of developing varieties that are better adapted for the low latitudes farther north. The West-Central region, where the soil is less fer- tile, recently has shown a tendency to expand soybean production. Although these soils require heavy initial amounts of fertilizers, 48 the yields have been above the country's average. This fact, together with the relatively low cost of land, has helped offset at least part of the increased cost due to the heavier application of fertilizer and to transportation because the West—Central states are farther from export ports and crushing facilities. The expansion of area devoted to soybean production has come about not only through the opening of new land, but also at the expense of other crops and pasture. For this aggregate model, wheat and pasture were considered the major competitors for area. It is common to see studies in which price of wheat is expected to be posi- tively related to the supply of soybeans because of the possibility of double-cropping between soybeans and wheat. In this study, there were two main reasons to expect a negative sign. First, as pointed out above, double-cropping with wheat was more common when soybeans got their start in the South. Although no figures have been reported, it is known that the double-cropping system has declined relative to total production in Rio Grande do Sul. In the rest of the country, soybean is grown as a single crop. Second, agricultural researchers common believe that a significant negative interaction may exist between double-cropping and soybean yield. Because of weather con- ditions, frequent delays in wheat harvesting and consequently in soybean planting can reduce soybean yield up to 25 percent (Thompson, 1979). The price of beef was chosen as a proxy for the return of the product of pastureland. In Brazil, practically all cattle are grass 49 fed, and an increase in the price of beef would lead farmers to increase the number of animals and consequently the area devoted to pasture (Williams, 1977). For this reason, the price paid to farmers for beef cattle was expected to be negatively related to soybean hectarage. errors in ASB = ASB YSB ‘ TPSB = where: ll ASB LASB = ERSB ERW ‘ LPB The estimated equations are presented below (standard parentheses). -205625 + .9689*** LASB + 409.059*** ERSB (.0380) (81.431) - 363.596*** ERW - 25.414 LPB + 10394.8 TREND (107.220) (19.703) (21458.9) R2 = .99 0-w. = 1.80 41962.2-+.9997*** LASB + 427.635*** ERSB (.0121) (44.890) — 276. 294*** ERW - 29.167** LPB + 550451*** 0 (65. 284) (11.891) (117978) R2 = .99 0-w. = 1.67 h = .74 72 -1125.33 + 34.1579*** TREND (7.5200) R2 = .53 0.0. = 1.32 ASB * YSB soybean hectarage (hectares) ASB lagged one year expected return of soybeans per hectare (Cr$l,000/ha) expected return of wheat per hectare (Cr$l,000/ha) price paid to farmers for beef cattle lagged one year (Cr$/MT) 50 TREND = equals 61 for 1961 and increases by one each year D72 = dummy variable equals 1 for 1972, otherwise zero YSB = yield for soybeans (kg/ha) TPSB = total soybean production (MT) In the initial estimation, all the variables had the expected signs, and the coefficient of determination was very high (.99). The LPB variable had a coefficient greater than the standard errors, and the coefficient of the TREND variable was not significantly different from zero. All other variables had relatively high t-ratios. As in the case of corn, the TREND and LASB variables were highly correlated (.94). For this reason and because of the low t-ratio, TREND was deleted. There was controversial information about the observed data for 1972. The data used in the analysis were based on the USDA revised calculations. However, the new—soybean-area information was about 30 percent higher than the older data, and for this reason a dummy variable for that year was added to the final equation. The resulting equation showed the same goodness of fit and had increased t—ratios. The low value for the h-statistics indicated the absence of serial correlation. Supplies of soymeal and soyoil.—-The supplies of soymeal and soyoil are directly related to the domestic crush of soybeans. The rates of extraction varied during the study period because of the adoption of modern technologies. Then the observed yields were used. In recent years, however, yields became stabilized at about .776 for meal and .185 for oil. This means that for every metric ton of 51 soybeans crushed, 776 kilos of meal and 185 kilos of oil are produced. The remainder is moisture loss. The supplies of soymeal and soyoil are then expressed as: TPSM = YSM * DSB TPSO = YSO * DSB where: TPSM = total production of soymeal (MT) TPSO = total production of soyoil (MT) YSM = extraction rate for soymeal (percent) YSO = extraction rate for soyoil (percent) DSB = domestic crush demand for soybeans (MT) Nhgat,--Brazil imports about two—thirds of its total wheat con- sumption. A long-term objective of the government has been self- sufficiency in wheat production. To help achieve this goal, the government tried to stimulate production through making credit avail- able at substantially negative real rates of interest for acquisition of machinery and other inputs such as fertilizer and seeds, and by announcing price supports before planting time, always above world— market levels. The real differences between domestic and inter- national prices, because of persistent overvaluation of the exchange rate, are really smaller than they might appear. The Marketing Department for National Wheat (CITRIN) is the sole buyer for domestic wheat and the sole importer. CITRIN buys the entire crop at the established support price and sells the imported and domestic wheat to the mills at a uniform price that is higher than 52 the import price but lower than the support price. This is done in such a way that revenues and costs are equalized, eliminating the necessity of direct government resources to cover possible differences. Wheat production increased during the 19705, reaching a peak in 1976 (3.2 million metric tons). The wheat yield has been very low, averaging only .85 MT/ha in the last decade. The excellent prices for soybeans have contributed to the expansion of wheat production. It seems that soybean production is now leading wheat production rather than the contrary, as occurred in the 19605 (Williams, 1977). Soybeans are considered a complementary crop to wheat because of the double—dropping system. The expected return of soybeans per hectare is expected to have a positive sign. The production of wheat is concentrated in the South, and most of the area traditionally devoted to its cultivation can be used on a double-cropping basis with soybeans because of climatic conditions. Wheat is planted in May or June, cultivated during the Brazilian winter, and harvest begins in September. Soybean is planted from October to December, grown during the Brazilian summer, and the harvest begins in March. The expected return of wheat per hectare is expressed as the product of wheat support price (the only price) and the expected yield. The results of the estimated equations are pre- sented below. AW = —6389230 + .5006** LAW + 228.075 ERW (.2070) (231.501) + 152.427 ERSB + 86329.5** TREND (143.659) (40214.5) R = .91 D.W. = 2.32 h = -1.91 53 AW = -6110110 + .5795*** LAW + 220.225 ERW (.1655) (175.192) + 136. 256 ERSB + 81527. 5** TREND (114. 982) (34059. 4) R2 = .96 o.w. = 2.07 p = —.40* (.21) YN = 299.471 + 7.46** TREND (3.76) R2 = .49 D.W. = 2.66 TPW = AW * YW where: AW = wheat hectarage (hectares) LAW = AW lagged one year ERW = expected return of wheat per hectare defined as wheat support price * expected yield (Cr$l,000/ha) ERSB = expected return of soybeans per hectare (Cr$l,000/ha) TREND - equals 61 for 1961 and increases by one per year YW yield for wheat (kg/ha) TPW total wheat production (MT) As discussed before, the use of a lagged dependent variable as a regressor raises the possibility of serial correlation. The first equation presents an estimate for the h-statistic that is high enough to detect the presence of serial correlation. That equation was re-estimated using the Hildreth-Lu estimation procedure because it is the appropriate method for equations with lagged dependent vari- ables (Maddala, 1977, p. 282). The new results appear in the following 54 equation. The estimate of p was -.40 with a t-ratio of 1.9, indicat- ing that the value of this parameter was significantly different from zero. According to the Rao and Griliches rule of thumb, there is an indication of some gain in efficiency by using estimation proce- dures that take serial correlation into account when the absolute value for the estimate of p is greater than .3 (Maddala, 1977, p. 283). This study assumed that the true residuals were first-order auto- regressive. The R2 increased from .91 to .96 in the second equation, and all of the estimated coefficients were more robust than in the first one. Supply elasticities of cerealggrains.--The results presented in Table 4.2 show the hectarage response elasticities derived directly from the estimated crop hectarage equations and evaluated at the average expected return and hectarage values. Not only changes in prices but also changes in yields were taken into consideration. These elasticity estimates must be interpreted with caution because the underlying ceteris paribus conditions usually do not hold for simultaneous systems of equations. Although no direct comparisons among this and other studies can be made because of the approach adopted here, in which farmers were considered to respond to expected crop yield as well as to expected price, some confidence can be placed in these elasticity estimates because they were found to have about the same order of magnitude as those expected and empirically obtained in other studies .1. 55 (Williams, 1977; Vilas, 1975; Thompson & Schuh, 1977), using the tra- ditional approach. Table 4.2.--Elasticities of hectarage response.a Expected Return Per Hectare Crop _ D Corn R1ce Soybeans Wheat Beef Cattle Corn .1247 -.0136 Rice .. .2076 -.0683 .. Soybeans .. .. .5267 -.2696 -.1428 Wheat .. .. .2767 .3599 aThe formula for the elasticity computed at the means is %%-. %:, where A represents area (hectarage) and R the relevant expecte return per hectare as defined in the text. bIn the case of beef cattle, price received by farmers was used instead of return per hectare. Soybean supply (hectarage) was found to be more responsive than the other crops. Although the direct elasticity of corn was rather inelastic, it still implied that price policy and research efforts could contribute to expanding its production. In terms of cross-price elasticities, only the results in the soybean and wheat equations seemed to be relevant. The cross-price effects between soybeans and wheat had opposite directions, but both had a significant effect on the hectarage equations. A change in the price of beef cattle implied some effect on the soybean hectarage through the changes in the area devoted to grazing. 56 Meat Production The livestock subsector was included in the model because of its interactions with the feed—grain market. Although beef cattle are almost exclusively grass fed, they were included in the model because the interrelationships between beef and other meats in the consumer-demand side influence the demand for corn and soymeal as feed input, and because of the competition for area between pasture (grazing) and crops, especially soybeans. The economic model used to represent the structure of the meat sector was based on the assumptions that producers, as in the case of crops, try to maximize their profits when deciding the amount to be produced; that their product is considered to be homogeneous; and that individually they do not influence product or input prices—- that is, perfect competition is assumed. Beef production.--The cattle industry presents a particular characteristic when compared with other enterprises. Its product is, at the same time, both an investment and a consumption good. Ideally, one should have a set of equations including investment and slaughter functions for both males and females. However, this refinement was limited by the availability of data. In Brazil, the official data on the stock of cattle on farms are not considered to be reliable. Because of these data problems, a single equation was postulated to explain the production of beef in Brazil. For this purpose, annual data of cattle slaughtered under federal inspection were used. 57 As in the case of crops, it was assumed that, in the short run, a beef-cattle farmer was likely to remain in the same enterprise, adjusting the size of his herd in response to changes in economic con- ditions. In this way, beef production is determined by the previous year's output as well as price. Therefore, it was assumed that the beef-cattle farmer would move in the direction of the optimum by an amount proportional to the difference between the optimum and the present position. The model should also specify a period long enough to account for the time that elapses between the moment a farmer decides to increase production and the moment the product is ready to go to the market. In the case of beef production in Brazil, this period was believed to be about four years. According to Nerlove (1956), farmers react not only to the current price but also to the price they expect to receive in the future. He suggested that the expectation concerning future prices can be formulated based on past prices. Therefore, in the supply of beef there is a delayed adjustment to changes in price. Following Coirolo (1979), a polynomial-lag model was chosen to estimate the supply function. In this distributed-lag model, the parameters of the lagged variable are allowed first to increase and then to decrease and vice versa. These a priori restrictions on the parameters reduce the problem with degrees of freedom. Another advantage of this technique is that it deals indirectly with the problem of multicollinearity, which cer- tainly would be present if the function were estimated as in an 58 ordinary multiple—regression equation because the lagged variables would probably be highly correlated (Kmenta, 1971, p. 473). Several lagged structures in terms of degree of polynomial, lag length, and constrained or unconstrained lag parameters were attempted. The best equation was achieved with a second-degree polynomial approximation of the farmer's beef price lagged for five years, with the last lagged parameter constrained to be zero. One additional variable should be the price of feed (pasture), but this was not possible because of the difficulty of constructing such a variable representing the opportunity cost of land for pasture. The results for the estimated equation are given in Table 4.3. Table 4. 3. --Parameterestimates for constrained second- -degree polynomial lag model. Dependent variable. PB Independent Estimated Standard _ . Variable Coefficient Error t Rat‘° Constant 21207.2 138934 .15 TPB (t-l) .7660*** .0721 10.63 PB (t -17.04*** 4.82 -3 53 PB (t-l) 2.61 2.89 .90 PB (t-2) 14 14*** 3.01 4.70 PB (t-3) 17.55*** 3.08 5.69 PB (t—4) 12.83*** 2.15 5.96 R2 .95 N 20 D W 1.52 h 1.13 where TPB = total production of beef (MT e.c.w.) PB = price of beef cattle received by farmers (Cr$/MT e.c.w.) t = current year e c.w. = equivalent carcass weight 59 In evaluating the results of this equation, it should be noted that the statistical significance of the individual coefficients was relatively good. The R2 was quite high for this equation, and all of the coefficients had the expected signs. The value of h led to the rejection of the hypothesis that there was serial correlation in the equation. Based on these results, it can be concluded that the value of the total slaughtered beef at any given time depends on a weighted sum of the present and past values of the independent variables. The effect of beef prices on the total slaughtered animals is negative in the short run. Given an increase in price, the beef- cattle producer expecting the price to increase more will withhold animals from slaughter in order to increase future production. In the long run, the opposite is expected; that is, the effect of a change in price is positive-—as the stock of animals increases, the number of animals ready to be slaughtered becomes higher than previous levels. The elasticities presented in Table 4L4illustrate these results. The short-run price elasticity of slaughter was -.15, and the long-run elasticity was .23. Three to four years would be necessary to reach again the level of slaughter supply occurring at the moment of the price increase. To gain some confidence, it is interesting to compare the single and cumulative elasticities of slaughter derived in this study with those obtained in other studies. (See Table 4.5-) Although none of these studies used exactly the same approach or considered the same period, the values of the elasticities were of the same order of magnitude. 60 Table 4.4.—-Simpleandcumulative price elasticities of beef slaughter. Time Period Simple Price Cumulative Price E1aSthlty E1ast1c1ty t -.15 -.15 t-1 .02 -.13 t-Z .12 -.01 t-3 .14 .13 t'4 -10 .23 Table 4.5.--Comparisonpfcumulative elasticities of beef supply in three stud1es. Time Period (T025113 %§§§;3 %¥:%:?Kc t -.15 -.11 -.26 t-1 -.13 -.04 -.30 t-2 -.01 .04 -.19 t-3 .13 .11 .00 t-4 .23 .17 .19 aPresent study, period: 1961-1980. bLattimore (1974), period: 1941-1971. CCoirolo (1979), period: 1956-1975. Hog production.--0nly recently has the swine industry started to expand, as a result of improved management techniques and the use of specialized breeds. The supply of hog meat can be derived in a manner similar to that used for the supply of beef. The assumption thatatleast part of the live animals may be kept on farms as inventory carryoverswhen 61 producers expect higher future prices can also be applied here. However, the length of time within the weight range at which the slaughtered hogs are marketed is shorter than in the situation with cattle (Myers & Havlicek, 1975). In some cases it is possible to have two pig crops per year, and some pigs born in one year are not slaughtered until the next. Ideally, empirical analysis of the hog-meat supply should focus on a time period shorter than one year-- a quarter or at least a semester--to capture the changes that might occur within a year. The use of annual observations in a more sophis— ticated approach could lead to erroneous conclusions. However, economic data for hogs in Brazil for the study period were available only on a yearly basis, and, as in the case of cattle, the only data considered reliable were the observations of hogs slaughtered under federal inspection. Because of these data problems, it was decided to specify the hog-meat supply as a simple function in which the total meat supplied was a function of the hog price received by farmers lagged one year and the price of the principal input, corn in this case, also lagged one year. A negative sign for the price of hogs was obtained in several specifications where the hog price appeared as an independent explana- tory variable, and because no apparent reason was found for that, it was decided to work with a ratio of the two prices instead of two independent variables. A great number of animals were eliminated, mainly during the 1971-75 period, because of the problem with the African Swine Fever. In the absence of data reporting the number 62 of animals discharged, a dummy variable was added to the explanatory variables. The results are presented below. TPH = 15437.1 + .9416*** LTP + 5164.83 LPH/CORN (.0927) (3421.24) - 83244.8*** 07 (16394.0) ‘75 R2 = .87 N = 20 0.w. = 2.09 h = -.24 where: TPH = total production of hog meat (MT e.c.w.) LTPH = TPH lagged one year LPH/CORN = ratio between the hog price received by farmers (Cr$/MT e.c.w.) and the price of corn (Cr$/MT), lagged one year D7175 = dummy variable equals 1 for years 197—75, otherwise zero The R2 seemed reasonable for this kind of specification. All of the coefficients had the correct sign and were greater than the respective standard error. The variable that expressed the ratio between the prices of hogs and corn was the least robust, with a t-ratio of 1.5. The coefficient of the dummy variable that was used in an attempt to take care of the discharge of animals because of the African Swine Fever appeared to indicate that the supply of hog meat in the period 1971-75 was significantly less than in the other years. The h-statistic was small enough to reject the null hypothesis con— cerning the existence of serial correlation. Although the "price“ elasticity of hog slaughter (.0849) was rather inelastic, it should be remembered that it reflected the change 63 in supply of hog meat to changes in variables with effects operating in opposite directions. Poultry production.--The rapid expansion of the poultry-meat industry in Brazil occurred in the last decade and resulted mainly from the use of modern, intensive systems of production based on new imported technologies. In 1961, Brazil produced 7,700 metric tons (MT) of poultry meat. The 1980 production was about 864,000 MT. This should be viewed as an unusual circumstance and not as indicative of a new trend. Another factor that contributed to the rapid transformation of the poultry sector was the expansion of the mixed—feed industry. Corn and soymeal are used extensively in the production of feed rations. Improvement in the quality of these rations has helped the new poultry breeds to convert feed more efficiently. In the case of broilers, for example, conversion is about 2.5:1 (Nogueira & Criscuolo, 1979). This indicates the high-technology level of the Brazilian poultry industry. This rapid expansion of the poultry sector allowed the country to begin exporting poultry meat in 1975 (3,500 MT). Including the Middle East among the importing countries, in addition to the factors described above, made it possible for Brazil to export about 169,000 MT of poultry meat in 1980 and approximately 275,000 MT in 1981. It is difficult to model a sector with such rapid techno- logical change. Also, the same data problems as those described in the section on hog production existed here. In addition to the difficulties described above, real prices for meat declined more throughout the study period than did the cost 64 of feed, mainly soymeal--the principal component of the rations used in the poultry industry. Regressions based on these variables always resulted in wrong signs, notably negative coefficients for the real prices of chicken received by the farmers. Several attempts were made to obtain theoretically reasonable estimates but did not improve the results. Examples of alternative forms are nominal rather than real prices and supply and demand based on two—stage least—squares estimation. The choice of the final model used here was based on the study by Peterson (1981). According to concepts drawn from the theory of the firm, this writer concluded it was reasonable to assume that relatively new sectors, like the poultry industry in Brazil, are still moving toward equilibrium. Over the historical period, this sector has had declining profits, but the writer believes it is still positive. Only in the long run would equilibrium be reached, with economic profits for all industries equalized at zero. It was hypothesized that the poultry sector will continue to increase as long as profits are greater than those of the hog industry, which is considered the alternative investment possibility because of the vertical integration of those sectors with the feed industry. Ideally, one should construct return measures based on the gross margin for poultry relative to the gross margin for pork. Because of data unavailability, a ratio between the prices received for chickens and hogs was used as a proxy for the relative return measure. Therefore, the total-meat—production (MT of slaughtered 65 animals) equation has as arguments the chicken-hog price ratio; the price of soymeal, considered the main element in the compounded feed ration; and a trend to represent the technological change of the sector. Only observations beginning in 1970 were used in the estima— tion of this equation because of the dramatic change Brazil experi— enced in this sector during the last decade. The results are presented below. TPCH = -5033320 + 193417 PCH/HOG - 111.236** WPSM (162214) (44.793) + 73813.1*** TREND (6928.5) R = .97 D.W. = 1.33 where: TPCH total production of poultry meat (MT e.c.w.) PCH/HOG = ratio between the price of chicken received by farmers (Cr$/MT e.c.w.) and the price of hogs received by farmers (CrS/MT e.c.w.) WPSM - wholesale price of soymeal (Cr$/MT) TREND = equals 70 for 1970 and increases by one each year The explanatory variables seemed to explain relatively well the variation in total meat production. All of the coefficients were greater than the respective standard errors and had the correct signs. The Durbin-Watson test was inconclusive regarding the existence of serial correlation. The estimation of the same equation using the Cochrane—Orcutt procedure did not improve the results. In fact, the residuals did not show evidence of serial correlation. 66 Although the equation used in this model seemed to account very well for the rapid growth of the broiler industry, a more realis- tic formulation is needed to take care of the long-run equilibrium. Validation To validate the model, the period 1971-80 was simulated, based on the estimated equations, and compared with the actual values of the variables. To make the model self—corrective at each itera- tion, actual lagged values were used rather than those estimated by the equations. The results related to the area and total-production equations are presented in Figures 4.3 to 4.6. The accuracies of forecasts are given by Theil's U—statistics presented in Tables 4.6 and 4.7, which include the correlation coefficients for actual levels (R1). This measure is based on the regression of actual and pre- dicted values (Y: = d + BYE) presented in Chapter III. As discussed before, UM, UR, and UD should add up to one, since they are error proportions of the mean square error, but they may not do so because of rounding. Actual yields and predicted acreages were used to forecast total production. The percentage errors between the actual and simulated values of area and total production were generally small. As can be seen in the figures, most of the turning points were caught. Following the criteria that UM and UR have to be close to zero and UD close to one for a perfect forecast, it can be said that the equations' performance was satisfactory. The U-statistics suggest that the rice equations were somewhat weaker than the others as 67 AC AR ....... + - - - _ - - .. - - + - - - - 11.722 - - 6.572 - ‘ 11.573 - * - 6.402 - ' 11.423 - /; 5.232 - z - I 11.273 - t * /; - 6.062 : + 1.10123 ' 9' ’ 50892 ' * ' c Q — - * * 100973 "' */ - 5.722 - g .- 10.823 - + - 5.552 - fl\ ' 10.673 - // - 5.382 - + - - * + - _ T - 10.523 - +-+ - 5.213 - ' 100373 " * ' 50043 ' - - - - 1k ‘k - 10.223 - - 4.873 - * ' _ _ .. + — 10.073 - * - 4.703 - +\/\+ -_- 9.924 - + - 4.533 - + - ’o'o'o'o-o'o'o'o‘o'o 'o'l'o‘o‘o'o’o'o'o’o 71 73 75 77 79 71 73 75 77 79 AH ASB - -------- + - ---------- t 3.748 - - 8.602 ' /; 3.584 : : 8.028 : /+ : 3.420 - - 7.4.54 .. / - .- ‘k i - — 3.256 - ' 6.880 - + - 30092 " t + E 60307 _ /+/ _ 2-928 - + \ - 5.733 - + .- - 5 - — t - 2.764 - + - 5.159 - / - - D - + - 2.600 - * - 4.585 - - 2.€36 - - 4.012 - - 20212 ’ 4‘ " 30Q38 ' + ' 2.108 - / - 2.5-64 - / - - i c - fi .- 1.944 - + w - 2.290 - / - 10780 " * '2 ’ 10716 ‘ '2 " 'o’o’o'o'o‘o'o’o‘o'o 'o'o'o'o-o‘o"o"o'o'o 73 75 77 79 71 73 75 77 79 Figure 4.3.—~Actual (+) and simulated (*) values for area: corn (AC), rice (AR), wheat (AW), and soybeans (ASB), in million ha, 1971-80. TPR 9.628 9.369 9.111 8.853 8.594 8.336 8.077 7.819 7.561 7.302 7.044 6.785 6.527 Figure 4.4.—-Actual 68 TPC 20.115 - ----- 19.557 - 18.999 - + 18.441 - 17.883 - * - + 170324 ' 16.766 - 16.208 - +-+ 15.650 - 15.092 - * 14.533 - //\\ 13.975 - + 13.417 - * 71 73 '75 77 TPW —----§—-—- - ——————— t-- r 3.135 - + - 20943 ' i + g - — + - * 20751 ‘ //\v If - - + § ' 20559 ' - i - - * - ' 20367 ' - - _ * - : 20175 ' ' * : 10983 ' * * ' ’ 1.791 - - 1- - -* + - * + ‘ 10599 ' - *‘k U - c- n/ ' 10907 ' - + - — — \\: - 1.215 - - // - 1.023 - - + + - - - ¢ ' 0831 ' ‘ 'o’o'o'o'o'c'o'o'o'o 'o‘o-o‘o'o'o'o'o‘o-o 71 73 75 77 79 71 73 75 77 79 (TPC), rice (+) and simulated (*) values for production: corn (TPR), and wheat (TPW), in million MT, 1971-80. 69 14.820 13.758 12.696 11.634 10.572 9.510 8.449 - // P+§ 7.387 6.325 5.263 4.201 3.139 2.077 TPSM TPSO ---------- T l——-------— 9.133 - - 2.232 - 80452 ' ' 2.066 ' 7.770 - * - 1.901 - 7.089 - - 1.735 - 6.408 : */»—+ I 1.570 : +/+--r 5.727 - * . - 1.404 - 4 4 50045 ' # ’ 10239 - 3/ 4.364 - fi/ - 1.073 - 3/ 3.683 - // - .908 - /; 30001 ' 7 - .743 ' * 2.320 - // - .577 - // - - - + 1.639 - /» - .412 - +[/4 .957 - r/A - .246 - v/ 'o'o'o'o'o'o-o‘o'c'o '0'0'0'0'0‘ ' '0'.“ 71 73 75 77 79 71 73 75 77 Figure 4.5.—-Actual (+) and simulated (*) values for production: soybeans (TPSB), soymeal (TPSM), and soyoil (TPSO), in mi1110n MT, 1971-80. 70 TPB - - u u - - - - * - - 2.445 - + _ 2.376 - \\ - - * - 2.308 - I.- - 2.239 - - - * o 2.171 - + - 2.102 - +\: ‘ + 2.033 - 4 - - i 1.965 - - 10896 " - 10828 ' '9'—+ - .. \ 3, _ 10759 ‘ * + t + _ - g, - 1.690 - fi\// - 10622 " + - ‘0'. ’ '0' ’0' " -. 71 73 75 77 79 TPH TPCH - - - .. - - - - _ - T - ———————— - .Q. 696.899 ' * - 833.573 - f .- + - - 680.127 - 773.547 - 4 663.356 - - 713.522 - - - — _ «p - 646.584 - - 653.496 - + - — t - — 629.812 - + " 593.470 - */ - - - - 4. - 613.041 - - 533.445 - / - .. - .. * - 596.269 - - 473.419 - * + - 579.497 - + - 413,393 - // - - - - * + - 562.726 - \ 4 1 - 353.368 - +/ - -. * - - - 545.954 - * - 293.342 - // - - + + - - + - 529.182 - * *\ - 233.316 .. / - 5120411 - " 1730291 - */+ .. - 'k - - /+ i - 4950639 ' ' 1130265 ' + .- ‘o'o'o'o'o'o'o‘c‘o'o '0'0'0'0'0'0‘0'0‘0'0 71 73 75 77 79 71 73 75 77 79 Figure 4.6 --Actua1 (+) and simulated (*) values for meat production: beef (TPS), in million MT, andhogs (TPS) and ch1ckens (TPCH) in thousand MT, 1971-80. 71 forecasters. This is probably due to decreasing expected yields, which may have affected the decision to plant, represented by the expected revenue variable in the acreage equation. On the other hand, the wheat equations led to relatively large forecasting errors in cer- tain years. One explanation for that could be the omission of a vari- able to represent credit in the acreage equation, since its availability is crucial to the planting decisions. The figures show that the soy- bean equations were the best forecasters. Table 4.6.--U-statistics for area equations. Equation U1 U2 UM UR UD R1 Corn .0171 .7864 .0002 .1185 .8813 .73 Rice .0316 .5462 .0097 .1308 .8595 .88 Wheat .0596 .5696 .0077 .0559 .9364 .84 Soybeans .0080 .1164 .0239 .0877 .8884 .99 Table 4.7.--U-statistics for total-production equations. Equation U1 U2 UM UR UD R1 Corn .0164 .1988 .0000 .0144 .9856 .97 Rice .0319 .4126 .0069 .2110 .7820 .92 Wheat .0600 .3101 .0002 .1564 .8434 .91 Soybeans .0076 .0598 .0304 .0790 .8906 .99 Soymeal .0422 .3694 .0008 .0048 .9945 .98 Soyoil .0422 .3547 .0017 .0031 .9951 .98 Beef .0247 .5133 .0028 .0795 .9177 .93 Hogs .0245 .5286 .0280 .0154 .9566 .91 Chickens .0352 .3573 .0043 .0998 .8958 .99 413 CHAPTER V DEMAND Introduction Consumer-demand theory is, in fact, an allocation problem. The consumers are concerned with the amount of their incomes to be allocated to each particular good. Each time they make a decision, it must be based on the knowledge of the total amount of money avail- able for spending and the prices of all goods. Therefore, consumer demand should not be analyzed on the basis of individual commodities, but as part of a complete system of demand equations. Recently, an increasing number of researchers have dealt with problems of specifying and estimating such systems (see Barten, 1968, 1977; Theil, 1975, 1976, 1978; Brown & Deaton, 1972). Working with systems of demand equations has the advantage of recognizing the interdependency of demand for the various goods while using the basic concepts of the utility function and budget constraint. The results are usually pre- sented as a set of demand equations, which show demand as a function of income and price. Without misplacing the importance of the relationship between the quantity demanded of all commodities within the budget, this study did not focus on a complete system of demand equa- tions. One of the reasons for this decision was the unavailability of data for all commodities needed for this particular procedure. 72 73 Another was the fact that even if one has the proper data, it is almost always necessary to group the commodities to cope with the problem of degrees of freedom, and this researcher was interested in having coefficients for specific commodities and not for a group of them. Individual demand equations were estimated for several com- modities, based on the consumer-demand theory. This theory explains demand in terms of the maximization of consumer utility subject to an appropriate budgetary constraint. As Labys (1973, p. 9) pointed out, sev- eral important considerations surround demand equations. Among them are a set of assumptions and conditions under which the behavioral relationship between quantity demanded and prices must hold. These points have been fully explored in many studies (Theil, 1978; Barten, 1977; Labys, 1973) and are not repeated here. Fulfillment of these conditions and assumptions was assumed in this study. For example, it was assumed that all consumers face an equivalent price, that individual responses to income changes can be approximated by the average quantity response to income changes among consumers, and that utility maximization is adopted by the consumers subject to their income or budgetary constraints. To simplify the presentation and discussion of the several equations, they were categorized in three different groups: one for consumer demand (soyoil, meats, rice, and wheat), one for crush demand (soybeans), and another for feed use (soymeal and corn). The consumer-demand theory suggests that the quantity demanded of a good depends on the price of the good, the prices of all other ..fi-.:. 74 related goods, and consumers‘ income. In the case of retail demands considered in this study, it was not feasible, according to the pre- ceding discussion, to interact with all of the considered products in each equation. For the purpose of this study, only the most important food items were considered explicitly as connected products. It was assumed that a doubling of price and income variables would have no apparent effect on consumption. Thus, postulating the absence of money illusion, prices and incomes were deflated by their respective price indexes. The consumer-demand theory indicates a negative relationship between the quantity demanded of a particular food item and its price. The prices of the products considered as substitutes in consumption are directly related to the quantity demanded of the considered product. A similar direct relationship holds for income and popula— tion variables. Initially, it was intended to use per-capita dispos- able income and population as distinct explanatory variables and to express the dependent variable as total (aggregate) consumption. Because of the high correlation between per-capita income and popula— tion, per—capita consumption was used as the dependent variable in all retail-demand equations, eliminating the population variable from the right-hand side among the independent variables. A major problem in using this approach is that no data series exist for the volume of consumption for most of the agricultural products in Brazil. For this reason, series of apparent consumption had to be constructed for all the products studied. In general, 75 apparent consumption was calculated by subtracting from total pro- duction the uses as seed and industrial processing, net trade, and changes in stocks. Each case is considered in detail in a separate section of this chapter. Unfortunately, only data for stocks of soybeans and products were available. The calculated consump- tions of the other items diverge from their true values to the degree that the assumption of no carryover stock is improper. This is crucial, mainly in the case of rice. Another problem in using this approach is that exports, in general, are reported on a calendar-year basis for January through December, but most of the products are harvested by the end 0f the first quarter in the central and southern regions of Brazil. This means, in some cases, that exports realized at the beginning of the year are derived mainly from the previous year's production. These factors could prejudice the calculation of apparent consumption described above. The results of the estimation of each demand equation are given in the following sections. Twenty observations covering the period 1961-80 were used in most of the cases, with the exception of broilers and pork, for which the 11 most recent observations were considered. The estimated standard error of each coefficient appears in parentheses beneath each of the coefficient values. 76 Consumer Demand M In the last two decades, as a result of rapid urbanization and increase in per—capita income, demand has shifted from lard to vegetable-origin oils, and per-capita consumption of edible oils has risen. At the same time that the production of soyoil increased dramatically, the production of both peanut and cotton oil declined in Brazil. As a result, soyoil now accounts for about 90 percenttrfthe total domestic consumption of edible oils. In 1980, domestic demand for soyoil was about 1.5 million metric tons. The demand for soybean oil was initially postulated to be a function of its own price, the price of a close substitute (lard), and real income. No satisfactory results were obtained with this fonnulation, mainly indicating an erroneous sign for the price of soyoil. To help explain the rapid expansion of consumption of soy- oil in Brazil in recent years, variables such as lagged consumption and trend were tried, but because both were highly correlated with income, these formulations could not be carried on. The results of a final version of the equation are reported below, where availability of soyoil, which includes total production plus stocks at the beginning of the year, was included as one of the explanatory variables. It is believed that in the case of soyoil in Brazil, the dramatic increase in per-capita consumption was not only a result of urbanization and growth in per-capita income, but was mainly due to 77 availability as a consequence of the rapid growth of the soybean sector. 050 = -1.415 - .0614 P50 + .0524 PLARD + .0032*** AVSO (.0657) (.0523) (.0006) + .2558** INCOME .1147) ( 2 _ R — .99 D.W. = 1.92 where: DSO = per-capita domestic demand for soyoil (kg) PSO = retail price of soyoil (Cr$/kg) PLARD = retail price of pork lard (Cr$/kg) AVSO availability of soyoil (MT) INCOME = disposable income per capita (Cr$/inhabitant) The explanatory variables had the expected signs. The less- significant coefficient was the price of soyoil. Its estimate had about the same size of standard error as reported below each of the coeffi- cient values. The goodness of fit in the preceding equation can be attributed mostly to the income and availability variables. As expected, changes in per-capita consumption in the period considered can be attributed to changes in these two variables. The own-price elasticity was -.26 at the observation means. The estimated cross—price elasticity of soyoil with respect to a change in the price of lard was .27 in this study. A previous study by Williams (1977) reported an own—price elasticity of -.15. However, no strict comparison could be made between these two results because they were based on different equation formulations and because the 78 latter was based on an even less significant coefficient. The income elasticity of soyoil was .8352. Ripe In Brazil, rice is a staple food that usually provides about 25 percent of the caloric and 15 percent of the protein intake of the population. Because no data exist for the volume of rice consumed in the country, the apparent consumption was calculated based on an average milling rate of 68 percent. The following formula was used: Disappearance = .68 (total production - seed [76 kg/ha] - waste [10 percent of total production]) - exports + imports. The absence of data related to the stocks of rice affects greatly the apparent- consumption measure. It is believed that significant amounts of rice have been held by processors, producers, and wholesalers in addition to the government in most recent years, but accounting systems seldom report inventory data. Per-capita consumption is more or less constant in most regions of Brazil, with the exception of the Northeast, where per-capita con— sumption is about half of the average for the rest of the country. The major reason for this is the fact that the northeastern region does not produce rice, and transportation costs to ship the product from the southern and central regions arevery high, considerably increasing the price to the consumers. In this region, manioc flour is the main substitute for rice, whereas wheat flour is the natural substitute for the rest of the country. Bearing in mind the data limitations, the demand for rice was postulated to be a function of its own price, per-capita income, and 79 the price of two possible substitutes--wheat flour and manioc flour. The statistical results of the estimated equation were as follows: DR = 36.1603 — 1.9966 RPR + .8866 RPWFL + 3.6019** RPMFL (1.1977) (2.0071) (1.4220) -(:92;2)INCOME R2 = .43 0.11. = 2.15 where: OR = per capita apparent consumption of rice (kg) RPR = retail price of rice (Cr$/kg) RPWFL = retail price of wheat flour (Cr$/kg) RPMFL = retail price of manioc flour (Cr$/kg) INCOME = disposable income (Cr$/inhabitant) The multiple-correlation level was considered very low, but this equation was kept because it was the best that could be obtained with the information available. The price variables presented the expected signs, and the results showed that manioc flour can be con- sidered a relevant rice substitute. The coefficient for the wheat— flour price had a low t-ratio; this meant that the coefficient was not statistically different from zero. A possible explanation for this unacceptable result could be related either to the data problem dis- cussed before or to the fact that wheat consumption during the sample period was heavily subsidized, disguising the true effect of wheat— flour price on the consumption of rice. The negative sign accompany— ing the income variable was not expected, but its parameter was not statistically different from zero. 80 The own-price elasticity was -.325 at the observation means, and the cross-price elasticities were .112 for wheat flour and .374 for manioc flour. According to these results, it appears that changes in price of wheat flour did not affect the consumption of rice, whereas the opposite was more likely in the case of manioc flour, in which the estimated elasticity was greater in absolute value than the rice-price elasticity. The insignificant and negative income elasticity (—.Ol4) was less worrisome because consumption of rice remained practically constant over the sample period. Several attempts were made to improve the results of the rice- demand equation, but no success was obtained. According to Labys (1973), the commodity-demand behavior in several situations is more appropriately described dynamically. Consumers spread their response over some period of time when income or prices change. The introduc- tion of consumption in the previous period as an additional explana- tory variable altered significantly the results of the equation, increasing not only the goodness-of—fit measure but also the signifi- cance of all independent variables. However, the sign for the coeffi- cient of the lagged endogenous variable (consumption) turned out negative, implying a coefficient of adjustment greater than one, which was not easy to interpret and suggested overaction by market partici- pants. As a result, this modified equation was not considered in the final version of the model. 81 M Wheat is an important element in Brazilian nourishment. Because of the rapid growth in subsidized domestic consumption, imports have been increasing, despite a doubling in wheat production during the past decade from about one million to more than two million tons. In 1980, Brazil imported about two—thirds of the wheat con- sumed during that year. Lack of reliable data prevented the final demand for wheat from being expressed as final demands, such as for bread, paste, etc. A series of apparent consumption was estimated as follows: domestic consumption = .75 (total production — seed [100 kg/ha] - exports + imports - waste [3 percent of total wheatJ), where .75 indicated the conversion rate from wheat t01~heatflour. The absence of data related to carryover stocks of wheat apparently did not affect the apparent consumption measure as seriously as in the case of rice because there have been some indications in recent years that only a small percentage of the total wheat has been held. The aggregate domestic demand for wheat flour was postulated to be a function of its own price, the price of potential substitutes (rice and corn), and disposable income. The price of rice did not show satisfactory results in any of several models attempted to estimate the demand for wheat; therefore, it was dropped from the final version of the equation shown below. DWFL= 37.1263 - 2. 3850*** RPWFL + 1. 7295 PC + .2384 INCOME (.7263 (2.1522) (.2643) R = .91 D.W. = 2.17 82 where: DWFL = per-capita apparent consumption of wheat flour (kg) RPWFL = retail price of wheat flour (Cr$/kg) PC = producer price of corn (Cr$/kg) INCOME = disposable income (Cr$/inhabitant) The equation had a reasonable R2, and all of the parameters had the correct expected signs. The results indicated that the coef— ficient for the retail price of wheat flour was the most important and the only statistically significant parameter affecting the domestic consumption of wheat flour. The failure to represent all the forms of government intervention taken to protect the urban consumer and the use of aggregate demand to represent the wheat demand probably affected the true response estimates of the explanatory variables. The estimated elasticity of wheat—flour demand with respect to a change in its own price was -.46. This was very close to the estimated price elasticity (—.50) obtained in a recent study conducted by the Commission of Production Financing (CFP, 1981). The low cross-price elasticity (.09) indicated that changes in the price of corn did not affect significantly the quantity of wheat flour consumed. The income elasticity was also small and positive (.10), which was reason- able because the main final product of wheat flour is bread and it can be considered a necessity. 83 Meat In this study, three basic meats--beef, pork, and broilers-— were considered. It was intended to explore the interrelated nature of these three important products for Brazilian consumers. These results are essential for the final phase of this study, in which the model developed in this study represents an attempt to illustrate the interrelationship between the feed-grain and the livestock-meat sectors. The model allows for the simultaneous inter- action between the consumer and farm levels of the market system for several important commodities. As in the case of supply, one ideally should have monthly or at least quarterly data to analyze the allocation of available sup- plies of meat. Lack of data, however, prohibited 8 more refined approach, so the analysis was carried out in terms of annual data. The theory of consumer demand, discussed at the beginning of this chapter, suggests that the quantity demanded of a particular good depends on the good's own price, the consumer income, and the prices of the most important substitutes. Beef, pork, and broilers consti— tute the three most important items in terms of meat consumption in Brazil; therefore, at the retail level,beef, pork, and broilers are considered substitutes for each other. The effects of storage changes were not considered because of unavailability of data. It was assumed that year-to—year differences in cold—storage holdings are relatively small and that yearly consumption of each of the three meats is close to domestic production minus net trade. This seems reasonable because 84 most of the cold storage, mainly in the case of beef, occurs during the dry season (June through August). The statistical results for the respective retail-demand equations for beef, pork, and chicken are as follows: DB = 19.2472 - .4325*** RPB + .0119 RPP + .1718 RPCH (.0813) (.0475) (.1148) + .3355*** INCOME (.1066) R2 = .67 0.w. = 1.58 U I ll 01 .2995 — .2925*** RPP + .0763 RPB + .2995 RPCH (.0834) (.0699) 1992) + .0715 INCOME (.1551) R = .88 D.W. = 2.36 DCH = .2675 - .1515** RPCH + .0448 RPB + .2891*** INCOME (.0614) (.0307) (.0454) 2 R = .98 D.W. = 1.39 where: DB = per-capita consumption of beef meat (kg/inhabitant) OH = per-capita consumption of pork meat (kg/inhabitant) DCH = per-capita consumption of broiler meat (kg/inhabitant) RPB = retail price of beef meat (Cr$/kg) RPH = retail price of pork meat (Cr$/kg) RPCH = retail price of broiler meat (Cr$/kg) INCOME = disposable income per capita (Cr$l,000/inhabitant) 85 In general, the estimated coefficients were consistent with a priori expectations. The coefficient for price of pork in the broiler- demand equation was expected to be positive. In the initial run this coefficient was negative, and because the ratio of the estimate to its standard error (t-ratio) was too low, suggesting that the price of pork did not significantly affect the quantity of broilers demanded, that variable was deleted from the equation. In the demand equation for beef, the price of pork again was not significantly different from zero, but it had the correct sign. The own-price and income variables seemed to explain a great deal of the variation in the three meat demands, with the exception of demand for pork, in which income had a relatively low t-ratio. Estimates of elasticities and cross-elasticities for each product evaluated at the data means are presented in Table 5.1. BEET is usually considered to present a more price-elastic demand than pork. However, the results of this study indicated a more price-elastic demand for pork than for beef during the sample period. The price elasticity for beef (-.5811) was similar to that (—.555) found in a study by Lattimore (1974). This inelastic demand for beef with respect to its own price reflects the fact that beef consumption in Brazil is highly concentrated in the upper-income classes, and the variation in price does not greatly affect the quantity of beef con- sumed by these people. The small elasticities of substitution between beef and pork, and between beef and broilers, and the low income elasticity help support this point. Chicken is most responsive to 86 changes in income and to changes in its own price. In general, the computed cross-elasticities of price indicated that beef, pork, and broilers are substitutes for each other. Table 5.1.--Averageelasticitiesand cross-elasticities of demand at theretaillevel for beef, pork, and broilers. Elasticities or Cross-Elasticities With Respect to: Variable RPB RPH RPCH INCOME DB —.5811 .0194 .2198 .2452 DH .3354 —l.4842 1.0213 .2050 DCH .3580 .. -.9596 1.4838 Crush Demand (Soybeans) The total demand for soybeans in Brazil is derived from the export demand and the domestic crushing demand. The demand for soy- beans as food or feed is negligible and, therefore, can be considered practically a crush demand. This crush demand is derived essentially from the demand for soyoil at the retail level and the demand for soy- meal as an important component, because of its high protein content, in mixed feeds (Williams, 1977). The domestic soybean crush was postulated to be determined by crush capacity, availability of soybeans, and the crushing margin or relative profitability of soybeans. The soybean-crushing demand was assumed to be a positive function of the capital invested in crush- ing facilities, represented by the crush capacity, and of the availa- bility of beans, represented by the total production of soybeans. 87 The larger the availability and the capacity, the greater can be the volume of soybeans to be crushed. Soybeans, when crushed, yield two joint products: soymeal and soyoil. The variable used to represent the crushing margin was calculated as the ratio between the price paid by the crusher to the farmer for a ton of soybeans and the value of the meal and oil in a ton of soybeans (Spriggs, 1981). Thus, this variable was expected to have a negative sign. The larger the expected value of soymeal and soyoil relative to the price of soybeans, the greater the volume that can be crushed. The results of the equation used to represent the crushing demand are presented on the following page: DSB: 743111 -1216310 CRUSHM + 288. 410*** CRUSHC + .361463*** TPSB (1052470) 37. 435) (. 051133) R2 = .98 D.W. = 2.53 where: DSB = domestic crush demand for soybeans (MT) CRUSHC = soybean-crushing capacity (1000 MT) TPSB = total production of soybeans (MT) CRUSHM = crushing margin for soybeans (Cr$/MT), defined as: PSOYBEANS CRUSHM = YSOYMEAL * PSOYMEAL + YSOYOIL * PSOYOIL where the letters P and Y at the beginning of the words mean, respec- tively, the price and the yield for the corresponding product. The rates of extraction were not considered to be fixed during the study period because of the increasing level of technology 88 adopted by the soybean-processing firms. The prices of the two sub- products were considered to be expected prices; therefore, they were the observed prices for the preceding year. All the coefficients were greater than their respective standard errors and had the expected signs. The estimated price elasticity (margin) at the observation means was -.24. Although inelastic, this result indicated that changes in the relative prices seemed to exert some influence on the volume to be crushed. No direct comparisons with other studies were possible because the only inves- tigator known to have estimated a similar equation (Williams, 1977), in which the crushing margin was defined as the difference between the equivalent value of soymeal and soyoil and the price of soybeans, reported no results because this variable had a low t—ratio. In the present study it seemed that the variable used to represent the crushing margin was to a certain degree dominated by the other fac- tors (relatively high t-ratios). Feed Demand Soymeal Brazil has become a major force in the world soybean-products market. In recent years, Brazil and the United States have shared the position as the largest exporter of soymeal and soyoil in the world. In addition to this dramatic expansion in exports, the internal con- sumption of these soybean products has also increased markedly in the last decade. From 1971 through 1980, meal consumption increased tenfold. 89 Soymeal is the main component of livestock feed because of its high protein content (40—50 percent). In Brazil, soymeal is used pri- marily in the broiler industry, which consumes about 75 percent of the total. Swine and dairy cattle consume a small percentage of mixed feed, whereas beef cattle are mainly grass fed and fattened (Nogueira & Criscuolo, 1979). In this study it was hypothesized that demand for soymeal in Brazil is a function of its own price, the price of corn, and broiler production. The price of chicken should have been used instead of production, but it was not possible to obtain any acceptable results with this variable. The reasons for that situation are not known, and the problem remains to be explained. A negative relationship was expected between feed demand and the price of soymeal, whereas a positive relationship was expected between feed demand and production of broilers. The production of broilers should capture the effects of income and population in the poultry-demand equation; therefore, the income and population variables are not repeated here. Whereas corn and soybeans are complementary in terms of nutritional aspects, they can also be considered economic substitutes (Williams, 1977). Their relative economic costs determine the final combination of these products in poultry feed rations; therefore, the sign for price of corn in the demand-for-soymeal equation depends on which of the two effects is dominant. The results of the demand-for-soymeal equa— tion are reported below: .3..." 90 DSM = 401756 - 342.716* PSM + 273.493 PC + 2.431*** TPCH (174.512) (245.100) (.219) 2 R = .91 D.W. = 1.59 where: DSM = domestic feed demand for soymeal (MT) PSM = wholesale price of soymeal (Cr$/MT) PC = producer price of corn (Cr$/MT) TPCH = total production of broilers (MT in e.c.w.) The variables chosen as arguments seemed to explain relatively well the feed demand for soymeal in Brazil. All the coefficient esti- mates had the expected sign and values greater than their respective standard errors. The result for corn, although less robust, indicated that the substitution effect was dominant, whereas poultry production could be considered the main determinant of the demand for soymeal during the period under consideration. The estimated elasticity in relation to price of soymeal was —1.54; the elasticity with respect to the price of corn was .78. These results seemed plausible, although Williams (1977) reported a higher own-price elasticity in absolute value (-3.74), which can be explained by his use of a different sample period and different explanatory variables. 2933 Only about 10 percent of all the available corn in Brazil is used for human consumption, whether industrialized or "in natura.” The remaining 90 percent is used primarily for animal consumption. In 1974, about 35 percent of the total domestic disappearance of corn 91 was used for the mixed—feed industry, and 53 percent was used on the farms for seed and animal feed. In 1981, those figures were 71 percent and 17 percent, respectively, while human consumption remained more or less constant (CFP, 1981). Ideally, one should have several demand functions, one for each different use. Unfortunately, data for such an approach were not available. The only possibility was to have an aggregate function for domestic disappearance of corn, which was calculated as volume of corn produced - seed (20 kg/ha) - 5 percent for waste - net trade of corn. Another problem was that no reliable data existed for the carryover stocks of corn in Brazil for the study period. Therefore, the data used to represent domestic consumption of corn were a residual and as such were subject to random fluctuations. The demand for corn was then hypothesized to be derived from the final demand for hog and poultry meats. In the following equation, corn demand is expressed as a function of its own price, the price of a closesubstitute(soymeal), the production of hogs, the production of poultry, and the price of wheat flour. As in the case of soymeal demand, income was not included among the explanatory variables; it was hypothesized that the income effect was captured in the demand equations for hog and poultry meats. The price of wheat flour was included in the corn-demand equation to represent the price of sub- stitutes in human consumption (Thompson & Schuh, 1977). The statisti- cal results for the equation used in the analysis were as follows: 92 DC = 430284 - 2039.93* PC + 2086.54** PSM + 732.447* RPWFL (1157.86) (718.78) (425.906) + 3.650 TPH + 17.333*** TPCH (3.273) (3.146) R2 = .94 0.11. = 1.81 where: DC = domestic disappearance of corn (MT) PC = producer price of corn (Cr$/MT) PSM = wholesale price of soymeal (Cr$/MT) RPWFL = retail price of wheat flour (Cr$/MT) TPH = total production of hog meat (MT e.c.w.) TPCH = total production of chicken meat (MT e.c.w.) In general, the statistical results for this equation were satisfactory. The signs of the coefficients were all consistent with a priori expectations, based on the economic theory and on the back- ground given above. All the variables had coefficients larger than their respective standard errors. The estimated own-price elasticity at observation means was -.2458. This result was not far from the —.2010 figure obtained by Thompson and Schuh (1977) for the sample period 1947-70. The cross-price elasticities were .3979 for soymeal and .3303 for wheat flour. Both elasticities were greater in absolute value than the own-price elas- ticity, clearly indicating that soymeal is a strong substitute in animal consumption and that wheat flour is a strong substitute in human consumption. 93 Validation As in the case of supply, the estimated equations were simu— lated over the 1971-80 period to validate the model. The results are presented in Figures 5.1 to 5.3, which include the actual values. Table 5.2 presents the values of Theil's U-statistics for all the products considered. The demand equations were estimated based on apparent-consumption data. This procedure leads to errors larger than usual, since they are calculated as residual. The unavaila— bility of consumption data is aggravated by the fact that, even when available, data on carryover stocks are not reliable. Table 5.2.--U-statistics for demand equations. Equation U1 U2 UM UR UD R 1 Corn .0454 .7047 .0042 .0693 .9265 .82 Rice .0349 .4842 .1548 .0116 .8316 .83 Wheat flour .0280 .5707 .0289 .4587 .5124 .97 Soybeans .0423 .3656 .0013 .0048 .9939 .98 Soymeal .1484 .6377 .0001 .0064 .9936 .88 Soyoil .0455 .4731 .0362 .1684 .7954 .98 Beef .0249 .5652 .1279 .3278 .5449 .94 Hogs .0510 1.0554 .0213 .4195 .5592 .85 Chickens .0254 .3421 .0025 .0297 .9677 .99 The U-statistics suggest that the demand equations for wheat flour, beef, hogs, and soyoil are not as good forecasters as the others. The least reliable, though, was the wheat-flourdemandequation. The failure to include variables explaining government intervention in the domestic market mainly in the cases of soyoil and wheat flour may DR 50.575 49.311 48.046 46.782 45.518 44.253 42.989 41.724 40.460 39.195 37.931 36.666 35.402 94 D80 - - - - - - - - - - + 11.963 - f - 'k 11.126 - - 10.289 - * - . - +—+/’ - 8.615 - 4 * - 7.778 ' i c 6.941 - * - 6.104 - +" - 5.266 - * - 4.429 - - - + - 3.592 - /fi - 2.755 - +~+ - 10918 " * ' 71 '73 '75 '77 '79 ' DWFL * - - 44.448 3 --------- 42.887 - - 41.327 - «0 39.767 - t - 38.206 - +\t 3 36.646 - /C + i i - - + , . - 35.085 - 4 .- I' i — 33.525 - // * - .- - 31.965 - +-+ +/+ * - _ * - 30.404 - - 280844 ' + + t - - t + - 27.283 - t y/ - - +—+ - 25.723 - ’/ ‘o'o‘o'o'o'o'o‘o'o'o 'o'o’o'o'o'o'o-o'o 71 73 75 77 79 71 73 75 77 79 Figure 5.l.—-Actua1 (+)and simulated (*) values for demand: soyoil (DSO), rice (DR), and wheat flour (DWFL) in kg/capita, 1971-80. OIIIIIIIIIIIIIIIIIOII‘II‘l‘+ 95 DB - u - - - - - + - u - 20.221 — ‘9 - 19.760 - 3 - * — 19.299 - - - * — 18.838 - - — § .- 18.377 - - 17.916 - 3 - * 17.455 - . - 16.994 - + 3 16.533 - - - i 4- 16.072 - . + - 15.611 - +_+ . I 15.150 - \\ 3 140689 ' + : 71 '73 '75 77 79 ' DH DCH — t ------- - — ---------- + 7.320 - - 5.643 - - § fin - - * 7.085 - - 5.261 - - - - - + - 6.850 - - 4.8 ' - - - * _ 7d - +// - 6.614 - - 4-496 - J// - 5.379 - 4 - 4.113 - I - - - * - 6.144 ‘ - 3.731 - / - 5.908 - + 3.348 - +/’ I - + - fl - 50673 ' - 20966 - — - — - * - 50‘238 ' - 20583 ’ - 5.203 - + . 4 2.201 - // I .- * - .- * - 40967 ‘ / - 10818 ‘ - - + + - - / - 4.732 - //\\// 4 - 1.436 - /; - - + 'k - - - 4.497 - t + t - 1.053 - + - 'o'o‘o'o'o'o'o'o‘o'o '0'0'0‘0'0'0'0’0'0'0 71 73 75 77 79 71 73 75 77 79 Figure 5.2.-—Actual (+) and simulated (*) values for meat demand: beef '(DB), hogs (DH), and chickens (DCH) in kg/capita, 1971-80. 96 - -.-.-.- - - -.- DSB - - - - - - - - - - + 11.942 - t 11.061 - - 10.179 - - 9.298 - . 2 8.417 - - I - +/:o- + - 7.536 - i - 6.655 - . - - q. - 5.774 - v/ - 4.892 - + - 4.011 - J/ - 3.130 - // - 2.249 - - - 1/1 - 1.368 - +/' - 71 '73 '75 '77 79 ° DSM DC --------- - I: - 2.664 - I 20.325 - 2.447 - 19.614 - 2.231 - - ‘18.903 — 2.015 - . 18.192 - 10798 - - 17.481 - 1.582 - - 16.770 - 1.366 - j/’ - 16.059 - 1.149 - i + - 15.348 - - +\ * - .. 0933 ' + */ ' 140637 ‘ .717 I * \\// 2 13.926 I * L/ .500 - + - 13.215 - \ .284 " * 4' " . o - * _ a\,/ _ 12 504 _ .068 - + . - 11.793 - + 71 '73 '75 77 79 ° 71 73 C 75 ‘77 79 Figure 5.3.--Actua1 (+) and simuiated (*) vaiues: soybean crushing demand (DSB), and feed demand for soymea] (DSM) and corn (DC) in miiiion MT, 1971-80. + L Ol'lllllllllllllllltlll : ‘ i .. 5% in part responsible for the poor 9513*53 In the case of beef and hogs, other omitted variables and/or more detailed and accurate data prevent the model from forecasting well. CHAPTER VI PRICES, STOCKS, AND NET TRADE Price Determination A single price relationship cannot describe the price mechan— ism correctly without taking into consideration the assumptions included in other equations within the same model. Unfortunately, . there does not exist a general theory that considers actual rather than ideal market structure. Factors such as expectations are also important to speculation and hedging as well as to physical trading (Labys, l973, p. 92). According to Labys, several factors have con- tributed to this situation. Among them is the fact that in most studies, price relationships have been derived simply by inverting or normalizing the demand relationship. In the absence of such a theory, it was decided to concentrate on an approach that considers separately the factors that best appear to explain price behavior for each com- modity. For most of the commodities analyzed in this study, Brazil is relatively unimportant in world markets. This is not the case for soybeans and products. However, because it was beyond the scope of this study to estimate a world model for soybeans and products, the same basic approach was adopted for soybeans as for the other commodi- ties. That is, Brazil was assumed to be facing a horizontal world demand forall the commodities analyzed. Brazil was then considered a 98 99 price taker in the world market, and the world prices are given exogenously. In spite of the fact that most world prices are given exogenously, the government often intervenes to establish domestic market price. The intervention is usually through trade policy, with the purpose of keeping domestic prices below the level that would exist under other circumstances. The device most persistently used by the government during the study period was the maintenance of an overvalued exchange rate, which constitutes an implicit export taxation. Whereas this policy could be considered reasonable in the past when Brazil faced an inelastic international demand for coffee and consequently could shift the burden of the taxation onto the foreign consumer, even then it was harmful because it discouraged the export of other commodities that had a potential to become important in the world market. As pointed out before, the present government has selected the agricultural sector as its top priority because it is believed that this sector can help improve the balance of trade by increasing its exports. But to expand agricultural exports, the government has to adopt a series of policy actions. One important policy related to the export market is the adoption of a more "near equilibrium” exchange rate. This, it is believed would improve the competitiveness of Brazilian products in foreign markets and at the same time trans— late into higher producer prices, which could be an incentive to increase production. 100 Price Relationships: Soybeans, §gymgal, Soyoil, Rice, Corn, and Beef It was postulated that the level of government intervention in the export market for the agricultural commodities considered in this study could be explained through economic variables. According to Lattimore (l974), the level of intervention can be assumed to be predictable on the basis of certain exogenous and predetermined vari- ables. He showed that the level of intervention can be estimated as the difference between the FOB price evaluated at the equilibrium exchange rate and the domestic price of a particular product. To explain pastlevelsof intervention, Lattimore formulated an equation relating the calculated level of intervention to variables intended to account for the reasons behind the governmental decisions with respect to agricultural export policy. The current rate of infla- tion was hypothesized to be one measure of the government's interest in holding down the price of food items to domestic consumers. Another factor considered—-the desire to increase foreign—exchange earnings, which would reduce the level of intervention-—was hypothe— sized to be a function of the overall position of the balance of payments. In this study, those factors suggested above were incorporated directly into a price—detenmination equation. It is very important to have a relationship between domestic and international prices, which would allow for analysis of a possible removal of the restric- tive policy considered. 101 According to Abbot (l979), a country may choose different directions in determining its domestic prices. The situation gen- erally used in equilibrium analysis is that the domestic price of a commodity is equal to its world-market price at the country's border (DP = WP) or the world price times one plus an ad valorem tariff (DP = WP [l+t]). Another situation is when a country ignores the world market and controls its domestic prices by using, for example, a constant quota or a variable-levy system. Finally, in some cases there is a limit to which domestic prices can f0110W world prices- Only a partial adjustment may be allowed in some periods, but domestic prices should, in the long run, follow international market prices (DP = b NP). It was believed that the assumption of a partial-adjustment model was the most adequate in the case of Brazil. Although for most of the products Brazil can be considered a price taker in the international market, the domestic prices only partially follow the international prices. In order to accomplish some domestic social goals, the government intervenes through the overvaluation of the exchange rate and/or other policy measures. The aspects discussed so far were incorporated in the follow- ing equation for determining the domestic price of a particular agricultural commodity. DP = b0 + b1 WP + b2 EXCH + b3 INFL + b4 BOP + b5 TPROD where: DP = domestic price WP = world market price l02 EXCH = exchange rate INFL = rate of inflation (the rate of increase of the general price level) BOP = the overall position of the balance of payments TPROD = total domestic production of the particular commodity b0 = constant parameter b1...b5 = parameters measuring the influence of the respective variables on domestic price The estimate of parameter b] is used to calculate the elas- ticity of price transmission,1 which provides a measure of the response of domestic prices to changes in world prices. Usually, for sim- plicity, a perfect price transmission (D1 = l) is assumed, but given the evidence that the internal price is to some extent insulated from the world—market price, the size of international price adjustments cannot be ignored. The reason for the use of the world-price and exchange—rate variables as separate regressors was based on the study by Chambers and Just (l98l). They argued that there are differential adjustments to price and exchange-rate movements, and, therefore, the use of one variable, like the world price expressed in cruzeiros (world price times the exchange rate), cannot be considered to represent both effects. Production may affect the degree chosen for a controlled domestic price because in bad years a higher price may be permitted 1See Bredahl, Meyers, and Collins (l979) for a discussion of the importance of this price-transmission elasticity. 103 in order to cope with the temporary shortage (Abbot, 1979). Based on this evidence, total production was included as an explanatory variable. Because the prices of food items have a significant effect on the general price index, the level of intervention was hypothe- sized to be a positive function of the rate of domestic inflation. Therefore, a negative relationship was expected between the domestic price and the rate of increase of the general price level. The controlled domestic prices were postulated to be an inverse function of the expected overall position of the balance of payments. The need for foreign-exchange receipts (when the balance—of-payments account is in deficit) leads policy makers to lessen the degree of interference in the agricultural export sector. The domestic prices were expressed in national currency, and the international prices were expressed in dollars. As with the other monetary variables of the complete model, the international and domestic prices and the exchange rate were expressed in real values, with the respective price deflators having the same base year (1977 = 100).1 The conversion to a different exchange rate for simulation can be made by using alternative exchange rates or directly by making appropriate changes in the respective coefficient of the price-linkage equations. 1Notice that the exchange rate was deflated by multiply— ing the nominal rate by the ratio of the wholesale price index in the U.S. to the wholesale price index in Brazil, both indices having a common base l977 = lOO. 104 It was impossible to find complete series of published data on the export or import prices for the considered agricultural com- modities that were traded by Brazil during the studied period. In the absence of such data, the alternative appeared to be the use of the average FOB price of exports in each category. However, when data for the value of exports were found, inconsistencies were noticed in a number of cases, making it impossible to use the data. As a result, series of international prices (published by the IMF, FAO, and USDA) were used in the analysis as proxies for the world (export and import) prices of Brazilian agricultural commodities.1 Follow- ing are the results for soybeans, soymeal, soyoil, rice, corn, and beef, which have similar specifications. The numbers in parentheses are standard errors; numbers in brackets are elasticities. PSB = 707.180 + 9.11789*** WOPSB + 38.1829 EXCH (.7 3809) (27.2374) [ 7389] [ .1906] - 5.4046l*** INFL — .O6930** LBOP - .04971*** TPSB (1.44436) ( 02818) (.00508) [-.0778] R2 = .98 D.W. = 1.94 p = -.6307*** (.2089) PSM = 1750.35 + 3.86164*** NOPSM + 13.6015 EXCH (.92260) (39.4756) [.3264] [.0843] - 6.40846** INFL - .01802 LBDP + .02434 TPSM (2.43584) (.03378) ( 01559) [-.0246] R2 = .77 0.w. = 1.67 1See Appendix A for the description of these variables. 105 PSO= -8854. 68 + 18. 6771;** WOPSO + 1332. 52*** EXCH 6 197 (447. 86) [.5860] [1.1985] - 6827.l6** 07374 - 5.9157*** TPSO (2807.66) (.8614) [-.2090] R2 = .80 0.w. = 1.36 PC = -873. 819 +(5. .72959*** WOPC + 140. 642*** EXCH 1 798 31) (30. 288) [.4057] [1. 3797] - 4.27928** INFL - .01417 LBOP - .00976 TPC (1.70078) ( 02199) (.01024) [-.0891] R2 = .73 0.w. = 2.31 PR = 4907.94 + 5.53340** WOPR - 1.11938 INFL (2.05251) (8.05618) [.5780] - .224934* LBOP — .44929*** TPR (.125674) (.14018) [-.8658] R2 = .56 0.w. = 2.06 PB = 3436.29 + .42563** LPB + 3.63778** WOPB (.18219) (1.29914) [.2934] - 108.866 EXCH - 13.2129 INFL (476.638) (27.3088) [.1033] R2 = .60 0.w. = 1.78 h = .84 where: PSB PSM PSO PR PC PB LPB EXCH MOPSB WOPSM NOPSO WOPR NOPC WOPB LBOP INFL D7374 TPSB TPSM TPSO TPR TPC 106 price of soybeans (Cr$/MT) price of soymeal (Cr$/MT) price of soyoil (Cr$/MT) price of rice (Cr$/MT) price of corn (Cr$/MT) price of beef (Cr$/MT e.c.w.) PB lagged one year exchange rate (Cr$/US$) world price of soybeans (US$/MT) world price of soymeal (US$/MT) world price of soyoil (US$/MT) world price of rice (US$/MT) world price of corn (US$/MT) world price of beef (US$/MT) the overall position in the balance of payments (US$l,000,000) in the previous year the rate of increase in the general price level (percent) policy dummy, =l, 1973-74, =0 otherwise, representing the years when the government imposed an embargo on the exports of soyoil total production of soybeans (1000 MT) total production of soymeal (l000 MT) total production of soyoil (1000 MT) total production of rice (l000 MT) total production of corn (1000 MT) 107 Overall, the statistical support given by the coefficients of the above price equations was considered satisfactory. The coeffi- cients for the world-price variables, in all cases, were highly significant, indicating that they play an important role in the determination of domestic prices. The price-transmission elastici- ties provide an idea of the response of domestic prices to changes in world prices. Only a fraction of the increase in the world prices was transmitted to the domestic prices. A l0 percent increase in the world price of soybeans, for example, resulted in only a 7.1 percent increase in the domestic price of soybeans because of several ways of government intervention. The elasticity results show how strong the policy mechanism was in offsetting the changes in world prices. The exchange-rate variable was highly significant in the cases of corn and soyoil price equations. Furthermore, the estimated structural exchange~rate elasticities for prices, both larger than unity, indicate that the levels of domestic prices of corn and soyoil were very sensitive to the fluctuations of the exchange rate. The exchange—rate effect was relatively low in the cases of the other commodities. In the case of rice, this effect wasnegative,and forthat reason the exchange-rate variable was dropped from the respective equation. The effects of the balance—of—payments and the rate-of- inflation variables in some cases appeared to contribute to the explanation of the trade-intervention policies in determining 108 domestic price. This contribution was more evident in the case of the rate of inflation. In terms of the total-production variable, the elasticity results indicate that only in the case of rice was the level of domestic price very sensitive to an increase in total production. Support Price of Wheat The wheat support price equation was postulated to be a function of the international price of wheat, the exchange rate, the level of foreign-exchange reserves,and domestic wheat production. Because of the planting schedule, the variables relating to the preceding period were considered. Self—sufficiency in wheat production has been a long-term goal of the Brazilian government. A decrease in wheat production forces the government to raise the support price of wheat to stimulate domestic production. Because the government regu- lates wheat imports, the level of foreign-exchange reserves was believed to explain variations in the wheat support price. When the level of foreign-exchange reserves is low, the government may be unwilling to maintain a low domestic price (see Abbot, l979). PW = 1510.27 + 12.1631*** LWOPW + 91.5264)EXCH (3.1341) (71.2834 [.4502] [.3704] - .15995*** LFER - .22867 LTPW (.05388) (.17622) [-.0835] R = .79 D.W. = 1.77 109 where: PW = support price of wheat (Cr$/MT) LWOPW = world price of wheat in the previous year (US$/MT) LFER = the level of foreign-exchange reserves in the previous year (US$million LTPW = lagged domestic wheat production (1000 MT) Only the coefficients for the world price and level of foreign—exchange reserves were highly significant. The estimated structural exchange rate elasticity, although inelastic, indicates that the devaluation of the exchange rate would bring some increase in the domestic price of wheat. Producer Prices of Hogs and Chickens Because of the insignificance of the swine and poultry sectors in agricultural exports during most of the sample period, a different price relationship was considered in these cases. The prices of hogs and chickens were assumed to be a function of their total production, the price of beef, and a trend variable to capture the effects of any variables that might have been omitted. The results are presented below. PH = 26837.1 - 6.1689 TPH + .2288* PB - 177.164 TREND (4.4689) (.1291) (105.295) + 5190.09** D74 (1388.86) R = .81 D.W. = 2.65 110 PCH = 37210.9 — 3.5089 TPCH + .3248*** PB - 403.192*** TREND (2.3458) (.1026) (105.922) R2 = .88 0.w. = 2.96 where: PH = price of hogs (Cr$/MT e.c.w.) TPH = total production of hogs (MT e.c.w.) PCH = price of chickens (Cr$/MT e.c.w.) TPCH = total production of chickens (MT e.c.w.) PB = price of beef (Cr$/MT e.c.w.) TREND = trend variable, measured as the last two digits of the calendar year 074 dummy variable, =l in 1974, =0 otherwise The results suggested that the beef price was the most impor- tant explanatory variable in the above price equations. A dummy variable was added to the hog-price equation because of the dis- crepant price that occurred in 1974, probably as a result of the significant decrease in the supply of pork meat caused by the African swine fever. Retail Prices The retail margins were not considered to be constant but rather to be affected by several factors such as the quantities of commodities produced, processing costs, and the levels of producer prices (see Myers & Havlicek, l975). The procedure used in this study was to estimate directly a retail-price equation in which the retail price was hypothesized to be a function of the respective producer price, the urban salary, and a trend variable. The «flu-w- 111 influence of inflation and quantity produced was assumed to be cap- tured in the respective producer-price equation. The urban salary was used to represent the costs in the pertinent processing industry and marketing services. Following are the results related to the retail- price equations for hogs, beef, chickens, and rice.1 RPH = -36l64.9 + l.0482*** PH + l4.4857** WAGE + 360.992* TREND (.2428) (5.1796) (165.687) R2 = .91 0.w. = 2.29 RPB = -2627.67 + l.0887*** PB + 3.3672 WAGE + 47.8662 TREND (.1093) (2.9623) (106.548) R2 = .95 0.w. = 1.55 RPCH = 26074.8 + .9448*** PCH + 2.0540 WAGE - 271.210** TREND (.1935) (2.4302 (113.860) R2 = .94 0.w. = 2.56 RPR = 1841.6 + 1.0427*** PR + 3.7699** WAGE - 61.0173 TREND (.1909) (1.4408) (53.2931) 02 = .87 0.w. = 2.14 where: RPH = retail price.>.). 995.511004834621350233 09286190416224701.90070 00000000000000.0000003 1].! a\ . 0’ V037 ‘1 I. 0‘ 057d... . : 000.99C4:7.C:799=750 1.73561 4 .4 ...... 4.. .0000... 71 ..7 l. V 07900;:4199276581553392 0000000000000000000002 02°66357334649104046R. 3.1).". 04657.720r640781m311945 04444444433329.4163_0222 0000.000 ...! 3.... .. .....Ir 7. I... 092725315948941180149 0000000000000000000001 04298980930399.2“31128 (...... 003854011369842464126 035:63244432333452231v2 0 O 0 0 ......OOOOOOIOIOOIOO 1234 5678 901234 56 78 90 66666655677777777778 7.0.777", 11111101110L111_1111111 163 PCH PH P8 P50 2113800Q.960128$30126 00.000000000000000725 .27-02489018664341 o o 0 04. 076777817687108340302 099613585162769429828 066637545132142201285 011111111111111119997 041195526005029260955 ooooooooooooooooooooo 0295512874926796025161~ 008814754779431865453 08061841417571.04297.411 06n‘09833543023406132R02 011111111111112111111 89795403551305130732 0.0000000000000000000 05758508020140490541.72 069602108683320105307 079457234697431222963 043312743135601843628 0111111111.11122111.121 64595062477037 200067 00000000000000.000002 03591366413021.0205... 01 0520—53522587286553149 022316446019415172469 066675077665327623117 021122211111111111117 o O O O ..OOOOOOOOOOOOOOOOOOO 12345678901234567890 66666666677777777778 99999999 9999 9999 9999 1.111111111111111 1111 5880 NTC ESSM ESSB 0.0.0................OOOOOOOOOOOOOO...0.0.0.600... 7 77 7777777 0 a 00 0000000 0 .9 ...... +¢+++++ 0 F. [E EEEEEEE 000 00 o o 00 .00 007714118 0000 000040550 05885406 0.00000050880995091634 000006373567049471115 000938522542070134555 04292512141176111.1111 #4566564 06 0 013 o 0 0 0 040 0 . . . . 000.000.0000 0 o t o o o 0 0000000 0 o 0 00000000 0 o o o o 0000000000003 00000000000000000 00000090000000.0168 00000045245500.4704 swam 02455681254111.1112 0 000000000000 0 00 0 0 000000 0 0 0 0 0 000 000 o o o o o 0000000000002 00000000000000000 00000000000065022 00.000044217603402 000mm23567812346811912 E+G7 000.000.00.000... 00000000000000000 00000000000000.0001 0.000000000000090 00 0.3176203942010911. 67903463391121027 0% 05111112122466 0311 0 0 12345678901234567890 66666666677777777778 99999999999999999999 111111 11111111101111.4111 NTSM NTSB MN NTR 777777777 000000000 +++++++++ EEC...EEEEEE .0000 o o o o o 0 0594856013 000008990657349506089 0004154536605103739014 000120935334083365178 001085454551450133415 059230823921112345556 03464111225901.0(0000 0.. 777777 7 000000 0 + ++++++ 0.00.... EEEEEE E 000000 0 00 0 0 0734407. 0 09 0 0 0 00 063 0802030158822 0716063434252143965983 05740825816139.2338198 027302147092077365940 036345205181112332621 079317136322 0 0 0 0 0 054 0 o . 0 O O 07.7777777777777777777 000C00000800000000000 0++++++++++++++++++++ 0EEEEEEEEEEEEEEEE.LEEE 095853527 266665807312 055028664459017300421 069618517 2271552684652 089828063793150820365 08.116844639789404 6367 01.2221222211122232434 0.0000000000000000... .0900... o 0.000 0000 000000000000 03 0.0 025 069.0350 0 o 0 09389 085015830123 0177224401 025070755562922665651 090028462198105678085 010. 084133033455890685 026613452112236894162 0 - . . . . 0 o 0 .....OOOCOOOOOOOOOOOO 12345678901234567890 66666666677777777778 9999999999999999 9999 17.111111141111111 1111 164 NTCH ......O......O................... NTH NTB 00000 0 0 0 0 0 03 06956104 932097 668808 492016 0 0 0313581. 0000000 00 0 0000 0 008 0003 3203500 2573900 66128 09 00000000000001511412 0000000000000 000000000 0 0 0 0 0 00 0 000 0 0 0 0 0 0 0 0 0 0000000 000 000 0000000000000000000 002 .000000000274004080009 002211769515702283806 007915103470921548151 053248539112221811261 0 000000000000000 0 0 0 0 0 0 000000 0 00000 0 00495255 03.. 0 0 0 053 0 0 0 007 08823591. 000146974559209382475 0 072764270732304427477 047436159507122658042 06444444filflv43692244547 0 12345678901234567890 66666666677777777778 99999999999999999999 11111111111111111111 DSB DNFL DR DC 7777777778 0000000000 + 0. 0 4 ... 0 + + 0. + EErbE—LEEEEF. 0 0 0 0 0 0 0 0 0 0 03583676883 051448218126016194962 0684042772161354840684 057475907061112872053 00993168473706.1231562 069356795991224568881 01122233458 0.0 0 0.0 0 0 0 0 0 040811478281701262400 093718303443978149878 0496226411922904157223 0214104 31777616061932 ......0.............. 056672.356445679226875 022222222222222333334 06008841101700477 4648 066870370485458470151 0895924906692015060222 056420149331993422445 ..................... 002360550957907313539 04444:..344343343454344 0 . O . 077778888888888888888 00000030000090.9000000 0+++++0++++++++¢++0++ 0r_E.._E_Lr:E...r:LEE::LEEEEEE 088853649257745847821. 021912257793516407930 0269371741356443737381 046317.057 239721126276 049070.010111334456460 08.8951111111111111112 ..................... .................... 1.2345678 9010234567890 66666666 67777777 7778 9999 9999 99999999 9999 1111. 1.1.1.1. 1.111. 1.1.1.1 1.1.1.1. . 046252596489563831079 H 0 DB 080 DSM 091706311783313217015 0113450034980492861514 034235893222932509552 049533971922492694929 ..................... 077777278787655444655 . 020987643462576000898 048847502271248981802 0592120042602188756463 037474821591866174224 ..................... 087777678877554680086 0 11 1.11111 11 11.11 1122 11. 006400170290195232107 003292193917511512951 01.44565950623221694082 077839522868948808703 0561.4031.006090~Q72105 0 0345568990000000000 02 0 0 0 0 0 0 0 0 0113246679991. 777 000 0 0 0 EEE 0416 004 04 0 0 06 05 0 0 0665 0 0 0 02 0 07. 0496 010481047 0382135220537734580111 0822109445102464374227 073396820610081616257 011201434038561461112 06781541811273 0959 0 0 0 0 O 0 0 000000000000000000000 6’07 12345678901234567890 66 6’066 66 67 77777777 78 99999999 99 9999999999 111111.. 11111111111111 165 RPMFL RPHFL RPR DCH 0986781.“.1243962480641 029465763299515950826 0.00000000000000000004 .98350150812974452866 (35423333334944565336 0.0.0.... o454902657§59147100006 0961.15.68530702717 6° 35 l : .E 14:. .86767667675566533221 022350695206601190850 006:62461u8690118620794 00000000000000.00700002 7))7I.l:.l.7 5. I. <01}. .7 6844 0M7016WIEW20941848.99216 : t. I. . . I... .... .....ICO‘ .70682932486936727982 . .. ...... 001112334455 0 O O O I 0.000000000000000000 1239 56789012 34.567890 .66666666677777777778 99999999 99999999 9999 111111111111 11111111 RPCH RPH .0000... RPB RPLARD 7 V V. i A. < c ... . .... t 4. 00.000000000000000... <51. Dora“. 71634.83 AVG 30016994 042m. < ...: 08373826.?13109_237’368H 5 .7 7): I V ......I...O..O.I.h... .11937709872182698178 I is“ I l)‘\~l.‘ '4. 0672623 2R.16774—31009 11 V 7 7 v V 7 ...... ..A ... . . t ... ......OOOICDOICOOOOIO 0099640934531928598601 093502893357027 438935 076207J30589264442251 07799999897585.4“170711 r... .. «J ‘4. 1 h A I 0 O I . 12345678901234567890 66666666677777777778 99??) HOPSB NDPH HOPR UOPC 000000000000000000000 000000004 ......l...l.....l.... 010037.026376017019877 010111210012497237699 (l.l. 00...... 03041.5.1- 67: I." 02.31952 5122000 0 0 0 0 0 0 0 0 000000000000074236191 ..7 62.1 (78 I. .1 i .114 (5.1". 1.31:... 00000000 05.3..”08890900526749304 5.. E00 .702 00.00000700000000000000 012211111111235433334 056482315099817690335 056419185644237891848 .7 7 7 . 51 c689~l5.93692296 o o ..2 n o o 000000000000002920055 078345747186573115012 Oakl- .l . O 12345678901234567890 MM6666%66777W7777778 111111111.1111111 1111 166 HOPB CCAP HOPSO oUUCDUU OUOUUOGUOUUD 00 0188346 91740631u0.00_401 40:...11. .... 1.74:: 0......- .94811015915512676144 037385113n917882353308 015789917353 0 o .90 a o o .3 ooooohooooaquSlooSSéZ 005812671638158511192 .43-35654557811189.1112 .00000500000000005000 .953276523406023027682 ..“OIOOOOOOOOCOOII... . ...... €27 .8.223661799C333637069 022222221123248545665 00...... coouooooofivooooo¢ooooo 088989199911131112222 O i O O IOOOIOODOIOOOIIOOOOO. 12 3a. 56 78901234. 5678 90 666666 66 67 7777 7777 7.8 9999 99 9999 9999 99 99 99 11111111111111 1111. 1... pop FOREXCH BOP EXCH 01468135833570.5031111 .7.77.7 I. . ... r. .114 .. 4|. .00 00000000000 0 o o o o a .6 o o3 014.4. 63592443413511.9224. 1.. .....4 . 3’33 . .. . .- .— 00.00000 3.1).! I37011623197463;D7357 91 a o o .11.2~(3445566811125 a I t o o cluoooooo'ooo-ouoooo 12345678901234567890 66666666677777777778 l.l... . 43 1 . «J IND4 0.0.0.004 IN02 HAGE INCOME Er.l. .0724q.56430383938 0000 .8 a u u .u c336825.39~0:10315 LLB! .086793871518133mu0000 .1. E . .. ...: . .615357385532189... o o n 9 3319780 0 u o a o o o o o .0837 .7 n o c - 032532608900312 o .1135811112233471124 00...... 094883134862181950910 .4361903730’6997900487 IQCOOUCOOOOOOUOQOOOCOZ 0458512:~6386342096.361 .3207? .5“..32335334445676T6.6r.. .11111.111111111111111 n o o c 7777777777 0 7 O 051245519631111122222 09.666667789 a o o a a o o o o o I cocooooooonoooooooco 123Q5678901234567890 $665m655677fl77fl77778 q 4 1. 4|. 1. .J 1.1. 167 INFL UPI IN017 INDE O noncoouooooaoouoooooo 900000000000000000003 o666024628549198450524. 006358022493396826790 ......COOOOCOCOOOOOIO 071506889090748712830 ‘35795322212112244351 O .0000 00030000 000000 00 0 00300000 0000 00000000 .990fl06706874908sooh. COCO-......COOOOQGBUB 7.47.”;‘1. . 4 4 4 l 000467988412158060000 c . 0134679111223461124 7.! 7.5!.» 1 (ICE .936875400218387107721 .4598382621787727 . u u o 05.072821 0 o o o o o o I .0726 0600.00.0469486790439 .01134791112223961124 .....ICCOOOOCOOOIOD. 123‘. 56769019.. 34567890 6666 66666777 7777 7778 9999 9.999999999999999 1111 1111 1111511111111 ifititit‘fiffitfiii'fiiiiiiiii..iitilifittiltitiiii'kfifiiiiiittfifitttiiifiiiitttfiiii APPENDIX C COPY OF THE FORTRAN SUBROUTINES USED TO SOLVE FOR THE ENDOGENOUS VARIABLES IN THE MODEL 168 169 DEL METHOD SIMULTANEO SDAIERS 1976) ENDOGENOUS VARIABLES M L M 0 RSH B FOHRSH B S H PPPCSB HSCTTT P HS A??TTTPPP?DDDNNN??R?HE I. i. I.- C13! 1. at: . ~ I 4 l HM C S R PPP USCRSHTTT P P S A??TTT?PPPDDDNNN?R?R?E 25814.7 03692581.“.1036925 1112222333044555566 40H 0H B B wsc M 8 SCR H S cs PPPRSHCSB TTP P S AA9.0.T.ITPPPDDD¢.?NNR7.R0.—L 7 ll. 1111222333444455566 0 H 3 ONE US B C7 PP M CS X99 PCP2 OOXRS YY €67 ON07 UHEYY?TT?EDD?BIPD 36925814703699.5841 111222333344‘5 R C D ”C 84 X R R3” H67 SEA HHHYY??T?EDDDFHRI 25814703692581470 11122223334945 S E L B A I R A V S U 0 N B 05 L ECSB N7 LFP XHHH?YY?T??TDD?IRC E 1.4103692581470369 1111222333444“ 170 l\ I ‘1 I 9 9 z i t u 3 6 C. 1 II 5)J I 8 I I2 10 8 9 fl. 0 95 11 9 o ) II 1 9 29 t 6 I 8 _ 4) 5 4. ) 82 97 5 3Y AV) 4. 6 ) 68 1J23 3 6 1 32 27 1 6 1) 3 3) 81 62 3 I(( 6. 1 . 9( )86 0 4I 01 0 OJ 3( 72 23ZY 7 5 J 1Y 710 6 1 ) I. I I 9 SY 12 I2(E 2 I IE 1.3 I 6) J J 6 8 12N I. 9...... I9 I 4. 9 2 I I 686) 3 3J I. O 9 #3 7 +2.19 I) ) ) ) ) 4. 1* 03 I)) 1 I! +5 4. ... .\ 5.9 I I II ) 1. 2 3 A. ) IO 3 3 24 ) 6 3 )2 )Z O ) 2) #1 1 1 1 1 1) Z -) a 60¢.2 O1. .91 )lI 1‘ ) 1 R10! ) )) 1I01 . II II ( (I .8 ) 2 ). I Jl. . ( JZ .Y 5 .I O) 1) - 5. )J VI Y Y Y J1. J. )1 1F. 1)) DY )J )2 O. J 1 J 1 .1 J94J II E E E E I( I. J( .5 ()J2E J! J ) ) 6I I. I. 02 ). J . I05 I. .3 I 9 I 9 SY 11. IY J7 1 94. I2 I+ 6 5 ( 4. Y 60 1J ’J 3801 J2 1E (a 2Y '6 § . 3‘0 54. 9 1 1 26 6) Y 64. . I 2 I (8 I3 I( ) ) ) ) ( Z. (I 38 J47. (( 4) ( ( 1 (4 (1 J2 1‘“ Yo ( AY) J J J J Y9 E Z (2 ) III ) 7.2 (1 Y Y I 0 Y2 1 Y1 I2 Y2 IIZ ( 4 I I I I I7 3 22 142I) 2. E E 7 ( 1 1° 0 IO YI- 1 2 3 A. I) 3 I0 I .0( 1 II J 12 IV. 3 I 29I (Y IY )I Y 1 1 1 1 J 97 . I JZ 7. II I I 78 Y 3 I Y 7)) I)... II ( ( ( 3I 51 DE . I I7J 41 5 76 7 7 5 I OI 3)11 J 2 Z Z Z 74 94. 40 1 4... 6O 6O 02 ) ) 6 7( 2 6) I 5 51.1 699 53 21 39 3) 2 104. 1. 1‘ J J I. 1 0 5) MOZI ) 1) 9.d6 23 I I I I 1( 7I 32 61 (7631 SE 4V I I 8 2I 2 91 1J OJ 7J I2 72) 4.2 I 59 1 . Y746( 81 I 6 5 1) 8 0 O. 2 I. 0 O 1 .43 9(J 1 1 1 1 9 5. .4 OJ 6° CY I9 8‘ 1 1 6 17 0 6J 72 54. I32 92 I 0 0 0 3 II 5A. 2 I I04 31 8 l. ( 01 I6 I I 72 .2 2(9 9(1 0 u 0 0 O) I 11 3 It I 0 2) Z z (I 3 O 6 81‘ l\ 0(VI I 5 o o o 0 #8 J 4 9 765 08 J BY + I O2 XDII I2 I2 Y ) I( I 9I . +( 7. )7 1 II I I 6Y 3 3 ( I ( ( IJ I Z I I I I 14 13 4 Z 6 .17 SI 5 II 72 3 7Y I 0 #I I A .4 84 9) 7 OI .6 3 11 ) ) 4 31 I I 0 I 9I 5 )1 29I ) ) ) ) 72 It.) I1 2I 3 )J0 I.) 1( A. 4 52 9 9 9. D 7 3 1)J3 I2 1 2 3 Q 33 321 7 -_ I 75)) 13 0 5 17. 3 3 as 0. 7) 0 0 34 01. “J 0.0. 261 I. ( 1 (I 48 7I2 DJ; 09 7 IJ126 5)2 31‘ ( ( 8. 7 05 6) 'MZVIG1 32 26 5 04.7.. 675 VI Y Y Y 5 8 l\ 9 9 14 6 9 . A. 0 7J( A. Y Y E A.) 16 86 XOFYO 29 73 532 92d. Y Y Y Y 1. .8Y 4.2 59 0 . 9J( 1 IY 7/ Y Y .1 .4 .( 12 ACN IX/ 26 48 82(I 1 I0 2E 4. 19 3 ( 92+ 3E 7 2) Y ( M O 0):. 33 .0 6. . (Z 6. 5 = = : : : ) : 9 : (. 5 6 : 29‘ :4 : ) : : :5 : (A. : E : Y V 2 I I 2(5 .5 ) 79 Z :1 I (II ) (I 4 1 Y6 Y N300 9E :0 .. — ...l O = ) ) ) ) )1 )2 ) ) I -) 7.. 4 )7. )2 ) ) )9 )Y( )0 )( 1C2YR I . I7 )0 1 2 3 4. 8. 1 I) 2I 3 GI 1 5 ) 4.0. 6 5 65 4 Y 5 6 I‘2‘NE ) )) _.)I 2 )) 1.. 1. 1 1 1J 241 2 2. :2 I9( 2I5 3Y 1 1 2o 6IE 6) 6I )XRIIVII 1. 2) 3 65 41) ( I. II II (I ( 1 (Q. l\ (9 4Y ( 1 (Y I\ l\ ( l\ 1‘) II JAAYVFF ( (1 012551 (.1 Y Y Y Y Y1 Y . .lx Y3 Y))OY87UY Y6( Y Y Y Y. Y2+ Y1 Y8 (HPLEEE Y) Y. HY228 YJ. Y Y Y Y Y2 Y Y Y9 YJ3HY773( YQY YI Y Y Y Y1 Y. Y4 RYNNRDO Y) YJ RY26 Y IJ ) ) ) ) )( ))Y )4. ) I2K)169 )EYD) ) ) )) )4) )J )2 TNT. OF C )1 ) I) )G3I )4 I 0 O 0 0 DY OJI 02 09( 0106I DOINOO 0 0 OJ 0 JJ 0 I. 06 ) III 0. U22I921 027 I I IN I I I I I2 I(Y I I I35 I4 A I1 I I I I .0 I I5 I6 N I 1U)N IJ I2( I2 I) 30.2 G G 000 GI 095 G I QZE 096 I4 0 I4MG3) GNG 09 037 02 0 I OZNOOO OI 0(Y GI ) 0Y( E E EIE E E(0 E E E I.0 ESOEEGANEOE E( E23 E( E8 NNI20P E1 EYE E 92 E Y I IN IT I I) I2. I. II o It . I .0 I130 21 I I2 I ( IY I F(2M I2 I I9 . II N)N)0)C) ) E ) ) ) ¢)E ).E )2(I).I) ) S)+Z ) )+ E IRYCOS )( )I O ) )J ) I0102I3U4N8. 1I9 2) 34)E¢U )2 5 4H41Y76C5 6I K4. $5. 6 NYYFYCN 1Y 2 3)) I A.) 1111T10101J 2 2 2J 261L28)10E2)5$3 EClUl 2 )C6)IK6 $6) ININE 0 ( ()) (29.4A‘J57(T(C(0(1( I (20 ( I. (2.1(1J.1C(J4U(. U(D(E(260(4 C()K(6 T. 12 11‘ TI Y)E1 I ITIEIPII 1 .1 71 E1 EIYJIIEIII E1 EIIII RE AN I. E 12A 1233 12N E A IN 12E 1A 123 1234M1123 120 H 0 123 I I M 10 I I ER C E Y R C E VI 6 R C E VI Y Y Y VI Y Y Y NO I H O O 1 H 0 0 0 .1 H .0 MW 0 O 0 0 0 0 0 AC R H s C R ul S H C R H s s s s s S S 8 CC C C C C C C C C C C C C C C C C C C C C .228766 (YY(29)IYY(28)) 9) 50857E-02 ' YY(19) I .326789 171 ) u 6 7 7 3 3 4 1 J 0 4 93 7 4 3 . I I) 5 7 ) 6 9 90 2 2 8 J 3J 1 5 I9 6 2 I 3. 9 3 4 I 7 1I 7 8 I 1 8 7 2E 19 6o 5 3 9 ‘3 6 4 4‘ 8) 1 3 54 1 7 4 1 9 YQ 3 I 4 SY IJ I 9 I7 I) I) I ( 6 ( I 4 5 IY I 4 7 J J Y 7 12 3 1 0 6 +5 I I #5 II I. 0 I I BI 4 3 5 5 5 . 3 )I I I 2 0 )( ) 7 )2 )4 )Q ) 1 26 22 3 2 6I TL 8( 9 ) I .6 ) I 2 5 5 2 52 52 5 7 J8 7 ) I3 (I ( I (I ( t I 9 ) '3 2 a ) 3 Y Y Y Y. Yt Y I 2 35 Q 2 9 )0 Y4 Y ) Y) Y Y Y 2 2o Y ( 2 17 0 5 ) 3 4 I ( (6 Y Y ( 21 I. I 2 I1 ID ID I Y Y1 Y Y ( E ( . . . 5 Y ( I Y Y. 29 2 Y “J 3E 3E 3 7 . I Y 06 0) Y 0’ 04 06 0) I Q I 5 I ) .9 .3 .6 .0 .3 .9 ( ) 6 5 I8 59 E3 I E6 E5 E5 E3 / 2 5) 8 1 7 1 93 1( 1( 55 95 1( ) 7 63 8 8 7 3( 54 OY 6 BY 13 31 7Y ) 2 26 U 4 7 9Y 6. SE 1 5 53 67 “E J 4 9( I 0 9 IY 9 3 7 3+ 29 20 1 I 0 BY 1) I) 4 9 9. 3O 0 1 3 9 50 8) I 5Y 7 18 9 3n 1 2 2 6) 9+ 2+ 1 4J 1) I +5 5 I 0 I) I) 9 I6 I I I) (I 2 I ( +( ) 29 J J 3 .1 ) ) J 20 +5 I 7Y Y 09 4 .9 .I ) ( .9 .9 .I I5 ( ) 6: 9E 5 .0 6 5 .5 9Y 5 5 5 )( 6Y 61 I I 8( 5 34 34 3 UY 2( 1( 54 32 CE I. 79 AI IY 46 0( 6( 6( 5( 7Y 5Y 4( 3 I 9J 2 6 4E 8o 62 22 5Y 1 “Y 9Y 52 (I 19 69 6) 1) 7 23 1 1 7E 4I 2 9 7 Y 4 43) 2J 6J 0+ 0 II II 1 I II 2I 6I Y) 8) 613 I 3’ 6 30 6 7 0+ 15 9 I 2 (3 1J 2(5 .1 .1 2) 4 32 33 4 .0 13 53 I3 6 O Y( 3 3 J ) 0 0 ) c 0 0 0 :( ..1 ..(Y ..( :l\ .. I :3 =- =— ..9 ..E :— :- ... Y 3 E 2 2 1 5 E E 1 3 E E E )E )( )/ ) ) )3 )( )7 )1 )( )2 )8 )4 )1 ( (IE()) (2 (2C( ( (0 (8 ( (1 (1 (9 (9 Y) Y CY3J Y6 Y9IYI YI Y6 Y3 Y! Y3) Y7 Y9 Y8 Y3 Y31Y69 Y6 YQRY Y Y3 Y2 Y YI6 Y1 Y2 Y2 )( )7R)(1 )aE)IP)1 )7 )0 )I )4 ) 3 )0 )I )I OY 01P0Y3 DICDO 02 04 0 0 08 0I( 0 0 0 IY Io IY‘ .7106LII I4 II II '1 I Y IO 0+ Dot 0 QILQ Z QQRG3IQ1 00 Q Do 03 Q)E Q 0 N0 EQEE1IE* EE PE AE7 E2 E) NE) E“ EJ E) E) AE) I CI6AI ICII I+TI2 I3 I3 AI1 II III I8 I7 MI7 ))1) T)) 1) L) E) )7 )5 ")2 2 7 )5 D)5 E)5 37R2.E3J6R7)18)R9. 1 2( E3(05 064) 7( N8( 09( FIRF2.F).RFIEFIEFZDFYIDF.I.F.AFYLF.IDF.EEF.EEF.EN K1 K1 P1 TI 1 711 I1H1 K1 1E TIEEI 11E 1E HIE KIERE IA 15 1A I1F 1C 1NI12E 12A 1M 10 12F 12 12C 1 c E E16 R c E Y Y E G I H 1 H E 0 H 0 1 H 0 0 E 0 H H R H B H C C R H S S B H C C C C C C C C C C C C C C C 172 )0 0A. 0 u 9 U I 4 0 20 \l \I 9 "c \I \I ) 1 1)22 1 2 4 . \l \I ) \1 ROM( 3 ( 6 J 6 7 8 9 YDAZ I! Y ‘ O 6 A. ‘ 4. .19—N O Y Y Y 5 0‘ (x ( l\ t ) Y Y 6 Y Y Y Y !Y to l (v Y E F. F. 0N)“ . g Y RIO O 6 u 9 O O VIC-OD ) 7. \I O Vic—L20 1 0 Q: \l ) \l ) 0(2 I. o ( \I J J J J 9 Yl‘ VI VI 5 I I 9 9 T OMY Y - Y 6 a 8 8 8 I)AT ( k. A. Q. 4 RUNC t ) t Y ( l\ ( ( co A 2 Y Z Z Z z 2 9 ’ 2 1 a 9‘), 0 0| 1 — Q t Q t X 00 0 Y o A634 Y ) ) \I \l \I M02 I . - 5 6 7 8 9 RXl‘o l 3 3 3 3 3 YEZO ) ‘I ( l\ l\ ( l\ N L2 1 0 4 Y Y Y Y Y 9N( 1 9 1 Y Y Y Y Y Y) Y a! 0 II X00, Y Y . . . . . A2)) Y ( Y "(00 ) ) ) ) ) CNDU a a n 5 6 7 6 9 N022 1 1 1 1 1 P(( 5 8 5 l‘ ( ( l‘ ( 'HZYG 9 6 9) Y Y Y Y Y XOFYO o o .9 Y Y Y Y Y ACN OX ) Q H O. .)E = = 2 2 ( = = = = .— 7.))”. 6. Y N300 ’, \l )0 )E ) \I \I ) ) 102Y 1. 2v: Q. 5 6 7 8 9 1(2(N 4 4E 4+ 4 Q h 4 4 XR(YI ( l\ 1'. l\ l\ l\ ..\ 1‘ )AAYVF Y Yo Y) Y Y) Y Y Y JHPLEE Y Y Yd. Y Y6 Y Y Y ‘YNNRD \I E)\l )3 \l \l“. .0: \l \p 0”} 9P n 00:“ 0‘ 0 01V 0 a o N \l 9 9 o A o c. CY - oY o o I E , ’0)N nu RGBEGY 0 GE 0 0 Q PZNOOO E Tut—“DE EE E E EEE YNIZOP o o‘A o _ D n F. to o 00 o leZMn ) T)ZR) A) D) ) )A) E ORV-(0E1 E2 T4)R5 A6)E7 8R9, NYYFYCD‘. N“. “.qu R‘lDfl-rlaTA. ININE A‘ ( TI\— 1\ T( _ A(D( ( T RRG EG)EGJTG) GJRGAGTG UN ETO)C02NO 9.5.05T0 ITOROEON 00 G x11x3 XA‘quEXS XTXNXR Run ETE4RE(SE6 ....(NEETE .L EU BH TE(( (YN((L(Y ((E(T(N(TD U0 NNFYDFYAFYAFELFYNFEFEFEN SC I IEEI El E1 II INIKIRE 123 N 1L 18 1M 10 1F C R L Y Y Y E G I 0 I 0 0 0 E 0 H C In S S S B H c C C C C C C C C YNNRDX 3 ))))\I)\I\l "1 ‘P ‘4213 11223344 F \l ..M997 I 333314333 E C 9°)NR5 0.0.5 (((l\(l\l\( F‘ZM .8291 NNNNNNNN : 1:: = : __ : ININE Iv__: ...—E T R ))))U)\I))))\I\l UN EOJJJJN11223344N 00 GD 9 O I 212222222A£R RM E11234T(((I\(((1\U BM T 2222NZZZZZZZZTD U0 N0(((I‘OFLFLFLFLEN SC IDZZZZCNNNNNNNNRE 123 o 0 1 BIBLIOGRAPHY 173 BIBLIOGRAPHY Abbott, P. C. "Modeling International Grain Trade With Government- Controlled Markets." American Journal of Agricultural Economics 61 (1979): 22—3l. Adams, F. Gerard, and Behrman, Jere R. Econometric Models of World Agricultural Commodity Markets. Cambridge, Mass.: Ballinger Publishing Co., l976. Agricultural Statistics. U.S. Department of Agriculture, various issues. Alves, Eliseu R. A. "A Pesquisa e 05 Ganhos de Produtividade em Culturas Alimentares no Brasil." Paper presented at the Congresso Interamericano sobre Agricultura e Producao de Alimentos, Forum das Americas, $50 Paulo, set/l98l. Anuario Estatistico. Ministerio da Agricultura, Comissao de Financia- mento da Producao (CFP), various issues. Anuario Estatistico do Brasil. Ministerio do Planejamento e Coorde- nacao Geral, Fundacao (IBGE), Rio de Janeiro, various issues. Ayer, H. w., and Schuh, G. E. "The Effects of Brazil's Export Policy on the Cotton Fiber Market: A Trade-Off of Agricultural Develop- ment and Foreign Exchange Earnings for Immediate Consumer Bene- fits.“ Paper presented at the Workshop on Price and Trade Policy and Agricultural Development, Purdue University, August 9-ll, 197l. Bacha, E., et al. Analise Governamental de Projetos de Investimento no Brasil: Procedimentos e Recomendacfies. Relatdrio de Pesquisa No. l. Rio de Janeiro: IPEA/INPES, l97l. Barten, A. P. "Estimating Demand Equations." Econometrica 36 (April 1968). . "The Systems of Consumer Demand Functions Approach: A Review." Econometrica 45 (January l977 . Bergsman, Joel. Brazil——Industrialization and Trade Policies. London: Oxford University Press, I970. 174 I75 Bergsman, J., and Malan, Pedro. "A Estrutura de Protecao Industrial no BraSil." R.B.E., Rio de Janeiro 24(2) (l970): 97-144. Boletim Estatistico. Fundacao IBGE, Departamento de Divulgacao Estatistica, various issues. Bredahl, M. E.; Meyers, W. H.; and Collins, K. J. "The Elasticity of Foreign Demand for U.S. Agricultural Products: The Importance of the Price Transmission Elasticity." American Journal of Agricultural Economics 61 (1979): 58-63. Brown, A., and Deaton, Angus. "Surveys in Applied Economics: Models of Consumer Demand." The Economic Journal 82 (December l972). Brundy, J. M., and Jorgenson, D. W. "Efficient Estimation of Simul- taneous Equations by Instrumental Variables.” The Review of Economics and Statistics 53 (August l97l). CFP--Comissao de Financiamento da Producao. "Estudo do Consumo de Alimentos Basicos no Brasil.” Vol. I: Resumo e Conclusoes Finais. Outubro l98l. (Mimeografado.) Chambers, R. G., and Just, R. E. “Effects of Exchange Rate Changes on U.S. Agriculture: A Dynamic Analysis." American Journal of Agri— cultural Economics 63 (February l98l): 32-46. Coirolo, Luis 0. ”An Econometric Analysis of the Beef-Cattle Industry of Uruguay." Ph.D. dissertation, Michigan State University, 1979. Conjuntura Economica. Rio de Janeiro: Fundacao Getulio Vargas, various issues. Durbin, J. "Testing for Serial Correlation in Least—Squares Regres- sion When Some of the Regressors Are Lagged Dependent Variables. Econometrica 38 (l970): 4lO-2l. Food and Agricultural Organization of the United Nations (FAD). Production Yearbook. Rome, Italy: FAO, various issues. Trade Yearbook. Rome, Italy: FAO, various issues. Gemmill, Gordon. "Estimating and Forecasting Agricultural Supply From Time Series: A Comparison of Direct and Indirect Methods. European Review of Agricultural Economics 5(2) (l978): l75-9l. Hall, Lana L. "Evaluating the Effects of P.L. 480 Wheat Imports on Brazil's Grain Sector." American Journal of Agricultural Economics 62 (February 1980). Homem de Melo, Fernando. "A Agricultura nos Anos 80: Perspectiva e Conflitos entre Objectivos de Politica." Estudos Econ6micos-- IPE/USP 10(2) (l980): 57-lOl. l76 International Financial Statistics. Washington, D.C.: International Monetary FUnd"TIMF), various issues. International Wheat Council. Review of the World Wheat Situation. London: International Wheat Council} various issues. Johnston, J. Econometric Models. 2nd ed. New York: McGraw-Hill Book Co., 1972. Kloek, T., and Mennes, L. B. M. "Simultaneous Equation Estimation Based on Principal Components of Predetermined Variables." Econometrica 28 (l960): 45-61. Kmenta, Jan. Elements of Econometrics. New York: Macmillan, l97l. Knight, Peter T. Brazilian Agricultural Technology and Trade: A Study of Five Commodities. New York: Praeger Publishers, Inc., 197l. Kreinin, Mordechai E. International Economics: A Policy Approach. 2nd ed. New York: Harcourt-Brace-Jovanovich, Inc., l975. Labys, W. C. Dynamic Commodity Models: Specification, Estimation and Simulation. Lexington, Mass.: D. C. Heath and Co., 1973. Lattimore, R. G. "An Econometric Study of the Brazilian Beef Sector." Ph.D. dissertation, Purdue University, l974. Leff, Nathaniel H. "Export Stagnation and Autarchic Development in Brazil, l947-62.” Quarterly Journal of Economics 8l (May l967): 286-301. . "The 'Exportable-Surplus' Approach to Foreign Trade in Underdeveloped Countries." Economic Development and Cultural Change l7 (April l969): 346-55. Lopes, M. R., and Schuh, G. E. A Mobilizacao de Recursos da Agri- cultura: Una Analise de Politica para o Brasil. CFP--Colecao Analise e Pesquisa. Vol. 8. l979. Maddala, G. S. Econometrics. New York: McGraw-Hill Book Co., l977. Mesquita, Alamir. "0 Desempenho da Agricultura Brasileira Nos Ultimos Vinte Anos e as Perspectivas da Soja no Mercado Internacional." Speech presented at 11 Seminario Nacional de Pesquisa da Soja, Brasilia, Fev. 198l. MSU Agriculture Model: Long-Term Forecast of U.S. and World Agri- culture. East Lansing: Department ongricultural Economics, Michigan State University, Spring l982. I77 Myers, L. H., and Havlicek, J. "Monthly Price Structure of the U.S. Beef, Pork and Broiler Markets." In Quantitative Models of Commodit Markets. Edited by Walter C. Labys. Cambridge, Mass.: Ballinger Publishing Co., l975. Nerlove, M. The Dynamics of Sgpply: Estimation of Farmers' Response to Price. Baltimore: Johns Hopkins—Press, l958. "Estimates of Elasticities of Selected A ricultural Com- modities." Journal of Farm Economics 37 (l956gz 496-509. Nogueira,~Jr., and Criscuolo, P. "A Soja na Avicultura." Agricultura em Sao Paulo, SP 26(ll2) (l979): l37—5l. Paiva, R. M.; Schattan, 5.; and Trench de Freitas, C. F. Setor Agricola do Brasil. Editora Forense, Universitaria/USP, 1973. Peterson, E. Wesley. "Adjustment of the Spanish Feedgrain and Live— stock Sectors Following Accession to the European Community." Ph.D. dissertation, Michigan State University, l98l. Pindyck, R. S., and Rubinfeld, D. L. Econometric Models and Economic Forecasts. 2nd ed. New York: McGraw-Hill Book Co., l98l. Prebisch, R. 0. ”0 Desenvolvimento Economico da America Latina e Seus Principais Problemas.” RBE, Rio de Janeiro 3(3) (Set. 1949): 47-lll. Precos Minimos. Ministério da Agricultura, Comissao de Financiamento da Producao. Precos Recebidos Pelos Agricultores. Fundacao Getulio Vargas, Instituto Brasileiro de Economica, various issues. Rao, P., and Griliches, Z. "Small-Sample Properties of Several Two- Stage Regression Methods in the Context of Auto-Correlated Errors.“ JASA 64 (1969): 253—72. Reynolds, R. M.; Heady, E. 0.; and Mitchell, D. 0. ”Alternative Futures for American Agricultural Structure, Policies, Income, Employment, and Exports: A Recursive Simulation." CARD Report 56. Ames: Center for Agricultural and Rural Development, Iowa State University, June 1975. Schuh, G. Edward. "The Exchange Rate and U.S. Agriculture." American Journal of Agricultural Economics 56 (February l974): l-l3. "A Politica Cambial e 0 Desenvolvimento da Agricultura no Brasil.” Anais da XIV Reuniao da SOBER, l976. 178 Spriggs, John. An Econometric Anal sis of Canadian Grain and Oilseeds. Technical Bulletin No. 1662. Washington, D.C.: USDA, Economic Research Service, December 1981. Theil, H. Introduction to Econometrics. Englewood Cliffs, N.J.: Prentice-Hall, 1978. Principles of Econometrics. New York: John Wiley and Sons, 1971. . Theory and Measurement of Consumer Demand. 2 vols. Amsterdam: North—Holland_Publishing Co., 1975-l976. Thompson, R. L. "The Brazilian Soybean Situation and Its Impact on the World Oils Market." Journal of the American Oil Chemists' Society 56 (May 1979): 391—98. "Structural Relations for Agricultural Trade Policy." Paper presented at the Annual Meeting of the American Statis- tical Association, Business and Economic Section, Chicago, Illinois, August 15-18, 1977. , and Garcia, J. C. "The Export Demand for Maize From Brasil." Revista de Economia Rural (Brasil) 16(4) (1978). Thompson, R. L., and Schuh, G. E. "Trade Policy and Exports: The Case of Corn in Brazil." Lafayette, Ind.: Department of Agri- cultural Economics, Purdue University, 1977. (Mimeographed.) Valdes, A. “Trade Policy and Its Effect on the External Agricultural Trade of Chile, 1945—1965." AJAE 55 (Jay 1973): 154-64. Veiga, Alberto. "The Impact of Trade Policy on Brazilian Agricul- ture, 1947—1967.” Ph.D. dissertation, Purdue University, 1974. Vilas, A. T. "A Spatial Equilibrium Analysis of the Rice Economy in Brazil.“ Ph.D. dissertation, Purdue University, 1975. Williams, G. W. "Economic Structure of the Brazilian Soybean Industry: A Prototype Model.” M.S. thesis, Purdue University, 1977. Zockun, M. H., et al. A Agricultura e a Politica Comercial Brasileira. Série IPE/Monografias, V011 8. $50 Paulo: IPE/USP, 1976. “‘WWWm