V E!!! .!E .El . !!!.!W1"E‘EE..!!E!E !!!!!E.E!!!!E!!-E!! ! . :E‘E!!!!! Em!!! -. E! Eh! E!!! E! .. E!!!" E. EEEEE! E..!E!!E!E IE.” TEEEEEE! E“. EEEEEEEEEE EEEEE! EEJEI E!EEEEEEE!!!EEEEIEEEEEEEEEEEEEEWQEEgEEEEEEEEEEEEEEEEW ‘. . E! E EE 1!! '1 ‘E! E1! 1‘ EEE EEEEEEEEE'EE’EEEEEEEE!"1 E... E EEEEEE .. E EEE‘ }i!!! 1!! E1 ‘EEEEEE E!!! E! E!!!EE WE E!!! 1.! EEEEEEEEE EIEEEEEEEEEE E! E! E!!! E! !It E!!! E! EEE. E!!! E1 ! E!!! E E ', E!!! E. 1!"!!! . EEEEEEE ’o E!!! E! 1 E“: E”. * EEE‘ IEEEEEEEEE!E.E:EIE_EEEEEEE!EE!E1!!EEE!!E1 !! !! E !!E W o O . . ) h EE! l!!! E! E! E! E E... E E EE . EEE I E '. E. I. E! E. E!!! -H—w 3H1 M NEE”? i ’EEEE E!‘ 1!! ‘EEE EEEEEEEE! E! 1!. 'EEEEEE'EEE 1 :EEEEE 1!! ”MEN. i... EAEEE! EEEEEFE EEEE h u... ‘ . - .— - - '~‘ M‘ o "-3- _ww " .4 h... =._____,w wm—I-H— wfi WW .EE 332:1?" fwkm‘ .. EE . . I.“ “EVE”!!! :EE!!!’ E! EEEEEEEEE ! WEE“! i!!! 1"! E“!!! E”. EEEEV a!!! -E EEM E... E!!! ! E!!! E .E’EEEE ""E 1511.!!! EEE““1E..EEE!E“. E! E: 1.1 .1!!! E!!! W!!!“ Ema! WW. EEEEEE I‘EE! *1! E"! 'E'!!! !EEEE!E‘!”E "'- .3! NE! W . 4!)!!! EE!!*...!!'E'1E.E EEE E!!!“ E!!! 1.! E! WEE“ ‘1 "E E E- !!! !E.! E E 5E .' WEE!!! E~1!E!!!!.!!“’! E}!!! H“!!! “3.331%?!“ at!!! EN ‘11!!! P!:'~. WW!E.‘~:"3.J;3!'!'EEF-'!!!!E!!7 E“ '. M!!! Mum H! E ME: 1."!!! E! M!!! E m....EEEEEEE;E!EE;EEE.:3.:iEE‘E-EnggE 1-123! 3". , t . ~ 2‘5: . . , . a! FELL!!! ;. 1‘“! - .’~ 1: ,z‘, BEE...” i!!!" E! EEEEE'EEEEEEEEE !3!EE""”EEE ...EE.EE!‘!E!!1JEEE. E‘E “E. E! . 3'7.!!‘~'.~.E.-...~!-.;EEEE' HEJE' . .' E E!!!1 WE. E52; ENE... ELM UK! E!!! “I! m.” E“!!! E! .n. EN. 'E' 1!! IE!!! ' 'E! _ if! I!!!EEEE!EE . RE!!! 5!! 4E!!! “.9 IE3; H‘!!!. ‘9!!! 2‘ E! E E !!'E iii!!!” V!!!- . 2!“! "I p .1 .t ' E.’_ H E J Pf!“ -!!E ‘3' i! .1'.~1-.'E"!E5"!!EE !‘ 7.3!! . EE. - " EEE.E‘!:.!E!E! ' Em! EEEE'EE! E!!! ”E“!!! W §E1E!"'E'EE=E-!!Erzt::evEEEEE1."!!! .i - t ! . .: EHEEEEKEEEEEEEE ,‘31.‘ 4 H!!!“ EL!!! E!!! “E!!! .EEE‘ EH. 1 .EEE EEE?!E!1.I‘E!!!!!!!13!"EI! ELIE!!!) .1; . E! *1 . "'EIEEEEEIEEEE'EEEEEI E. .E P .15.. . E! NW 1'1!!! .w .11.! ‘jEE'iEEEE‘EETE..‘EEEEEEEEE‘ ' . E‘. . . EEEEEE!!!E=! E, EEEEM 4! 3!:. ”.9! E. E)! E! ’E‘IEEE EEE. EEE!,.!!.E,‘,!‘..iE“E_;.IE,E1.E;;'.'",‘§=::! - ..EEE1:!E!EEEEEEE E1 E!” E”. EEE. E!!- E ‘1' ‘ E'EEEE.EI;T'21':1E'="1'31“?! . v 1...”! EEE .1!!! *' 11 ""1"“! '11.!!!“ EE . 1...”. EE . E!!! E r!E!EE NW N EE! E! H EEEEEEE!!!!E! EEE .1le E! !!EE 1...! 'E E'H"”'!'I!!'E'E1l!g E! .!E}EE!I! UM“! E1!!!“ E... E!!! EEK!!! EEEEEEE'E '. IE. EEEE'E '1!!! .E! WEE!!! . !!!!E-r.:,1E!!!!. EEEE!!E' Q. l' 11; E! a... I. 5!! NE!!! !! EE'i'E' lEE'E!"EE !EE! .‘E' E!!! EE‘E E!.! I'Er ”WEE?! ENE! !!:! EE!EE,E' MEEEEE E! EEE!EE ‘H! EE'EE 'E'.11;i E.1EEE‘111E ~. EEEE . . . EL”!!! ‘ | t'l "E!!! !:.E '.II -!!!':!: “E!!! E‘EEEELEE‘!’ ‘!5 EEE EETEEEEE’ELTflEz’EE 4!!" u!!! EEEEEE EHEERNEEEEEEEPE L!” E! HEEEEEE! EEE EEE E! E 1' EEE E EE! E!!! "‘E E-! E? E.‘ E!!! . HE"! WE 1.1:”... EEE ' E'E... .E'! E!!! "3! r! 14.".1'“?! .‘ .'!! E!’ EE'E E i" r!!! «HE!!!» E!!! EnH E! 4!}!!! W! (0."!!! 7!!!! E!!! 1! E!!! ‘.E!!!!! E! ! EEEEEEE‘EMHEEEEE'! ENE E... E!!! E‘s!EE,'! .EEE EE! . E!!! E!!! E2!!! EEEEEE!EEEEE..EEEE.!E “E! ‘E!!! ”I!!! ,EEE..E!,EE EEEEE E!!!" ”E! SIEEEEEMI E 'EH‘MEJ E ’7! EEIEEEH EE.-LEEEE!E .rEfE’JEEEIIE, EEE'E'!‘ !. EEH EEE‘E E!!! W!!! E!!! E!!! E!!! EEEEE. EEE IEEEHEEEEEE E1] .!! . E!!! Equ 1E !EE!E. EE 1!! NH. ..,~...E!EE!‘!EE.EEIEE. EEE. .EEIEE E!!! .E... EEEE‘. E‘E!E.!EEWE' . E! E E! .!!E E!!! NEE. ,1. EE ”WEE! E!..E EEEE ”E!!! E1.» ' . 1!!!! !!!! E“!!! E! EEE "EE .1!!! EE' EEE!!! E! ..E! El" -. E!!! E E1"! E"’ E... 111 a!!! 'EEEEm!!! ! E! E 11!!!! E! E. EEEE E'EEE . EEE J2, !!1!!!!!!.!!!!!E!E!:.E'.!IE'E l LlfiRARY Michigan State University This is to certify that the thesis entitled Measuring Aggregate Elasticities with a Multi-Commodity World Trade Model presented by Loreen Marie De Geus has been accepted towards fulfillment of the requirements for Master of Science degree in Agricultural Economics DateW’7 0—7 639 MS U is an Affirmative Action/Equal Opportunity Institution 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. Measuring Aggregate Elasticities with a Multi-Commodity World Trade Model By Loreen Harie De Geus A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Agricultural Economics 1987 ABSTRACT MEASURING AGGREGATE ELASTICITIES WITH A MULTI-COMMODITY WORLD TRADE MODEL BY Loreen Marie De Geus The effects of reducing grain prices in order to increase exports of 0.8. agricultural products and reduce burdensome government stocks are measured in this study. The Michigan State University Agricul- ture Model, an annual, eleven-region simultaneous equation model is used to measure the price elasticity of export demand for 0.8. wheat, feedgrain and soybeans. The prices of the three commodities are changed proportionally and simultaneously to capture an aggregate effect. Farmplevel supply elasticities are calculated using change in harvested area to a supply variable and gross revenue per hectare as expected price. Export supply and import demand elasticities are calculated for regions other than the 0.8. Farm level supply elasti- cities are low for most regions. Alternative specifications of revenue are explored. Price elasticity of export demand for 0.8. exports is found to be very inelastic. ACKNOWLEDGEMENTS A thesis is always the work of many individuals and I am indebted to many people for the final product. Without the guidance and encourage- ment of my major professor, Roy Black, this project would not have been possible. II am also thankful for the valuable input of my committee members, James Hilker and Daniel Suits, who advised me during the writing of the paper. For their friendly patience and thoroughness in tracking down inter- national data, I appreciate the assistance of Henry Haszler and Ed Allen. Special thanks are due to Shayla Shagam, who put considerable effort into rebuilding the Ag Model and who taught me how to use it. I also owe many thanks and favors to Nayne Whitman who assisted me time and again in data management and logistics. Harm thanks and appreciation go to my family for support and encour- agement throughout the degree process and for providing the foundations that enabled me to pursue an advanced degree in the first place. Finally, I owe a special debt of thanks to my husband, Robert, for two years of patience, support, editing and housework while earning his second Masters Degree vicariously with me in Agricultural Economics. 11 TABLE OF CONTENTS Page Chapter I. IntrOductionOOOO...0.0.00...OOOOOOOOOOOOOOOOOOIO0.00.0001 1.1 Objectives or the StUdyOOOOOOOIOOO0.0.0.0.000...0.0.0.00002 1.2 Organization of the Thesis................................3 Chapter II. Theory and Literature Review...........................u 2.1 Price Elasticity of Export Demand.........................A 2.11 Factors Affecting Price Elasticity of Export Demand..6 2.12 Methods of Price Elasticity of Export Demand Estimationo.OOICOOOOOOOOOOOO0.00.00...0.0.0.0000000007 2.13 Price Transm1881on0OOOOOIIOIOO..0.00.0.000000000000008 2.2 Aggregate Response...COO...IOOIOOIOOOOOOOOOOO0.0.0000000009 2.3 Policy Relevance of Price Elasticity of Export Demand....12 Chapter III. Measuring Supply Responses...........................15 3.1 The A8 MOdeIOIOOOOO00.000.00.00....OOOOOOOOOOOIOOOO00.00.15 3.2 Measuring Supply E1asticities............................19 3.3 Aggregate Response.......................................26 3.31 Measuring Aggregate Supply Responses................26 3.32 Sources of Low Correlation..........................28 3.33 Relationship of Cropland Base to Gross Revenue......33 3.3A Relationship of Internal Prices to Border Prices....38 Chapter IV. Country-Level Investigation...........................A5 “.1AustraliaOOOOIIOOOOOOOOCO0.00.0...OOOOOOOOOOOOOOOOO...00.0u6 ”.11InternalPriceSOOIOIOIOOI...I0.000IICIOOIOIIOOOOOOOOué b.12 Gross Revenue vs. Net Revenue.......................n8 ”.13 competing EnterprisesolOIOIIOOOOIOIIIOOOIOOOOOOOOIIOH8 "01“summary.O....0...OIOOIOOOOOOOOOOOOOOOOOOIOIOOIOOIOIIng 1‘02 ArgentinaOOCIOOIO00......0.......00.00.000.00.00000000000051 “.21InternalPriceSOO0.00.0.0...O...0.0.0.0000000000000051 iii TABLE OF CONTENTS (cont'd) Page ”.22 conlpetins Enterprises.00....0..00.0.000000000000000052 1$.23 surlmarL’OOOOOOOOOOOO...O...OOOOOOOOOOO00.0.000000000053 1‘03:'1ultiCOllineal‘ityOO...O....0.0.00.00.00.00...OOOOOOOOOOOOOSH CO Chapter V. Elasticities from Model Simulation......................5 03 5.1 Effect of the 0.8. Loan Rate on Prices...................S 5.2 Aggregate Price ElastiCitieS.O...0.0.0.0.000000000000000061 '5 f" 50fi1ReSUItSeeeeeeeeeeeeeeeeeeeeeeseeooeeeeeeeeeeeeeeeeeeé') I" 5.22 DiSCUSSionO0.0000000000000000.oeeeeeeoeeeoe ......... 7.) 5.3 Single Commodity Price Elasticities... ...... ...... ..... ..76 5.31ReSUItSeeeeeeeeeeeoeoeeeeeeeeeeeeeeeeeeoeeeeeeeeeeee76 5‘32 DiSCUSSionIOOOOIOOOOOOOOOO0.0.0....0.0.0.0000000000085 Chapter VI. Summary and Conclusions...............................89 6.1 Summary of Results........................................89 6.2 Conclusions...............................................91 Appendices 1. Ag Model Regional Groupings................................95 2. Price Elasticities of Import Demand........................96 3. Cropland Base Estimation, tatistical Results..............97 3. Raw data from Sources Other than the Ag Model.............10h 5. Nominal Loan Rates........................................107 6. Argentine Harvested Area Estimation, Statistical Results..108 7. Usefulness of the Ag Model for Policy Analysis............111 BibliograthOOOOOO...0.0....0.0......0..0.00.0.0.00000000000000000116 iv Table Page 1. Revenue Elasticities of Supply at Farm Level...................24 2. Percent Change in Price in Response to a 20 Percent Change in Wheat and Feedgrain Loan Rates......................59 3. Price Elasticities for Exporters - Aggregate Run...............o7 A. Revenue Elasticities of Supply at Farm Level -Aggregate RunOOOOIOOOOO00.0.0000...O...OI0.00.00.00.0000000069 5. Price Elasticities of Demand for Importers -Aggregate min...CCOOOOOOOOOOOOOOOOOOI.0.0.0.000000000000000072 6. Market Shares (i) - Aggregate Run..............................7M 7. Price Elasticities for Exporters - Single Commodity Run.......77 8. Revenue Elasticities of Supply at the Farm Level -Single commOdity RunOOCOOCOOIOIOOOOOOOCCOO00.......00000000079 9. Price Elasticities of Demand for Importers - Single CommOdity Runeeooeeeeeeeoeeeeeeoeeoooeoeeee00.000.00.82 10. Market Shares (i) - Single Commodity Run.......................84 11. Price Elasticities of Export Demand for 0.3. Wheat.............86 12. Comparison of Price Elasticities of Supply -IIASAFIOde1 and AgliodeIOOOOOOOOOOOOO...00.0.00000000000000088 13. Ag Model Regional Groupings....................................95 13. Price Elasticities of Import Demand............................96 15. Argentine Price Data..........................................10u 16. Brazilian Price Data..........................................105 17. Australian Price Data.........................................106 18. Nominal Loan RateSOOOOOOOOOOOOOOOO...0.00000IIOOOOOOOO0.00.00.107 Figure 10. 11. 12. 13. 111. 15. 16. 17. 18. 19. Cropland Cropland Cropland Cropland Cropland Cropland Cropland Cropland Cropland Cropland Cropland Cropland Weighted Base Base Base Base LIST OF FIGURES Page Estimations"'A‘r‘gentinaooooeeases-00.000000000000029 Estimations-BraZilOOO...OOOOOOOOOOOOC0.0.0.0....29 Estimations‘-Australiaoeeeeeeeeoeo0.000000000000030 Estimations-canadaOOOOOOOOOOOOO0.0.00.000000000030 Estimations - United States.......................31 Estimations - Developed Harkets...................31 vs. Gross vs. Gross vs. Gross vs. Gross vs. Gross vs. Gross Revenue Revenue Revenue Revenue Revenue Revenue Argentina.....................35 Brazil........................35‘ Australia.....................36 Canada........................36 United States.................37 Developed Markets.............37 Avg. Revenue, Border Prices vs. Internal Prices - Argentina.000......000......OOIOOIOOOOOOOOOOOOOOOOOOO'Cono Weighted Avg. Revenue, Border Prices vs. Internal Prices - BraZileeoeeeeeeeeoeeeeeeeeeeeeeeeeeeeeeeeeeeeeoeeeeeeee0,40 Weighted Avg. Revenue, Border Prices vs. Internal Prices - Australia...0.0.0.0...00.000000000000000.0.00.00.00.0000u1 Weighted Avg. Revenue, Border Prices vs. Internal Prices -Developed PiarketSOOOOOOOOOOOOOOOOOOOOOOIOOOOOOOOOOOOO0.0&1 Real ComDOdity Prices-ArgentinaCCCOI.OO..OOOOOOOOCOCOOOO0.0.0.55 Real Commodity Prices - United States...........................55 Base and Scenario Revenues for One Commodity, 1986 - 1996.......62 vi LIST OF FIGURES Figure Page 20. Supply of Exports by Countries Facing Infinitely Elastic DemandOOOOOOOOOOOOOOOIOOOOOOOOOOOOOOOOO.0.00.00.00614 21. Demand for U.S. Exports.........................................6u 22. Change in Net Import Demand...................... ...............65 vii CHAPTER I. INTRODUCTION In order to reduce the burdensome level of grain stocks held by the government, the Food Security Act of 1985 reduced the loan rate, and hence the price, of major crop commodities. The reduced price was expected to increase U.S. exports of grains and soybeans and thus remove excess supplies of these crOps from the U.S. market. The expectation that exports would increase was based on the belief that the net effect of a decrease in price would lead to a greater than proportional increase in demand for these crops, or that the price elasticity of export demand (PEXD) was greater than one. In that case, gross revenue from commodity sales would rise with a decrease in price. The question of whether PEXD is less than, equal to, or greater than one can be investigated empirically by measuring the change in U.S. exports in response to a change in price with a mathematical model of agricultural trade. Frequently, the effects of price shocks are simulated on a commodity—by-commodity basis. However, the change legislated by the Food Security Act is a simultaneous reduction in the prices of all program crops. Since a simultaneous price change reduces the degree of substitution between commodities, the PEXD for a change in prices for all three commodity groups would reasonably be smaller (closer to zero) than for single-price changes. The smaller the PEXD, the less effective the new price policy in reducing excess stocks. 2 A model of agricultural trade may consist of a single demand equation for exports or of hundreds of equations that interact to determine exports and imports from internal supply and consumption equations. One such large model is the Michigan State University Agriculture Model (Ag Model). It is an annual, multi-region, multicommodity, long-range forecasting model. Its scope includes three commodities important to U.S. agriculture: wheat, feedgrains1 and soybean products. The Ag Model does not include other crops that may substitute for these three on the farm, such as cotton, or in the world market, such as rice and other oilseeds, nor are effects on livestock considered. While some simplifications are necessary to model the complex world grain and soybean market, the structure of the Ag Model does allow analysis of several crOps at once, as opposed to the single- commodity approach. 1.1 OBJECTIVES OF THE STUDY This study uses the Michigan State University Agriculture Model (Ag Model) to measure the effects of reducing the loan rates of program commodities on world prices, trade and U.S. ending stocks and exports in particular. Attention is paid to the aggregate effect of a proportional price change on all three commodities as a group. The study calculates the short and long run supply elasticities of wheat, feedgrain and soybeans and measures PEXD using the Ag Model. Components of the PEXD are also considered, specifically the supply response of competing exporters at the farm level and the effect of imperfect price transmission between international and domestic markets. 1Feedgrains include corn, sorghum, oats, barley and millet. 1.2 ORGANIZATION OF THE THESIS In the following chapters, the measurement of PEXD and other elasticities is explained, executed and analysed. Definitions of the economic concepts used and a review of the literature regarding aggregate elasticities appear in Chapter 2. Chapter 3 deals with farm- level supply response and the choice of variables. Chapter A contains a more in-depth study of two of the Ag Model's eleven regions. Finally, in Chapter 5, the Ag Model is solved to calculate export supply, export demand, import demand and harvested area elasticities. A summary of the results of the three preceding Chapters and conclusions are presented in Chapter 6. CHAPTER II. THEORY AND LITERATURE REVIEW The economic measures used in this study to evaluate the responses of producers to price are defined in Chapter Two. Previous theoretical and empirical investigations are reviewed. In addition, the policy context that makes the question of producer price response relevant is described. 2.1 PRICE ELASTICITY OF EXPORT DEMAND (PEXD) PEXD is the change in a country's exports of a commodity resulting from a one percent change in the price of that commodity. When PEXD is measured by a single equation, all other variables are held constant. When this value is determined by a system of equations that allows all variables to adjust, it is closer to an impact multiplier than a true elasticity because other endogenous variables are allowed to adjust to change in price (Gardiner and Dixit, 1987). An impact multiplier is the coefficient on a predetermined variable in a reduced form equation in a multi-equation system (Kmenta, 1986). When the relationship is measured over time, it is called an n-period impact multiplier. However, the elasticities reported in this study are expressed as percent changes, which are unitless, rather than simply as coefficients. Therefore, elasticity will be used in this discussion to describe both the single- equation measure reported for some studies and the multi-equation measure resulting from the Ag Model and from other studies. 5 In the short run, PEXD typically accounts for adjustments in net demand for exports, which include changes in excess supply in competing countries, but does not account for changes in output. Agricultural output cannot adjust immediately to price changes because crops are produced only once a year. In the long run, one year and over, PEXD also incorporates changes in production, in government policies, and in macroeconomic factors, such as exchange rates. However, prices of other goods and the supply of and demand for them, tastes and preferences, income, technology and population are all assumed constant when determining PEXD (Gardiner and Dixit, 1987). Knowledge of PEXD is of critical importance to policy makers in that it determines whether revenue will increase or decrease as a result of a price change. If the absolute value of PEXD is greater than one, it is elastic, meaning that a drOp in price will lead to a greater-than- preportional increase in the quantity exported and revenue is increased by decreasing price. Conversely, when the absolute value of PEXD is less than one, it is inelastic and the quantity exported will respond less than proportionally to the price change. Revenue is maximized in this case by increasing price. In the case of unitary elasticity, the absolute value of PEXD is equal to one, price and quantity change proportionally and revenue is unaffected by price changes (Gardiner and Dixit, 1987). In general, the lower the price, the greater the quantity demanded, thus a negative sign is expected for demand elasticities. In contrast, the higher the price, the more is supplied, yielding a positive sign on supply elasticities. Because PEXD is a net demand elasticity, a negative sign is expected. 2.11 Factors Affecting PEXD Several factors influence elasticity (Gardiner and Dixit, 1987). One is the availability of substitutes, a second is the share of the total budget that the product holds and a third is the extent to which the product is considered a luxury versus a necessity. In general, the elasticity will tend to be higher the greater the number of substitutes, the smaller the budget share it claims, and the less necessary it is. Demand for food as an aggregate is generally considered to be inelastic because it is a basic necessity and because there are no substitutes. For an individual commodity, such as wheat, however, demand will tend to be more elastic, because other substitutes exist. The greater the substitutability, the greater the elasticity is likely to be. In the case of a country that is small relative to the world market, the quantity of exports from that particular country will not be sufficient to perceptibly affect world price. Since many alternative suppliers exist, the country theoretically faces a perfectly elastic demand curve. In the case of the U.S., it is a large country relative to share of world exports. The quantities exported by the U.S. (38 percent of net world trade (excluding intra-regional trade) in wheat, 73 percent of feedgrain trade, and 87 percent of soybean trade in 198A) have a significant impact on world price. Because U.S. share is large and the quantity available from competitors is relatively small, one would expect PEXD to be less elastic for the U.S. than for a country that held a smaller market share. Government interventions in the markets for agricultural commodities also affect elasticities. For example, price supports, set- aside programs and export subsidies restrict the responses of market 7 participants by insulating them from market variations. An elasticity measured from such a constrained market is likely to be considerably smaller (more inelastic) than a value that would be observed in a completely free market (Peterson, 1981). 2.12 Methods of PEXD Estimation Gardiner and Dixit (1987) surveyed forty five studies that estimate PEXD for U.S. wheat, feedgrains, soybeans and other crops. The esti- mated short run elasticities range from -.1u to -3.3 for wheat, -.30 to -1.51 for feedgrains, and -.1u to -2.00 for soybean products. Differing sample periods and assumptions about the existence of free trade or government interventions may in part explain the large range of empirical values observed. From these studies, no clear consensus emerges as to the actual values of these elasticities, nor even whether they are elastic or inelastic. Common techniques for measuring PEXD include a) direct estimation of one single equation for the excess demand of the rest of the world; b) calculation of PEXD by summing net supply and demand for all countries; 0) simulation of the entire international market for a commodity or group of commodities; and d) synthetic methods which simulate the market using elasticities obtained from other models. The greater the complexity of the modelling effort, the more likely that the model accounts for the many variables that determine PEXD. These include supply and demand shifters for all countries, governmental actions and international agreements, and financial factors such as exchange rates and foreign exchange reserves. (Gardiner and Dixit, 1987) 8 Tweeten (1967) characterizes PEXD as: n PEXD =:[E‘E*Q£1_E*E*gii_] 5:343 US i=1 di pi Oef si pi Oef pi BPworld where: E and E = domestic price elasticities of supply and demand in 31 di country i Q and Q = quantities supplied and demanded in country i 31 di 0 = U.S. farm exports ef E = Price transmission elasticity for country i pi This is a calculation method of determining PEXD, the percent change in U.S. exports resulting from a percentage change in the commodity's own price. In Tweeten's, and many other studies (Johnson, 1977, Gardiner & D1X1t’ 1987)! Epi is assumed to be one. That is, world prices transmit perfectly to internal prices in each country. 2.13 Price Transmission Price transmission is the degree to which world price fluctuations are passed across a country's border to it’s internal market. Complete transmission would occur if price changes passed immediately between markets. However, many countries are observed to insulate their internal markets from the price variability of the world market (Bredahl, Meyers, and Collins 1979). In these cases, price transmission is low. The Common Market has an explicit variable levy that precisely taxes away the difference between the world price and the supported internal price. In other regions, the insulation is less obvious and less absolute. 9 Many studies have shown evidence of insulation to some degree. Bredahl and Green (1983) tested statistically for "causal" relationships between harvested crop areas and world prices and found no significant relationship for large exporters of coarse grains other than the U.S. Causality was only significant between exports and world price of coarse grains for France and the U.S. Bolling (1986) calculated price transmission elasticities in western hemisphere countries and reported values ranging from .07 (wheat in Mexico) to 1.0 for crops of interest. Low price transmission elasticities represent significant insulation from the world market and partially account for inelastic supply responses to world price. 2.2 AGGREGATE RESPONSE A variation of the normal elasticity calculation is an impact multiplier that is measured by estimating changes in quantities that result from changing prices, but holding relative prices between the commodities being measured constant. If wheat, feedgrain and soybean prices are changed proportionally, substitution effects between crops are eliminated. The resulting elasticity reflects only an aggregate response on both the demand and the supply side to changes in price. This aggregate elasticity would logically be expected to be less elastic than a single commodity elasticity. On the demand side, aggregate elasticity is reduced because the consumer has fewer alternatives than when an individual commodity price is changed while other prices are held constant. By the same token, supply is less elastic because substitution between outputs is eliminated. 10 Not only is aggregate elasticity less elastic than a single commodity elasticity, but the characteristics of agricultural supply and demand increase the likelihood that this elasticity would be low. In general, aggregate agricultural supply tends to be very inelastic because specialized inputs such as farmland and farm machinery often have no alternative use that would yield an income comparable to agriculture. If the individual farmer decides to reduce his or her hectarage, the land in question is usually rented or sold to another farmer who keeps the land in production. In other words, farm input supply is highly inelastic. Hectarage remains relatively constant in the face of price fluctuations. Although cropland area trends upward as new areas are cleared and as increasingly marginal land is brought into production when prices are favorable. Cochrane (1958) has suggested that agricultural supply may be characterized by irreversibility. In other words, producers respond more strongly to increases in price than they do to decreases. For a given year, he maintains that agricultural supply in the aggregate is almost completely inelastic. When farm prices are high, farmers adopt new technologies in order to reduce their production costs. Once adopted, new technology increases output per unit of land or labor. Although the supply curve is steep, the quantity supplied expands because the curve is shifting to the right. In times of low prices for agriculture, new investments are not made and the quantity supplied changes very little, since the supply curve is inelastic. This differential response to price changes causes an irreversibility in agricultural product supply. While farmers will respond to increased 11 prices with increased production, a drop in price has little effect on the quantity supplied. Cochrane described the cycle of continually increasing agricultural productivity as a treadmill (1958). Farming is an atomistic industry where the individual normally is not able to alter price. Early adopters of new technology benefit by reducing their costs, which they can control, but the technological changes do not affect price initially because the adopters are few in number. The price of the crop falls as more and more farmers' costs are reduced and the average farmer's profit margin is squeezed. Average and lagging farmers are then forced to adopt the technology in order to compete. As a significant number of farmers adopt the technology, cost, hence price, falls for a given quantity, or production expands at a given price. In other words, the supply curve shifts to the right. Farm incomes decline as crop prices come to reflect the lower costs of production and excess profits to early adopters are eliminated. As newer still technologies are invented, farmers innovate to capture the cost savings and the cycle repeats itself. Cochrane's focus is not simply on measurement of elasticities, but on how the production possibilities frontier changes. There are few empirical estimates of aggregate elasticities in the literature. Tweeten (1967) attempted to estimate the elasticity of demand for all U.S. farm output. He obtained a farm-level elasticity of -O.55 in the intermediate term (3-A years) and -1.1 in the long run. Tweeten listed sources of possible error in his estimates but the net effect of these potential errors is impossible to discern. However, his elasticities were weighted sums of elasticities estimated for categories of consumption for food or for individual food commodities and cotton 12 and tobacco. Further, be based his estimation on free trade and attempted to correct for institutional barriers to it rather than measure an elasticity based on the current set of political institutions. Cochrane (1958) estimated the elasticity of domestic demand for food at retail in the twentieth century. The short run elasticity given was approximately -0.16 for the period 1950-55, the most recent time period estimated. He suggested (1965) that the farm-level elasticity may be slightly more than half the retail figure. Other studies of domestic aggregate demand published in the early 1960's (Brandow, 1961, Burk, 1961, Waugh, 196A) gave low farm level elasticities of less than - 0.2. Buiding on his recent study of domestic food demand, Huang (1985) has calculated an aggregate demand elasticity for food in the U.S. His estimate of -.13 at retail may be considered an upper bound on the farmgate elasticity since demand for food products at the farm level is less elastic than retail demand. This study is based on 1953-83 data and suggests that domestic price elasticity of demand for farm products is very low. 2.3 POLICY RELEVANCE OF PRICE ELASTICITY OF EXPORT DEMAND Because PEXD is a succinct measure of the effects of price changes on revenue, it is an important variable to policymakers. The crucial question - whether it is elastic or inelastic - shapes the choice of pricing policies. A current issue of concern in the 0.8. is the high levels of grain stocks that have accumulated as a result of high domestic support prices. If the PEXD is elastic for wheat and feedgrains, these stocks can be decreased by lowering the prices of 13 these grains. The reverse will occur, stocks will pile up when prices are lowered, if in fact the PEXD's are inelastic. These excess stocks exist because farm prices have been supported by the government above world market equilibrium levels in order to protect domestic farm incomes. Profitable returns stimulate production by attracting more resources into agriculture and by encouraging innovation which raises the productivity of resources already invested in agriculture. In addition, at high prices, the quantity demanded is below the equilibrium quantity and even farther below the amount supplied, hence stocks accumulate. While some level of stocks is desirable to stabilize the market in short crop years, consistent overproduction has led to levels of stocks that greatly exceed the desired quantity and are costly to maintain. Because domestic demand for grains is extremely inelastic (Cochrane, 1958; Tweeten, 1967), attention turns to the export market. The more elastic demand is, the less price must fall in order to bring supply and demand into balance. The low elasticities of domestic supply and demand indicate that there must be a sharp decline in prices in order to reach equilibrium. Cochrane (1965) estimated that in the early 1960's agricultural prices would need to fall by as much as no percent to achieve balance. Such a drop, he maintained, would reduce net farm income in the aggregate by as much as 60 to 70 percent. If, on the other hand, the demand for exports is elastic, gross returns would rise because the expanded quantity demanded would more than offset the decrease in price. Lowering the price of grains, therefore, is based on the expectation that PEXD exceeds one. Many economists support this notion at least for individual commodities 1A (Gardiner and Dixit, 1987). For example, Schuh (1983) has argued that demand for exports should be highly elastic because importers are buying quantities of grain that are small relative to their domestic production. Tweeten's (1967) estimate of aggregate demand for farm output averaged inelastic domestic demand with a very elastic export demand of -6.u2. More recent estimates of individual commodity prices range up to -10.2 for coarse grain in the long run (Johnson, 1977). The actual value of PEXD for U.S. agricultural commodities is unknown because it is difficult to observe in isolation, because the value changes over time with new developments in technological innovation and governmental policies and because the actual situation with its many market imperfections differs considerably from the theoretical case. The latter reason in particular may be an argument that highly elastic estimates of PEXD are biased upward. Many of the studies cited assume free trade or make simplifying assumptions that reduce the insulating effects of government policies. The elasticities of supply at the farm level in other regions of the world affect PEXD as do demand responses. The same factors that make farm supply inelastic in the U.S. would be expected to apply in these countries as well. If PEXD is in fact inelastic, lower agricultural prices do not allieviate the problem of excess stocks. More importantly, the drop in price is not offset by increased quantity and farmers could suffer significant loss of income. Thus the impact of decisions based on this simple calculation are of considerable importance to agriculture. CHAPTER III. MEASURIfiG SUPPLY FESPOHSES This chapter is focussed on the measurement of elasticities implied by the Ag Model structure, particularly supply responses. First, the Ag Model itself is briefly described.1 Then, supply response is described and measured. Special attention is paid to the aggregate response of hectarage to revenue changes and to how closely these two variables approximate the theoretical relationship measured by a supply elasticity. 3.1 THE AG MODEL The Ag Model was constructed to simulate international trade of wheat, feedgrains and soybean products - beans, meal and oil - by dividing the world into eleven regions. The U.S., Argentina, Brazil, Australia, Canada, and China are each modelled separately. The remainder of the world is divided into five economic regions. The Developed Markets include Western EurOpe, Japan and South Africa; the Soviet Bloc is composed of Eastern Europe and the USSR; the Oil- Exporting Low Income Countries are Organization of Petroleum Exporting Countries, excluding Gabon and Qatar, plus Oman; the Newly Industrialized Countries are a small group of rapid-growth nations - 1Shagam (1986) describes the structure and statistical validity of the Ag Model in detail. 15 16 Hong Kong, Singapore, Taiwan, Malaysia and South Korea; and the Low Income Countries include the rest of the world (See Appendix 1). Within each region, the equations are arranged to solve sequentially for domestic supply and demand. The net import and export equations interact with those of other regions of the model and with price to determine price and quantities simultaneously. Yield is determined exogenously as a function of trend and harvested area is based on lagged harvested area and lagged revenues per hectare. From these estimates, production is calculated as an identity (harvested area times yield). Consumption is estimated for exporters from price and income and for importers, consumption is calculated as a residual. Net imports are estimated from income and price or policy variables. Ending stocks are a function of domestic production and consumption, net imports where applicable and policy variables for importers and exporters other than the U.S. and Canada (where it is calculated as a residual). Net exports for Canada are a function of residual demand from the rest of the world and are calculated as a residual the U.S. and for other exporters. Except for price, each equation contains only predetermined variables. The Ag Model is structured on the concept of the U.S. as a residual supplier of grain and soybeans to the world. In practice, the U.S. domestic price, supported for domestic farm policy reasons, sets a floor for the world price because of the large volume of grain stocks and world exports controlled by the U.S. (HacGregor and Kulshreshtha, 1980). Other exporting countries are able to price slightly below the U.S. and export most or all of their excess stocks. When this supply is 17 exhausted, importers turn to the U.S. to satisfy the remainder of their needs (McCalla, 1966; Bredahl and Green, 1983). The Ag Model formulation approximates this market structure by assuming that competing exporters other than Canada are "surplus exporters." These "surplus exporters" do not hold large stocks, but instead export all surplus production at or slightly below the world price. They do not necessarily subsidize exports substantially below world price. Canada is considered to be a "contingent surplus exporter" that competes in an oligopolistic way with the U.S. for the residual pool of import demand. The remaining unsatisfied import demand is filled by the U.S. (McCalla, 1966). If the residual supplier structure is correct, the supply response of competing exporters must be extremely inelastic with respect to world price (Bredahl and Green, 1983). Inelastic price response would be characteristic of countries that only hold enough ending stocks to satisfy domestic needs. These countries would export any excess supply at whatever price is necessary in order to dispose of the stocks and the U.S. would hold all excess stocks for the world. On the other hand, if exporting countries are price elastic, they would hold stocks for speculative reasons when prices are low and sell them when world market prices are higher. Which countries are holding the residual stocks would be indeterminate because the stocks could be spread amongst all the exporters evenly or one or more countries could hold all excess stocks. The case where competing exporters are inelastic to price is the simpler case to model because it ignores the specific distribution of stocks amongst countries and assumes that the U.S. holds all excess 18 stocks. This structure, with modifications for Canada to hold some stocks, is chosen because it approximates the world market system more closely than a purely competitive model (McCalla, 1966; Bredahl and Green, 1983). Support for the assumption of inelastic surplus exporters can be found in Australia, for example, where domestic consumption is small relative to exports. Excess stocks resulting from changes in price cannot be absorbed in the domestic market. Therefore, Australia relies on the international market to adjust stocks (Goodloe, 1984). Limited storage capacity further encourages the Australian Wheat Board to dispose of as much grain as possible each year (Spriggs, 1978). The U.S. Gulf price is considered to be the world price in the Ag Model. Not only does the U.S. occupy a large share of the world market, the U.S. market is relatively open. U.S. prices are therfore used as the basis for pricing decisions in other countries. (MacGregor and Kulshreshtha; 1980, McCalla, 1966; Spriggs, 1978). According to Gilmour and Fawcett (1986). "Wheat prices in the United States establish the competitive standard for most wheat entering world trade. Their visible and competitive pricing process provides a convenient branchmark (sic) from which other exporters can establish their export prices." Prices for each region are handled by converting the world price, as defined by the U.S. price, to a border price. Border price is obtained by converting world price to local currency through the exchange rate and deflating by the local consumer price index. Production and net trade, therefore, are estimated with respect to this converted world price. 19 The Developed Makets and the Soviet Bloc are exceptions to the method of determining price described above. In the case of the Developed Markets, where consumers and producers are well insulated from world prices, internal European Economic Community (EEC) producer prices are used to estimate supply and demand. For the Soviet Bloc, policy variables are used in place of price variables because the economy of this region is not based on a market with prices that carry information about relative cost or utility. The transmission of world price to internal economies is not addressed directly in the Ag Model because domestic prices are not used, except in the Developed Markets. There, price transmission is assumed to be zero. The effects of government policies that separate the domestic market from the world market, such as tariffs and subsidies, are implicitly incorporated by observing quantities produced, consumed and traded. Economic, rather than geographical, aggregation of regions makes monitoring of international transportation costs impractical. Because the focus of the model is on net effects rather than on the specific pattern of trade flows, transportation costs are assumed to be constant. 3.2 MEASURING SUPPLY ELASTICITIES PEXD is the net effect of the demand responses of importers and the supply responses of competing exporters to a price change. Before measuring PEXD, individual supply elasticities for each country are calculated at the farm level from the harvested area equations. These harvested area elasticities affect export supply at the national level. It is the interaction of the farm- and national- level supply responses 20 of each region with the internal consumption and import demand equations that determines PEXD. Use of the term ”supply elasticity" to describe the harvested area response at the farm level may be misleading. Production is calculated in each country or region as an identity, the product of harvested area and yield. With yields determined exogenously, it is harvested area that responds to price and other variables in the model, but it does not equate with supply. The farm-level acreage response is carried through to the national and international levels and thus understates the supply response throughout the system because factors other than acreage that affect farm-level supply are not captured. The "supply elasticity" in this case measures percent change in a single input with respect to percent change in price rather than percent change in output with respect to percent change in price. Since other inputs are excluded, the elasticities in this study are expected to be lower than ones which measure outputs. Price elasticity of supply can be calculated by adding the elasticities of each component of production (Chiang, 197k): a =dPR0!F/Pno=g_1_g!§'/EK+g_Lfii/Y=E +3 S dP d? d? HA Y where: E = elasticity PRO = production S = supply HA = harvested area Y = yield P = price Harvested area is estimated in the Ag Model using a partial adjustment framework after Nerlove (1958). The generic form includes lagged harvested area, lagged revenues for own and substitute crops and other variables as follows: 21 HA = f(HA {-1), REV (-1), REV (—1), Z) ij ij ij kj where: HA = harvested area for commodity i in region j, 13 HA (-1) = harvested area, lagged one period, 13 REV (-1) = gross revenue per ha, commodity i, lagged one period, 13 REV {-1) = gross revenue per ha, commodity k, lagged one period, kJ Z = other relevant variables including time trend, policy proxy. See also Shagam (1987) and Mitchell (1983). Use of the lagged harvested area structure assumes that a) farmers only partially adjust to a change in expected price in any given year due to uncertainty, high costs of change or other factors; and b) farmers use the previous year's price as an estimate of the current year's price. Revenue is measured on a per hectare basis - price per ton times a four year moving average of yield in metric tons per hectare. Yields are averaged in order to even out the impact of drought years on revenue. Crop yield is a variable that is difficult to model accurately. The major determinant of yield is weather, which is a stochastic factor, unaffected by economic variables. Advances in technology (including improved plant varieties and new methods of disease and insect control) and the use of fertilizer are two dominant considerations in addition to weather. Weather causes large changes in yield from year to year, while the effects of technological improvements on yield tend to be gradual and unidirectional. In view of the difficulty in predicting weather, yield is simply estimated as a function of a time trend (Mitchell, 1983). The trend variable represents factors that change gradually over time and is intended to incorporate changes in technology. 22 a If price is not included in the yield equation, then m is Y implicitly assumed to be zero. If E :0, then E0: EU, or dHA/dP * EVE}: In the case of the Ag Model, the elasticity calculation is complicated by including yield with the price variable to generate revenue as described above. A further complication is that yield is not simply the value for the current year, but a moving average of the previous four yearS-2 Therefore, the elasticity that is calculated is: dHA * (E?§§?EL)/§X as opposed to dHA ' PVEAI W) a?- The difference between results from these two formulations may not be significant. However, the inclusion of yield may make the revenue variable less volatile than price alone. The smaller the variability in a regression variable, the greater the standard error in its coefficient, all other things equal. Therefore, revenue elasticities may be statistically less precise than price elasticities. Short run and long run revenue elasticities are calculated from the estimated equations for each region, except the Soviet Bloc, where price variables are not used. * $733 ) E =fi. F/fi E = [3 Short Run Long Run (1-) The short run in this case is the first year harvested area response, while the long run is greater than one year. For importing countries, import demand is estimated with a single equation. Price elasticities of demand are calculated for those 2For soybeans, the four-year moving average was replaced by a time trend in order to conserve degrees of freedom. 23 equations that contain price variables. These elasticities appear in Appendix 2. However, many equations include policy variables, rather than price variables, which account for government actions to insulate domestic producers from price fluctuations. Hence, net import demand elasticities for all regions are calculated by observing the change in model results when price is changed. These elasticities are found in Tables 5 and 9 in Chapter 5. For a detailed discussion of importer behavior within the Ag Model, see Wilde, et al. (1986). Table 1 presents the harvested area supply elasticities calculated from Ag Model results. All regions of the model show inelastic supply for all commodities in the short run. In most cases, long run supply is also inelastic. The notable exception is Brazil, where both long run elasticities that are calculated exceeded one. While these results suggest that farmers' responses are inelastic to price changes, they should be interpreted with some caution because they are measured from harvested area, an input, rather than from total supply, the output. Nerlove (1956) has suggested that "the elasticity of acreage is probably only a lower limit to the supply elasticity.” If these elasticities are viewed merely as minimums, then the minimums are quite low and do not provide much information about the true value of the elasticity. Greater restrictions on the harvested area equations may imporve the precision of the elasticities. However, the use of harvested area rather than supply, and revenue rather than price, raises uncertainty in imposing standard restrictions. More robust estimations may be obtained with a less restricted model than with an incorrectly restricted model (Beattie and Taylor, 1985). 2” TABLE 1: Revenue Elasticities of Supply at Farm Level Short Long Cross Run Run Revenue Argentina Wheat .327 .H7A -.307 (S) Feedgrain .291 .A61 -.372 (W) Soybeans .3A1 11.370 -.165 (H) Brazil i Wheat .670 3.050 -.505 (F) Feedgrain .573 - -.312 (W) -.151 (S) Soybeans .18N 1.7A3 -.053 (F) Australia Wheat .095 .A88 -.360 (P) Feedgrain .35A - -.667 (W) Wheat (new) .097 1.7uu -.097 (F) Feedgrain (new).650 - -.650 (W) Canada Wheat .098 .207 -.u67 (F) Feedgrain .361 1.400 -.260 (W) Developed Markets Wheat .u36 1.9H9 -.285 (F) Feedgrain .196 .35A -.21A (W) Soybeans .098 .h36 -1.053 (F) Low Income Countries Wheat .055 8.670 -.111 (F) Feedgrain .103 .200 -.073 (W) Soybeans .121 .309 -.530 (F) Newly Industrialized Countries Wheat .331 7.931 -.936 (F) Feedgrain .39“ .607 —.122 (W) Soybeans .16A .258 - U.S. Wheat .195 .390 -.208 (F) Feedgrain .376 .A30 -.209 (W) Soybeans .367 1.A53 -.343 (F) Oil-Exporting Low Income Countries Wheat .169 .173 -.333 (F) Feedgrain .238 - -.10A (S) Soybeans .200 .311 -.197 (F) China Wheat .072 .106 - Feedgrain .111 .A93 -.07A (W) Soybeans - - -.167 (F) Sample Period '55-'3u '6A-'83 '65-'8u ten-van '68-'83 '68-'83 '6A-‘8u '6A-‘8A '6A-‘84 '6A-'8A '6A-‘8A '6A-’80 '63-'83 '63-'83 '65-'83 '6A-‘8A '6A-‘83 '65-'83 '63-'83 '63-'83 '65-'83 '6u-v8u '6A-‘8u '65-'8n van-v83 '68-'83 van-v82 '64-'8A '6A-‘8A '65-'83 25 While the cross elasticities are all inelastic with one exception (soybeans in the Developed Markets), many of them are larger than the short run and even the long run own-revenue elasticities in the same equation. These relatively strong cross elasticities indicate that substitution between crops is an important factor in predicting farm level response to price. In some cases, it is possible that these elasticities overstate the effect of a competing revenue. This may be the case where some overriding factor is overlooked in the specifi- cation. Multicollinearity may also blur the distinct effects of the price variables. An example of large cross-elasticities that will be discussed in detail in subsequent sections is that of Australia. The cross-elasticities for both crops exceeded the own-elasticities in the short run and also in the long run for feedgrain. Because wheat is the dominant crop in Australia, its revenue could reasonably have a strong influence on feedgrain hectarage. However, it is unrealistic that the reverse should be true at the same time - that feedgrain revenue would be more important in wheat harvested area than wheat revenue. One factor at work in this situation may be land clearing that brings more cropland into production despite a downward trend in both grain prices. A second factor is the degree to which the revenue variables move in the same direction. The competing revenue variables in both equations pick up the negative sign associated with downtrending revenue but it is difficult to separate the effects of the two variables when they are collinear. New equations are estimated that constrained the cross-elasticities to equal the own-revenue elasticities. While such a restriction may not be completely correct, a ratio of revenues reduces multicollinearity problems by halving the number of revenue 26 variables in the equation. These results appear under Australia and are labeled "new." 3.3 AGGREGATE RESPONSE 3.31 Measuring Aggregate Supply Responses Supply response includes the effect of substitution between crops and that of overall contraction or expansion of area. Relative price changes induce substitution between crops rather than a shift in total harvested area. On the other hand, when all prices change simultaneously, substitution amongst crops will be minimal and the dominant effect will be an overall change in total harvested area. The net effect on all three crops is an aggregate response. In order to measure the behavior of aggregate area, it must either be modelled directly, which focuses on the aggregate response, or with each crop modelled separately, which focuses on the individual response. The estimated areas are then summed over the three crops to arrive at a total harvested area, or cropland base. The formulation of the Ag Model follows the latter method because the intent is to capture the dynamics of each commodity.3 Cropland base of the major exporters is estimated directly to compare the accuracy of this method to that of the summation method. For each country, an average revenue, weighted annually by the proportion of each crop's area, is calculated as follows: n REV :2 [ HA“) * l1--yr avg. yield(i) ’ Pw’XRg ]) ] i=1 CLB 091(3) 3Land that is double-cropped is counted as twice the area. 27 where: REV = average weighted revenue, HA(i) = harvested area for crop i, CLB = cropland base, sum of the harvested areas, yield(i) = metric tons per hectare of crop i, Pw = world price, XR(j) = exchange rate for country/region j, CPI(J) = consumer price index, country/region j. Cropland base is regressed using ordinary least squares in a Nerlovian adjustment framework on lagged values of the weighted average revenue: CLB = f(CLB(-1), REV(-1)). A weighted average of revenues is used because of high correlation between prices. Multiple prices would increase the likelihood of multicollinearity in the independent variables and decrease the reliability of the coefficients. Because harvested area shares of the individual crops would need to be determined from the total area, the use of individual revenues in the cropland base equation would create simultaneity problems if those same variables are used in addition to cropland base in the individual harvested area equations. Each cropland base equation is then inserted in a model of the appropriate region that solved recursively for harvested area, production and the other variables as described in Section 3.1, but with revenue exogenous. The estimated values are then functions of estimated lagged values rather than actual values. The fit of this direct estimation method is compared to that of summed harvested areas, both 28 based on estimated lagged values. T-statistics are examined for the contribution of weighted average revenue to the fit of the estimated equation. In all cases where cropland base is regressed on lagged revenue and lagged cropland base no strong positive linear relationship appeared. For most countries, t-statistics for the revenue variables are non- significant. In the case of Australia, revenue is significant, but negatively signed (see Appendix 3 for statistical results). Figures 1 - 6 compare the fit of the forecast cropland base and the sum of estimated harvested areas to the actual area. In most cases, summed harvested area estimations are superior to the direct cropland base estimate. Only in the U.S. did direct estimation follow the actual cropland base more closely than summed harvested areas. However, in the U.S., as in the other regions, the summed harvested areas captured more turning points correctly, implying that they contain more information than the direct estimate. Therefore, the summation method of determining cropland base is retained. 3.32 Sources of Low Correlation Several factors contribute to the poor fit of the cropland base equations. First, while the revenues for own and competing crops capture the tradeoffs in the cropping mix, they do not reflect the important decision variables in determining aggregate response. Changing total harvested area is essentially an investment decision. The profitability of a non-farming investment alternative may be relevant, but the relative returns for various crops are not. Second, with a single revenue variable, only one coefficient can be attached to D: 29 CHOPLAHD BASE - ARGENTINA Direct Estimation vs. Sun of Harvested Areas 15m 15% 14M* 13M+ 12M‘ 11. 1m 1975 1976 1977 1978 1979 1988 1981 1982 1983 1984 -— Actual ------- Forecast --- Sun of Harvested Areas Figure 1: Cropland Base Estimations - Argentina 27588 25888 ‘ ~ 22588 17588 15888 t J , 1 1 J BOPLAHD DAGE- 88A81L ireot Estimation vs. Sun of Harvested Areas 19 71 112 13 711 7‘5 7'5 55 is 19 9'9 1? a so 04 — Actual ------- Forecast --- 81111 of Harvested Areas Figure 2: Cropland Base Estimations - Brazil 3:: as: GOO» 3O CBOPLAHD BASE : AUSTRALIA Direct Est1nation vs. Sun of Harvested Areas 28889 17388 19999 f. . . . , . . , , . . 9- . . A 1955 1959 1979 1979 1979 1975 1979 1999 1997 1994 — Actual ------- Fitted --- Sun of Harvested Areas Figure 3: Cropland Base Estiaations - Australia CROFLAHD BASE - CANADA Estimated Directly vs. Sun of Harvested Areas 25999 225W‘ 2999991 175999 15M1 12599 1 . I a 1 Y . . 1 . 4riZF-IT-IFI-f-F-f-F-f 1966 1968 1979 1972 1974 1976 1979 1989 1992 1984 q — Actual ------- Forecast ---Su11 oi‘ Harvested Areas Figure 4: Cropland Base Estimations - Canada D: “at” ’2 me— 31 CROPLAHD BASE- lIHITED STATES Direct tEstiuation vs. Sun of Harvested Areas 128888 k 1199999» 199999« 999999 39°99‘..- 788881 9 59999T9.9.;1..,,r.. 79 71 7a 77 74 7s 75 77 79 79 99 91 99 93 94 —- Actual ------- Forecast --- Sun of Harvested Areas Figure 5: Crepland Base Estimations - United States ASE DEVELO ED HARHETS flgg‘t‘mfiseintion vs. Sun of Harvested Areas tiSflW-T9 1966 1968 1978 1972 1974 1976 1978 1988 1982 -— Actual ------- Forecast --- Sun of Harvested Areas Figure 6: Cropland Base Estimations - DeveloPed Markets 32 the revenues of the three crops of interest. Some information and flexibility is lost by restricting the weight on each crop to its proportion of total area. The individual harvested area equations may perform better than a single cropland base equation because they contain more information, such as policy variables that influence supply of the particular crop. When they are summed, the fit is closer to the actual values than that of direct estimation (Figures 1 - 6). In comparison, the cropland base equations, identical for each country, contain only revenue and the lagged dependent variable. It might be possible to estimate cropland base equations that track quite closely to the actual values, but they have less ability to capture the dynamics of individual commodities and substitution effects, which are usually considered more interesting. Both specifications show cropland base to be unresponsive to world price. This finding supports the view that elasticities are low in the aggregate. Both the direct estimation and the summing approaches likely share some of the same weaknesses in attempting to measure supply responsiveness. As suggested above, the relative prices of the crops of interest would not be expected to elicit a strong aggregate response. However, a key variable may be revenue from enterprises that are not included in the model, whether they are products that are locally important such as sunflowers in Argentina or livestock in Australia or non-farm activities that compete for land. Poor returns to livestock production may cause a shift from pasture to small grains even in the face of declining revenues for grains. Including these other country- relevant variables may improve the predictive power of these equations. 33 Another important consideration is that gross revenue may not be as relevant to the investment decision as net revenue. Changes in the price of farm output may have no relation to changes in farm income, especially in countries where government intervention in the economy is considerable. In a supply elasticity, the relationship that is theoretically measured is farmers' intent to produce relative to expected price. Droughts and other supply-reducing events appear as outliers in the cropland base data. This is an imperfection in modelling farmers' intentions that is accomodated by the Ag Model. Intent is better captured in area planted, but cropland base is the sum of area harvested. The complexity of estimation and the possibilities for error increase if both planted area and harvested area are included in the specification or if supply must be estimated from planted area instead of calculated from harvested area. The relationship of the cropland base to revenue based on border price reflects the aggregate responsiveness of harvested area to world prices. While this is specifically the intent of the modelling effort, this relationship may not be strong if the border price is substantially different from the internal price that farmers actually face. The less world price is transmitted to the domestic market, the less a market participant will respond to changes in the world price, resulting in a low price elasticity value. 3.33 Relationship of Cropland Base to Gross Revenue Cropland base and weighted average gross revenue are plotted over the historical period to note obvious patterns or discrepancies. A 3h pattern of strong association would suggest that the two are related and high adjusted R273 would be expected (Appendix 3). Little or no association over time would support the idea that other variables or specifications are needed to improve accuracy. These plots appear in Figures 7 - 12. From these figures, cropland base shows little relationship to gross revenue in most cases. It is evident from Figure 7 that the large rise in world prices does not translate into increased hectarage in Argentina. Some interventions such as a tax or tariff may have prevented the market signal from ever reaching the farmers. Price stabilization policies could be expected to smooth aggregate response by insulating farmers from world market instability. In Figure 8, Brazilian cropland base has trended upward considerably, during a period of roughly constant revenue, before the sharp increase in revenue of 1973. During the late 1970's and early 1980's, cropland base leveled off, as did revenue, on average, but an association between the two is not clearly evident. In Australia (Figure 9) some overriding trend such as declining production costs or substitution away from other competing enterprises and into wheat and feedgrain may have swamped the effects of revenue. Again, farmers may be highly insulated from world prices. Whether there is little aggregate response to price or whether low price transmission disguises a stronger response is not clear from this figure alone. However, it does suggest that aggregate revenue elasticities would be low. No distinct association between cropland base and revenue emerges from Figures 10 - 12 of Canada, the U.S. and the Developed Markets. D: muses-1:: REV ozxmzuece-eau ES 35 AH IHA 680 HD BASE vs HEIGHTED AUG. REVENUE 17.5 9.‘”’""*.1s.9 2.14 .‘.”1 : .“ ... -;° 912.5 a. 04 .. .:..'.”"«.'0.-....°'. nuns... ..‘".'.. 1, 1'. a F; a3; 9‘ 1.54L / r 7.5 1.81M/ 8.5 59 ' 57 ' 54 ' 55 ‘ 59 ' 79 ' 77 r 74 Y 75 T79 ' 99 ' 97 ' 94 — 7991:1711: «- CROPLAHD BASE Figure 7: Cropland Base vs. Gross Revenue - Argentina 21!. 113919779 BASE vs HEIGHTED AUG. REVERE 299. 17s. 1561 .u a; . 125....-. a“. 199 58 59 i 52 ' 59 755 ' 59 ' 79 ' 77 ' 7h ' 7t 879 ‘ 99 ' 99 ' 99 — 112021812 «- CROPLAHD BASE Figure 8: Cropland Baee ve. Gross Revenue - Brazil 3: WW RED spasms-scam 36 A ALIA C1139§AHD 8A8! vs HEIGHTED AUG. REDEHIIF . 28.8' . "Mum ee .,.-'ue'°""""" .' . ' 15 I a 3.31 '- “ "'-.._ ..--- 91.2.5 31 9'1 .m,.-"'°°°°.... 3'. ‘ " 19 7 8 2.:*l~«"‘”° .. 7 5 3.8‘ \/\_ . 1.5: 1.81 9.5 . i . . 7 r . . A , . . , 5 . . rs . , . 68 62 64 66 68 78 72 74 76 78 88 83 84 — REDEHIIE «- CROPLAHD BASE Figure 9: Cropland Base vs. Gross Revenue - Australia HA 88092:!” RASE vs HEIGHTFD AUG. REDDIIE 22.5 728.8 6 ., 717.5 5‘ ~ “ '150‘ 3‘ 12.5 2 1 59 ' 57 ’59 159 ' 59 5 79 i 79 ‘79 ‘ 7t ' 79 ’79 T 97 i 99 -— 99911111: «- CHOFIAHD 8882 Figure 10: Cropland Base vs. Gross Revenue - Canada ..a. a........~.... g3 ' D: \mczmcman E b=\a-s=::s-s¢e-sale 37 UNITED STATES CROPLAHD BASE vs HEIGHTED AUG. 887721818 9. 8 7. 5.. 5. 4. 68 62 64 66 68 78 72 74 76 78 88 82 84 — REM-Jill! «- CROPLAHD BASE Figure 11: Cropland Base vs. Gross Revenue - United States DEVELOPEDH HAlHETS CROPLAHD 8A SEE vs HFIGHTFD AUG. REUEHIIE 48.8 .‘O‘... £4 7 4? I5 g' " 147.9 8 '“es 7‘ .. 945.1 5. 745.5 5. '450' 4. 3 59 5'7. 5'4 5'5 5'9 79 7’2 7'4 7'5 79 9‘9 9'7. — 99957111 -,CROPLAHD 9957: Figure 12: Cropland Base vs. Gross Revenue - Developed Harkets :9: 3°.“wa a 38 Cropland base and revenue are both relatively flat (except the revenue peak in 1973) for Canada and the Developed harkets, although for the DevelOped Markets revenue dropped to a lower plateau after 1975. In the U.S., cropland base trends upward in a fairly uniform manner despite volatility of revenue. Sharp drops in cropland base are due to factors other than price response, such as the government-induced reductions in hectarage in the U.S. in 1983 when the Payment-In-Kind program paid farmers with stored grain to idle a percentage of their crOpland. 3.34 Relationship of Internal Prices to Border Prices A low responsiveness of cropland base to gross revenue based on border prices raises the possibility that price transmission is low in some countries. In order to determine whether border price is a reasonable proxy for producer price, the two are compared graphically wherever internal price data is obtained. Argentina has had markedly different policies toward agriculture depending on which political group is in power. During Peronist regimes, agriculture was heavily taxed and in the intervening periods agriculture was more market-oriented. Peronist administrations controlled Argentina in 1988-55 and again in 1973-75. which was a period of high world food prices. Taxation of agriculture was extremely high during this period and prevented agriculture from receiving the benefits of high world prices. Export taxes and differential exchange rates separated agriculture from the world market. In recent years, exceedingly high inflation and frequent changes in government have created instability in the economy that may have reduced the response of farmers to any change (Rainio, 1983). 39 Wholesale market pricesH for Argentina (Bolsa de Cereales, 1984) are used to calculate an internal revenue in the same way that weighted average revenue is constructed from border prices previously (Section 3.31). These revenues are plotted together in Figure 13. While internal prices followed the general pattern of world prices in years of stable prices, they failed to reflect the extremes, such as in the mid- seventies. In the case of Argentina, border prices do not closely represent the prices that farmers receive. The Brazilian government has supported commodity prices to ensure a minimum income to farmers and an adequate supply for the domestic market. Domestic price ceilings, an overvalued currency and export quotas for soybean products in most years have all served to separate the domestic market from the world market (Williams and Thompson, 1988). Farm prices in Brazil are used to calculate an internal revenue which is plotted against border-price revenue in Figure 14. Internal revenue is lower than border revenue in all years but one, and followed the general movements of border revenue. However, internal revenue does not rise as sharply as border revenue in years of large price increases. As in Argentina, border price does not reflect the price farmers face, but because both revenues follow the same general pattern, border price may be an acceptable, though not ideal, proxy for internal price in Brazil. In Australia, wheat prices received by farmers reflect a weighted average of returns from wheat sold domestically and wheat that is exported. This "pooled" price is paid to all farmers regardless of where their grain is actually sold. Producer price is determined by the ”Data from sources outside the Ag Model are presented in Appendix A. DB: wan-em a: \mzmcmw 40 ARGEHTIHA . _ HEIGHTED AUG. REUEHUE Border Prices vs. Internal Prices 2.5 2.87 1.51 so e. .e .. .C e e 1 9. x = 8 . ee'. 'eeee. .e ' e e Q 8.5 59 59 ' 54 ' 55 t 59 ' 79 T 79 ' 74 ' 75 ' 79 ' 99 ' 973 94 — Border ----- Internal Figure 13: Ueighted Avg. Revenue - Argentina Border Prices vs. Internal Prices 21 IGH7ED AUG. REUEHIIE Border Prices vs. Internal Prices 3 299 175‘ 1594 125- 199 75 s, 58‘ " 25fi,e.,..i.,..... 68 62 64 66 68 78 72 74 76 78 88 -— Border ------- Internal Figure 14: Ueighted Avg. Revenue - Brazil Border Prices vs. Internal Prices »:\MI=:ZP1€M:I D: \mzmm 41 100‘ 8.5 59 r57. r54 155 .59 T79 -72 T74 ' 75 '79 99 92 '94 — Border ------- internal Figure 15: Ueighted Avg. Revenue - Australia Border Prices vs. Internal Prices D FLOPED MARKETS , , IGHTED AUG. REUEHUE Border Prices vs. Internal Prices 3 1 ' 1 ‘ : ‘ 1 ‘ . ' 1 ’ 1 r 1 ’ 9 ' | ' 1 62 64 66 68 78 72 74 76 78 88 82 — Border ------- Internal Figure 16: Ueighted Avg. Revenue - Developed Harkets Border Prices vs. Internal Prices 82 Australian Wheat Board which controls all wheat marketing in Australia. The Board bases the price of wheat for export on world prices (Spriggs, 1978) while various schemes have been used for pricing domestically consumed wheat. Currently, domestic wheat is priced at approximately twenty percent above the export price. With a population of only about 19 million, Australia's domestic consumption of grain is quite small in proportion to the harvest. Thus, pool price is dominated by the export price. The price of feedgrain in Australia is approximated by the domestic price of barley, the main feedgrain produced in Australia, as opposed to corn price, which represents world market feedgrain price in the Ag Model. Barley and cats are primarily grown in the States of Western Australia, South Australia, and New South Wales, much of the same area as wheat, where the climate is suited to winter crops (Spriggs,1978). Barley marketing is controlled by four marketing boards. Unlike wheat however, it is legal to sell barley privately and most barley for domestic feed consumption is handled through private channels. The boards handle all barley for export and for domestic malting purposes. In order to represent the price received by farmers for feedgrains, the gross value of barley is divided by total barley production. A revenue variable is constructed using these average returns for wheat and barley. In Figure 15 internal and border weighted average revenues are plotted for Australia. Internal price closely follows world price in this case. The only year where the two deviate substantially is 1973, a year of exceptionally high world grain prices. This difference is partly explained by the wide disparity between domestic and export price 43 in that year and the effect of averaging. Of the four regions studied here, Australia is the only one that appears to have an open market that bases prices on world prices. Near perfect price transmission is suggested by the closeness of the two plots in Figure 15. The European Economic Community (EEC), which is the majority of the Developed Markets region, has clear policies of farm price insulation (Jabara and Brigida, 1980). The main mechanism of price insulation is the variable levy. Imports from outside the BBC that are cheaper than the supported price of domestic agricultural products are subject to a levy. The value of the levy adjusts in order to raise the price of the imported commodity to a fixed threshhold price, which is greater than or equal to the price of the domestic product. Producers and consumers are completely insulated from the world price in this way. Figure 16 shows revenue calculated with producer prices used in the Ag Model for feedgrain and wheat as well as border price revenues for the Developed Markets. Although the overall trend in producer revenue follows the trend of revenue generated by world price, it does not respond to large swings in world price and producer revenues demonstrate very little variation. This revenue stability suggests that producers in the Developed Markets are effectively insulated from the world market. Therefore, border price would not capture the prices that producers in this region face. In most of the regions examined in this chapter, cropland base shows little association with weighted average gross revenue, both from plots of the two variables over time and from statistical regression. The lack of a strong correlation between the two could occur because cropland base poorly represents farmers' intentions to produce or HM because lagged revenue per hectare does not approximate expected price. Alternatively, the lack of a strong association may be either because world price is unrelated to the price farmers face (low price transmission) or because farmers simply do not respond to expected price. The association between border revenue and internal revenue studied in this section identifies regions where price transmission is an important factor, specifically Argentina and the Developed Markets and possibly Brazil. The use of internal prices, which circumvents low price transmission, and other factors that affect the relationship between price and supply are examined in the following chapter. CHAPTER IV. COUNTRY-LEVEL INVESTIGATION The relationship between farm-level supply and price is examined in further detail in this chapter to determine why Ag Model elasticity estimates are low. Low elasticity values may result from specification error in either the supply or the price variable, from low price transmission, or from low actual elasticities. To examine these possibilities, new country-specific variables are introduced in two of the Ag Model's eleven regions. The new variables include a) an alternative specification of revenue, b) internal prices, which eliminate price transmission difficulties, and c) prices of substitute products (for suppliers). The two countries studied are Australia and Argentina. Australia is chosen because of the apparent negative relationship between cropland base and weighted average gross revenue. It may be possible to explain this unusual result by including other variables in the specification. Argentina is identified in the previous chapter as a region where border price poorly approximates internal price. Re-estimation with internal prices may result in a higher elasticity. If so, then price transmission is the cause of low elasticity. If internal prices do not raise the elasticity or improve the fit of the harvested area equation, then the likelihood that the elasticity is in fact low is increased. To explore the importance of each of the considerations discussed above, the new variables are introduced in the harvested area equations. 45 46 Specifically, 1) internal real wholesale or producer prices were substituted for border prices to observe the responsiveness of supply, when imperfect price transmission is eliminated as an obstacle; 2) an index of prices paid by farmers is introduced, to deflate gross revenue to account for variations in real costs of production; 3) enterprises identified as likely competitors for crop area are included to examine their effects on hectarage decisions. A.1 AUSTRALIA n.11 Internal Prices Internal prices are the first variables tested in the harvested area equations. Because the internal price for Australia followed the world price closely, internal prices would not be expected to improve the equations appreciably. The old equations contain wheat and feedgrain revenues calculated from border prices. In the wheat equation neither revenue is significant, but both revenues are significant in the feedgrain equation. The original wheat equation also contains lagged harvested area, a time trend and wheat ending stocks, a proxy for government policy. The initial feedgrain equation contains only the two revenues and a time trend, but not lagged harvested area. The coefficient of adjustment for harvested area in this case is one and the partial adjustment specification is dropped. When both internal revenues are introduced into the wheat and feedgrain harvested area equations, t-statistics are lower than for border price revenues. Other measures of fit also worsen. Specifically, in both cases adjusted 82 and F-statistics decline and standard errors of the regressions (SER's) rise. 47 Because grower returns for wheat reflect a weighted average price for a particular year's crop, farmers in Australia do not receive the full price until the entire crop has been disposed of. However, at delivery farmers do receive an initial payment. This initial payment had represented 70-80 percent of the anticipated final price in the past. Beginning with the steep rise in world prices in the 1970's, initial payments did not keep up as a percentage of the final price. Since 1979 initial payments have equalled the guaranteed minimum price. If market prices fall below this level, the farmer receives a subsidy. Operating costs for the Board are deducted from wheat returns before growers are paid and transportation differentials are also charged depending on farmer location. Because farmers receive wheat price information over a period of years, an alternative lag structure is tested in order to measure the effect of the delay in receiving price information on supply response. The lags are intended to differentiate between the effects of the initial payments and the final price. A two-year lag for wheat revenue only is tested in both the wheat and feedgrain harvested area equations to capture the lag in determining the pooled wheat price. Feedgrain prices are not complicated by delays in price information, therefore only a single period lag is used for feedgrain revenue. A two-year lag for wheat revenue does not produce a significant coefficient for either border price or for internal price. This result suggests that farmers do not rely on final payments from a wheat crop two years earlier in order to form price expectations. While initial wheat payments may be important in forming expectations of the final price, initial payments have been set equal to #8 the guaranteed minimum price since 1979. A complete series for initial payments is not available, but the guaranteed minimum price is tested in the wheat harvested area equation and shows no statistical significance. H.12 Gross Revenue vs. Net Revenue The question of net revenue as compared to gross revenue is addressed by deflating both border and internal revenue variables by an index of prices paid by producers. This net revenue index variable is then substituted into the harvested area equations in place of gross revenue. First, border-price revenues. vet revenues shows no significance in the harvested area equation for wheat. In the feedgrain equation, net wheat revenue is significant, as is gross wheat revenue and net feedgrain revenue is slightly more significant than gross revenue (t=2.h9 vs. 2.3u). However, the overall fit of the net revenue equation is slightly worse with lower adjusted R2 and F-statistic and higher SER. Therefore, the net revenue specification is rejected. Second, net revenues based on internal prices are tried and again, yield poorer results than gross border revenues. ”.13 Competing Enterprises Experts have suggested that the relative unprofitability of raising livestock in Australia, particularly on marginal, droughty land has contributed to growth of the cropland base, especially wheat area. The average price of greasy (raw) wool is introduced into the wheat equation to test the significance of a competing enterprise. Although sheep and wheat are produced in overlapping areas of the country, primarily in the "wheat-sheep belt," the land that is shifting from livestock to small A9 grains is only marginally suitable for sheep ranching because of seasonal drought. Therefore, a measure of profitability per hectare of such land may be more appropriate. However, such a variable is not available. The price variable is not expected to show statistical significance because it does not capture the tradeoff between sheep pasture and wheat production. In fact, real wool price shows no significance in determining wheat or feedgrain harvested area. This result does not rule out the importance of sheep enterprises in decision-making for wheat, but merely suggests that some other variable is needed to capture that effect. 4.1% Summary In each step of the analysis, the initial Ag Hodel equation is used as a basis for comparison over the sample period, 1960 to 198“. Internal and net revenue variables are substituted for gross revenues in the initial specification while competing prices and policy variables are added to the initial explanatory variables. Statistical significance, as measured by t-statistics, is the primary criterion for the contribution of a variable to the equation. Contribution to adjusted R2, Durbin-Watson, standard error and F-statistics are also considered. None of the variables tested showed a clear improvement over the initial equations for wheat or feedgrain harvested area. However, the initial equations have the problem that in each case the cross-revenue elasticities are considerably larger than the own-revenue elasticities. This is not an unrealistic result a riori, but in Australia, where wheat harvested area is twice as large as feedgrain harvested area, it 50 is unlikely that the feedgrain cross-revenue would have such a large effect on wheat harvested area. Further, large cross-elasticities have the undesirable property of causing crOpland base to respond in a counter-intuitive (opposite) direction to simultaneous wheat and feedgrain price changes. As mentioned earlier, the strong downward trend in revenue for Australia may be responsible for the large relative magnitude of the cross elasticities (see Table 1). As no new insights are gained from the new variables in this chapter, the cross elasticities are constrained to equal the own elasticities by replacing the two revenue variables with a single ratio of own to cross revenues (lagged). This specification allows harvested area to respond to changes in relative prices, but does not address aggregate response to proportional price changes. In this regard, this specification is completely inadequate. However, it mitigates illogical behavior of the equations in the absence of more detailed information that would better explain Austrailian planting decisions. Australia is not a country where low price transmission inhibits farmer response to world price. Therefore, the use of border prices is appropriate. The inclusion of country-specific data does not improve the fit of these equations for Australia. In order to improve the Ag Model's representation of Australia, a considerably more detailed regional model would be necessary. Inclusion of the livestock sector and possibly weather-related variables would likely enhance the Ag Model's performance. 51 u.2 ARGEETIN H.21 Internal Prices In Argentina, a procedure similar to that for Australia is followed. Real internal wholesale prices, which differ from producer prices by a stable marketing margin, are substituted for real border prices in the calculation of revenue for wheat, feedgrains and soybeans. These internal revenue variables are compared to border-price revenues in the harvested area equations. In the case of wheat, the initial equation contains lagged harvested area, a time trend, and lagged wheat and soybean border price revenues. When these are replaced with internal revenues, soybean revenue remained significant, but wheat revenue does not (t=.91) and fit does not change significantly. For feedgrain harvested area, neither internal price, wheat or feedgrain, is significant and own revenue has an unexpected sign. The original equation for feedgrain contains lagged harvested area, lagged feedgrain and wheat revenues and a dummy variable for 1971 and 1979. years of drought. Soybean harvested area is the only equation that shows some slight improvement with internal prices. The original soybean harvested area equation consisted of lagged harvested area, lagged soybean and wheat revenues and a splined time trend beginning in 1976, when soybean production accelerated in Argentina. Wheat revenue, which is not significant in the original, is still not significant (critical alpha =.18 vs .25) but the variance of the coefficient is somewhat reduced. One important difference is that the sign on wheat is positive for internal prices and negative for world prices. At first glance, the 52 positive sign appears incorrect, but is plausible in view of the fact that soybeans are frequently double-cropped with wheat in Argentin m (Wainio, 1983). No suitable index of production costs is available for Argentina. Hence, the response to net revenues rather than gross revenues is not tested. 4.22 Competing Enterprises Important enterprises that compete with Ag Model crops in Argentina include beef, sunflowerseed and flaxseed production. Wheat is most important amongst the crops, but they are all subordinate to cattle ranching in central Argentina's Pampa region (Wainio, 1983). Revenues for each of these commodities are introduced in the harvested area equation for each Ag Model crop. Revenues for sunflowerseed and flaxseed are calculated in the same way as wheat and feedgrain revenues, using internal sunflowerseed and flaxseed prices. For beef, the wholesale steer price is deflated and used. However, beef price data is only available from 1970 onward. When beef price is used, sample size is reduced to 1971 to 1984. Internal soybean price data is available only from 1966. When it is necessary to change the sample period of the estimation due to data limitations, an equivalent original equation is estimated using the shorter sample period in order to compare similar equations.1 Enterprises deemed important by Wainio in his study of farmers' responses to grain prices under various political regimes are tested first. In the case of wheat, beef is the only relevant commodity 1The initial sample period for soybean equations is 196A to 198k 53 considered by Wainio. However, beef price is not statistically significant for wheat, nor for the other crops of interest. Subsequently, feedgrain, sunflowerseed and flaxseed revenues are tested. Only sunflowerseed is significant and signed as expected (negatively). Replacing soybeans in the original equation, it improves fit slightly With higher adjusted 32 and F-statistic and lower SER. The best equation, therefore, is a function of lagged harvested area, lagged wheat and sunflowerseed revenues, and time. In the feedgrain case, wheat and sunflowerseed revenues and beef prices are all deemed relevant by Wainio. Beef and sunflowerseed are not significant, but wheat is. Flaxseed is also tested and proved significant, but feedgrain is never significant, even in the equivalent original specification. The best equation includes wheat and flaxseed revenues only (in addition to lagged harvested area and a dummy). Adjusted R2 rose but SER and F do not change considerably. Wheat, feedgrain, sunflowerseed, flaxseed and beef are all tested in the soybean equation but none prove better than the original specification of wheat and own revenue (with the change of sign for wheat revenue), lagged harvested area, and a splined time trend. While fit is slightly improved by the inclusion of one more competing revenue, flaxseed, the coefficient is not significant. ”.23 Summary The inclusion of internal variables improves the fit of each of the harvested area equations for Argentina, but only slightly. In this case, aggregated, international variables can be improved upon by SA measuring separately the response of farmers to price and the relationship of internal prices to world prices. In order to include these new equations in the model, it would be necessary to forecast all of the price variables in order to simulate into the future. In addition, some relationship between the internal price and the world price must be calculated in order to relate these revenue responses ultimately to changes in world prices. Because the internal price series would be difficult for Ag Model researchers to maintain in the future, the new equations are not incorporated into the model. 9.3 MULTICOLLINEARITY One complicating factor that arises with many similar competing enterprises appearing in the equation is the degree of multicollinearity present. While there are no definitive tests for multicollinearity, the coordinated movement of the prices of substitutes strongly suggests some problem with collinear data. Figure 17 shows the movement of real domestic prices for the three oilseeds and two grain groups used in the Argentine case. Specifically, oilseed prices are highly correlated with each other and wheat and feedgrain prices are highly correlated, but oilseeds and grains are less so. According to Pindyke and Rubinfeld (1981), "A rule of thumb states that multicollinearity is likely to be a problem if the simple correlation between two variables is larger than the correlation of either or both variables with the dependent variable.” This rule serves as a first test for simple correlation but does not measure multiple correlation. Due to time and resource limitations, multiple correlation tests are not performed as part of this study. For H3\¢n=::>r'c~ou eta—09mg H3 \mrpwhdmza cam—dmhmmc 55 REAL COMMODITY PRICES - ARGENTINA 2.9 105‘ 1.01 aa‘cwm" ~ 0.0 ,,,,,,,, 6B 62 64 66 68 79 72 74 76 78 99 82 84 FEEDGRA AIH -— SOYBEAN ------- HHEAT -—- FLAXSEED —- SUHFLOHERSEED Figure 17: Real Commodity Prices - Argentina REAL COHHODIT‘I’ PRICES - UNITED STATES 15.9 12,5. ml 7.5‘ w 2.31 o'oI fiI ' I r r I ‘ I ‘ . ' I ‘ I r I ‘ I ' I f! 59 52 64 52 58 79 72 74 75 79 89 32 94 . 9093“" OIL -— "HEM ------- FEEDGRMN --- $093229" -- 5093“" "EM: Figure 18: Real Commodity Prices - United States 56 a discussion of methods for detecting multiple correlation, see Judge et al- (1985). Argentina provides an example of potential problems with multicollinearity because several price variables are involved. The best specification for each harvested area equation contains one grain price and one oilseed price. Correlation between revenue variables is thus minimized because very close substitutes are not included in the same equation (eg. two oilseeds). Correlations between the revenues are DOt as large as the R2 of the equations. Correlations in the wheat equation are the highest. With an R2 of .56 in the wheat equation, the correlation between wheat revenue and sunflower revenue is .Ri. Correlation between wheat harvested area and sunflowerseed revenue is .33, and between harvested area and wheat price, .37. multicollinearity is most likely a problem in this case, according to the rule of thumb. Whether a strong relation exists or not is obscured by the presence of multicollinearity. The problem of multicollinearity is present in any equation that contains the price of substitutes or complements in addition to own price because some degree of collinearity exists between any two price variables used in the Ag Model. In the Ag Model, prices in most regions other than the U.S. are generated by converting the U.S. price to a border price. Figure 18 demonstrates the relationship between U.S. wheat, feedgrain and soybean product prices. Simple correlations between pairs of real prices range from .41 between wheat and soybean meal prices to .93 between wheat and feedgrain prices. Multicollinearity obscures the relationships between independent and dependent variables. Where multicollinearity is serious, there is 57 little that can be done to measure the relationships between variables more accurately without introducing a priori information. Restrictions based on prior information increase economic efficiency when they are valid, but introduce bias if they are overly restrictive (Abbott, 1987). Hulticollinearity no doubt exists among other Ag Model variables, but is most obvious among the price variables. CHAPTER 5: ELASTICITIES FROM MODEL SIMULATION In this chapter, price elasticities of export and import demand are measured and reported from simulation runs of the Ag Model. Hectarage response is also calculated from these runs. The runs consist of solving the model for a best estimate of future prices and quantities, then solving again with specific variable changes. The difference between the two runs measures the effects of the altered variables. The first step is estimation of the effect of loan rates on world prices. 5.1 EFFECT OF THE U.S. LOAN RATE ON PRICES The loan rate is essentially the price at which the U.S. government stands ready to buy certain agricultural commodities. A non-recourse loan is an arrangement where the U.S. government will lend money to farmers using their grain as collateral. The loan rate itself is the price at which the commodity is valued for loan purposes. If the market price of the commodity is unprofitable for the farmer, he or she may repay the loan in full with the commodity and the Commodity Credit Corporation (USDA) has no recourse but to accept this payment. These stocks accepted by the government enter the market when commodity prices reach a release price set by the government, generally 10 to 35 percent above the loan rate (Knutson, et al., 1986). During recent years, the loan rate has supported U.S. wheat and feedgrain prices above equilibrium levels. When this is the case, one 58 59 would expect that lowering the loan rates on wheat and corn would cause an immediate drop in the prices of wheat and feedgrain. Because U.S. prices are used as world prices in the model, world prices would also adjust immediately to changes in the loan rate in a simulation run. While there is a loan rate for soybeans, market prices have been high enough that it has not effectively supported prices in most years. For this reason, the loan rate for soybeans is not included in the specification of the U.S. component and cannot be directly tested. However, soybean price is expected to respond to changes in wheat and feedgrain prices due to substitution effects. To determine initial baseline results, the Ag Model is solved for the period 1975 to 1996. This preliminary run, call it number A1, uses historical data through 1983, and 198A where available, and uses the model's standard projections for exogenous variables beyond 198D. Appendix 5 contains the loan rates used for each crop. In a second run, A2, wheat and feedgrain loan rates are reduced by twenty percent below baseline levels, beginning in 1986, to test the degree of adjustment. The model is again solved over the ten year forecast period and the simultaneous adjustment of wheat, feedgrain and soybean prices is estimated. The results are presented in Table 2. TABLE 2: Percent Change in Price in Response to a 20 Percent Change in Wheat and Feedgrain Loan Rates 1986 1990 1996 Wheat -20.1 -20.0 -20.7 Feedgrain -20.2 -16.8 -18.6 Soy Meal -16.9 -23.5 -20.1 60 The reduction of the loan rates has an immediate effect on all three estimated commodity prices. On the basis of a twenty percent decrease in both the wheat and feedgrain (corn, sorghum, oats and barley) loan rates, wheat price drops 20.1 percent from the baseline in the first year (1986) and 20.7 percent by 1996. Feedgrain price responds to these same changes by declining 20.2 percent in 1986 and by only 18.6 percent by 1996. Dropping the wheat and corn loan rates by twenty percent causes soybean meal price to fall only 16.9 percent in 1986 but 23.5 percent in 1990 and 20.1 percent by 1992. Soybean meal price is used rather than soybean price because it is the main component of soybeans that interacts with feedgrain and wheat as livestock feed and because soybean price is simply a function of meal and oil prices in the model. Soy oil price is exogenous and therefore does not respond to changes within the model. As a result, soybean price is biased upward and does not fully reflect the response of the soybean market to changes in grain loan rates. The results indicate that a change in loan rates leads to a comparable change in U.S. market prices at times when the market price is close to and supported by the loan rate. In the following simulations, world prices (which are U.S. prices) are changed in addition to loan rates to assure precision in calculating elasticities. However, the results correspond to the effects of changes in the U.S. loan rate alone under current market conditions for wheat and feedgrains, assuming immediate price transmission from the U.S. to the world market. 61 This simulation is conducted to determine the effect on world prices of changes in the U.S. loan rate. Subsequent runs measure the effect of those price changes on U.S. exports and export revenues. 5.2 AGGREGATE PRICE ELASTICITIES Aggregate1 elasticities measure the effects of a simultaneous change in the prices of all Ag Model commodities on the supply and demand of exports and on the harvested area response for each commodity. The price changes parallel equal percentage changes in the loan rates. Aggregate export supply elasticities measure the percent change in net exports between two model runs as a result of a fixed percent change in all prices. Aggregate import demand, export demand and harvested area elasticities are calculated in the same manner. A baseline for comparison, number B1, is generated by fixing reve- nues and ending stocks for the forecast period of 1986 to 1996 at the 1986 estimated values. These 1986 values are taken from the initial run (no. A1) which determined equilibrium levels for the forecast period. The results of this baseline (B1) serve as a basis for comparison with runs where revenues are lowered by a fixed percentage (numbers B2 and BB). Figure 19 illustrates the baseline and scenario revenue levels. 1 Aggregate refers to elasticities calculated for a group of similar products, such as wheat, feedgrain and soybeans. Use of this terminology follows Cochrane (1958) and Tweeten (1967). 62 ................ base revenue ...................... scenario revenue Revenue —. o.— .. a.- on..- -- as- one. 986 1996 Year Figure 19: Base and Scenario Revenues for One Commodity, 1986 - 1996 Two ordinarily endogenous variables are exogenized in the PEXD simulations (82 and B3), the first is revenue. Revenues are held constant rather than prices because the focus is supply response rather than demand response. While it is price that consumers face, revenue per unit (per hectare in this case) is a more relevant measure for producers. Crop revenues trend upward over time if prices are held constant because changes in technology cause yields per hectare to rise. This upward trend in yield is maintained in the PEXD scenarios, but is counterbalanced by a proportional downtrend in real prices, thus holding revenue constant. While real agricultural prices tend to decline over time, the incremental decline in trend price is calculated for forecasting purposes to Just offset, on average, the increase in revenues due to yield increases. As a result, wheat prices are reduced by 1.7 percent per year, which is approximately the worldwide average increase in wheat yields per hectare. Feedgrain and soybean prices are adjusted downward by 1.6 percent and 0.9 percent per year respectively. 63 Soybean meal and oil prices are decreased by the same increment as soybean price (0.9 percent per year). Another variable that is exogenized is ending stocks. Ending stocks are held constant to represent constant government policies because ending stock variables are used as proxies for policy in many countries. Ordinarily, changes in ending stocks would have a strong effect on prices. Because the forces that determine prices are ignored in this scenario, the effects of changes in ending stocks are also ignored. Constant ending stocks also eliminate effects due to changes in speculative and storage demand in importing countries (surplus exporters are not assumed to hold speculative stocks). Two exogenous variables are held constant to eliminate the effects of differential rates of change in these variables between regions. First, inflation is removed from these scenarios by holding the deflator constant in all countries. Similarly, exchange rates are held constant to ensure that relative prices between countries do not change. While the equations in the model are estimated with very few restrictions imposed on them, this assumption in the PEXD simulation runs is a strong one. By holding exchange rates constant, it is assumed that exchange rates act as real price effects and, as such, are not relevant to the measurement of PEXD. For an alternative opinion on exchange rate effects, see Chambers and Just (1979). The first simulation run, number 82, tests the effect of reducing all prices and revenues simultaneously by twenty percent. The only changes made from the initial baseline run (B1), are to lower wheat, feedgrain, soybean, soy meal and soy oil prices, which automatically 6A adjusts revenues downward by twenty percent also, and to reduce loan rates by the same percentage. The model has been structured to reflect the assumption that the U.S. is the residual supplier to the world market. Competing exporters are assumed to face infinite demand elasticity for their exports because their share of the world market is small and the quantities they export would not affect world price appreciably. If the demand they face is infinitely elastic, then quantities sold at different prices trace out the supply curve as shown in Figure 20. 'U'U \ Figure 20: Supply of Exports by Countries Facing Infinitely Elastic Demand The U.S. is assumed to be able to supply all demand that remains after the competitors have exported. Constant ending stocks in this scenario ensure that U.S. supply is infinitely elastic. When supply is infinitely elastic, changes in price trace out the demand curve for U.S. exports as in Figure 21. Figure 21: Demand for U.S. Exports 65 Thus supply elasticities for competing exporters and PEXD for the U.S. can be calculated from changes in exports as a result of changes in price. For importing regions, elasticity of demand for imports can be calculated. If a country imports, demand exceeds domestic supply. When price drOps, quantity demanded increases and quantity supplied domestically decreases. The elasticity calculated from the model is a net elasticity of demand for imports, which is total demand minus domestic supply and incorporates adjustments in both supply and demand, shown in Figure 22. P D \ / s \/ /\ P1 / \ P2 1 'X [I ll - — 0 increase in imports Figure 22: Change in Net Import Demand 5.21 Results Elasticities are calculated from the differences between the constant-price run and the simultaneous-price-drop run rather than at a single price-quantity combination, therefore they are are elasticities. U.S. PEXD is calculated for each commodity by dividing percent change from the baseline (constant prices for the entire period) in U.S. net exports by twenty (for percent change) and multiplying the result by one hundred. Export supply elasticities for competing exporters are 66 calculated in a similar way. Elasticities of supply at the farm level are approximated by measuring percent change in harvested area for a percentage change in revenue. Short run export and import elasticities are calculated for the same year the changes occurred (1986) when demand responds, but production does not; and for the following year (1987) when production also adjusts to changes in price. Long run elasticities are calculated for 1996, ten years after the change in prices began. Elasticities are presented for soy meal and oil equivalent exports and imports and for the three soybean products individually. Except for the U.S., exports and imports are calculated on an equivalent basis. Meal and oil equivalents include a percentage of the whole beans as well as the actual meal or oil. The percentages are the amounts of meal and oil extracted when beans are crushed. For meal, 79.5 percent is used for all regions, while 17.5 percent is used for oil equivalent. The remaining three percent is waste. Table 3 presents PEXD for the U.S. and supply elasticities for other net exporters. It also shows percent changes in world trade, calculated in elasticity form. As a result of a twenty percent drop in price and revenue, world trade increased for all commodities except wheat. The unexpected behavior of wheat may be due to income effects. When prices are lowered in unison, animal protein becomes more affordable and consumption of wheat, an inferior good, declines as more meat is consumed. Secondly, more feedgrains are fed to animals than wheat when soybean meal price is low because the extra nutrition of wheat can be compensated for with more soybean meal. More specifically, wheat is substituted for feedgrain when soybean meal prices are high 67 TABLE 3: Price Elasticities for Exporters - Aggregate Run' Same Year Next Year Tenth Year Expected 1986 1987 1996 Sign Export Demand - United States Wheat +0.09 +0.30 +0.39 - Feedgrain -0.07 -0.16 -0.15 - Soybeans -0.05 —O.1O -O.95 - Soy heal -0.15 -0.23 +0.56 - Soy 011 -0.08 -0.90 +0.17 - Meal Equiv. -0.09 -O.13 -0.52 - Oil Equiv. -0.07 -0.19 -0.59 - Export Supply - Argentina Wheat -0.10 -0.31 -0.38 + Feedgrain +0.02 +0.11 +0.11 + Soybeans 0.00 +0.15 +2.63 + Soy Meal 0.00 -0.11 -3.83 + Soy Oil 0.00 -O.13 -3.98 + Meal Equiv. 0.00 +0.06 -O.3H + Oil Equiv. 0.00 +0.06 +0.35 + Export Supply - Australia Wheat 0.00 0.00 0.00 + Feedgrain +0.23 +0.19 +0.12 + Export Supply - Brazil Soybeans 0.0 +0.33 +1.08 + Soy Meal 0.0 +0.08 +0.68 + Soy Oil 0.0 +0.13 +1.25 + Meal Equiv. 0.0 +0.11 +0.71 + Oil Equiv. 0.0 +0.19 +1.22 + Export Supply - China Feedgrain +0.36 +0.45 +0.93 + Soybeans 0.0 -0.72 -0.A1 + Soy Meal 0.0 -0.06 +0.01 + Meal Equiv. 0.0 -O.33 -0.16 + Export Supply - Developed Markets Wheat +0.39 +0.23 +0.08 + Soy Oil +0.00 0.00 -0.02 + World Trade Wheat +0.07 +0.17 +0.19 - Feedgrain -0.03 -0.07 -0.0A - Soybeans -0.05 -0.06 -0.0A - Soy Meal -0.05 -0.05 -0.0A - Soy Oil -0.03 -0.15 -O.22 - Meal Equiv. -0.05 -0.05 -0.04 - Oil Equiv. -0.05 -O.11 -0.10 - *Aggregate Run = simultaneous 20% drop in all crop revenues 68 because of its higher nutrient content. In this case, as soybean meal becomes more affordable, more of it plus feedgrain is fed to livestock and less wheat is demanded. PEXD's for all commodities in the U.S. are less than one, even in the long run. Similar to world trade, U.S. wheat exports decreased in response to lower prices, but at twice the rate. This implies that the 0.8. loses market share in wheat. Supply elasticities are quite low in general. This low response to price change is consistent with low elasticities calculated from the harvested area equations. 0.8 and Canadian net exports are based on the pool of demand remaining after the other countries export, rather than on price. Canada is the exception to the residual supplier structure of the model. It is not considered a surplus exporter, but a contingent surplus exporter, which means it shares the residual pool of world demand with the U.S. Since Canada does not face perfectly elastic demand for its exports, as the other exporters are assumed to, movements in equilibrium quantities do not trace out the supply curve. Therefore, export supply elasticities for Canada are not meaningful and are not calculated. Changes in hectarage in response to changes in revenue are presented in Table A. Elasticities are calculated for the short run, the year following the change, and the long run, ten years later. Planting decisions in all regions are assumed to be based on revenues from the year before, hence no response is possible in the first year of the price change, 1986. In most cases elasticities are low and some are negative, opposite the direction that is expected. Aggregate elasticities for wheat and 69 TABLE 9: Revenue Elasticities of Supply at Farm Level - Aggregate Run' Next Year Tenth Year Expected 1987 1996 Sign United States Wheat -0.09 -0.08 + Feedgrain +0.13 +0.16 + Soybeans -0.06 -O.27 + Cropland Base +0.03 -0.02 + Argentina Wheat -O.12 -O.26 + Feedgrain +0.03 +0.09 + Soybeans +0.06 +0.33 + Cropland Base -0.02 +0.05 + Brazil Wheat +0.07 +0.11 + Feedgrain +0.13 +0.27 + Soybeans +0.09 +0.53 + Cropland Base +0.11 +0.37 + Australia Wheat 0.00 0.00 + Feedgrain 0.00 0.00 + Cropland Base 0.00 0.00 + Canada Wheat -0.25 -0.52 + Feedgrain +0.19 +0.99 + Cropland Base -0.11 -0.19 + Soviet Bloc Wheat 0.00 0.00 + Feedgrain 0.00 0.00 + Soybeans 0.00 0.00 + Cropland Base 0.00 0.00 + Developed Markets Wheat 0.00 0.00 + Feedgrain 0.00 0.00 + Soybeans +0.03 +0.12 + Cropland Base 0.00 0.00 + Newly Industrialized Countries Wheat -0.51 0.00 + Feedgrain +0.61 +3.72 + Soybeans +0.28 +1.67 + Cropland Base +0.97 +3.06 + 70 TABLE 9 (cont'd) Next Year Tenth Year Expected 1987 1996 Sign Oil-Exporting Low Income Countries Wheat -0.11 -0.11 + Feedgrain +0.10 +0.09 + Soybeans -0.01 -0.02 + Cropland Base +0.02 +0.01 + Low Income Countries Wheat -0.09 -0.35 + Feedgrain +0.03 +0.06 + Soybeans -0.09 -0.06 + Cropland Base 0.00 -0.08 + China Wheat +0.15 +0.21 + Feedgrain +0.19 +0.91 + Soybeans -0.93 -0.53 + Cropland Base +0.09 +0.21 + aAggregate Run = simulation of simultaneous 20% drop in revenues for wheat, feedgrain, soybeans and soybean products 71 for feedgrain in Australia and the Developed Markets are zero because these harvested area equations have been restricted to yield zero elasticities when prices change simultaneously. In each case there is insufficient information to specify a more informative equation. In the Soviet Bloc, harvested areas do not respond because they are functions of ending stocks rather than revenue. In these centrally planned economies, farmers are assumed not to respond to price incentives, but to government policy, represented by ending stocks. Cropland base in the U.S. holds essentially constant over the ten year forecast period. In fact, in most countries, cropland base responds very little to a simultaneous price change in the three crops. The exceptions are the NIC's, where cropland base is highly elastic, and Brazil and China, where moderately inelastic responses of .37 and .21 respectively, are observed for the long run. Production of Ag Model crops in the NIC's has trended downward over time as these countries rapidly industrialize. Reduced revenue in this scenario may speed the removal of cropland from production. Table 5 presents demand elasticities for importing regions. All regions showed inelastic demand for every commodity, even in the long run, with the exception of feedgrain imports in Brazil. These are very elastic for the entire period. While the Developed Markets export some soy oil, the region is a net importer of oil equivalent. Thus the oil equivalent elasticity appears under import demand. Very few unexpected signs appear among the demand elasticities and most of these are very close to zero. The two exceptions are wheat demand in the LDC's and China. The unexpected signs for wheat suggest that the prices of other 72 TABLE 5: Price Elasticities of Demand for Importers - Aggregate Run' Next Year Same Year Tenth Year Expected 1986 1987 1996 Sign Brazil Wheat -0.05 -0.07 -0.06 Feedgrain -3.19 -7.95 -1.78 Soviet Bloc Wheat 0.00 0.00 0.00 Feedgrain -0.03 -0.02 -0.01 Soybeans 0.00 -0.03 -0.02 Soy Meal 0.00 +0.01 +0.01 Soy Oil -0.27 -0.17 -O.33 Meal Equiv. 0.00 0.00 0.00 Oil Equiv. -0.06 -0.05 -0.09 Developed Markets Feedgrain +0.02 +0.02 +0.02 Soybeans -0.06 -0.06 -0.05 Soy Meal -0.06 -0.06 -0.05 Meal Equiv. -0.06 -0.06 -0.05 Oil Equiv. -0.08 -0.08 -0.06 Newly Industrialized Countries Wheat +0.06 +0.09 +0.01 Feedgrain -0.09 -0.09 -0.02 Soybeans -O.11 -0.10 -0.03 Soy Meal -0.30 -O.31 -O.21 Soy 011 -0.50 -0.68 -0.23 Meal Equiv. -O.16 -O.16 -O.11 Oil Equiv. -0.11 -0.12 -0.09 Oil-Exporting Low Income Countries Wheat -0.02 -0.01 -0.01 Feedgrain +0.01 -O.29 -0.27 Soybeans -0.16 -O.20 -0.12 Soy Meal -0.18 -O.17 -O.11 Soy 011 -0.16 -0.15 -O.12 Meal Equiv. -O.18 -O.17 -0.11 Oil Equiv. -0.15 -0.15 -O.12 Low Income Countries Wheat 0.00 +0.93 +0.69 Feedgrain 0.00 -0.03 +0.09 Soybeans 0.00 -0.01 0.00 Soy Meal 0.00 -0.01 0.00 Soy Oil 0.00 -0.21 -O.19 Meal Equiv. 0.00 0.00 0.00 Oil Equiv. 0.00 -0.19 -0.13 73 TABLE 5: (cont'd) Same Year Next Year Tenth Year Expected 1986 1987 1996 Sign China Wheat +0.92 +0.26 +0.12 - Soy 011 0.00 -0.19 -0.5 - Oil Equiv. 0.00 +0.09 -O.6O - 'Aggregate Run = simulation of simultaneous 20% drOp in revenues for wheat, feedgrain, soybeans and soybean products agricultural commodities and non-agricultural products may be important in determining demand for Ag Model commodities. Market shares for the major commodities are presented in Table 6 for each exporter. U.S. market share in wheat tends downward in the baseline, but accelerates with the simultaneous price declines of the scenario to lose an additional two percent of the market. The U.S. gains market share in feedgrain and soybeans over the baseline estimations during the entire forecast period. While U.S. soy meal and oil lose some market share, this loss is compensated for by the gain of over ten percent of the soybean market. This shift in soybean product markets corresponds to a loss of nearly twelve percent of the soybean market for Argentina, with a gain of ten percent in the meal and oil markets for Argentina. U.S. feedgrain gains and wheat losses of market share are distributed fairly evenly amongst the competing exporters with no shifts of more than one or two percent. Brazil, however, loses market share in all three soybean products. 79 TABLE 6: Market Shares (1) - Aggregate Run* Baseline Aggregate Run 1986 1996 1986 1996 U.S. Wheat 50.9 99.9 50.2 97.9 Feedgrain 73.7 79.1 79.3 75.7 Soybeans 71.2 68.9 71.5 81.3 Soy Meal 35.8 31.9 36.5 27.7 Soy 011 38.3 91.7 38.7 38.5 Argentina Wheat 6.8 6.6 7.0 7.6 Feedgrain 12.9 12.9 12.3 12.0 Soybeans 15.7 22.9 15.6 10.5 Soy Meal 10.6 13.9 10.5 29.9 Soy Oil 7.8 13.8 7.7 23.8 Australia Wheat 12.5 8.5 12.7 8.9 Feedgrain 5.0 5.6 9.7 5.9 Canada Wheat 25.5 29. 25.5 29.5 Feedgrain 5.9 5. 5.9 5.2 Brazil Soybeans 9.0 9. 8.9 3.1 Soy Meal 97.5 96. 97.0 90.1 Soy 011 19.3 23. 19.2 17.1 Developed Markets Wheat 9.9 10.9 9.6 10.6 Soy 011 18.2 13. 18.0 13.2 China Feedgrains 3.6 2.2 3.3 1.8 Soybeans 9.1 9.8 9.0 3.1 Soy Meal 6.1 7.8 6.1 7.8 Soy 011 16.9 7.5 16.3 7.9 *Aggregate Run = simulation of simultaneous 20% drop in revenues for wheat, feedgrain, soybeans and soybean products 75 5.22 Discussion In this study, elasticities are measured from decreases in product prices. However, the linear relationships in the structure of the Ag Model cause it to respond equally in magnitude to price increases or to price decreases. If Cochrane is correct in stating that supply responses to price increases are greater than responses to price decreases, then the Ag Model results would tend to overestimate the actual supply response to a drop in price. However, it is necessary to keep in mind that the elasticities are based on harvested area rather than supply and as such represent probable lower bounds to farm level supply elasticities. In some cases, shifts occur between the amounts of soybeans and soybean products traded. In Argentina, soybeans has a large negative PEXD, but meal and oil have unexpected positive elasticities. These contradictory results are due to the behavior of the percentage meal equation that determines the share of meal equivalent exports actually exported as meal. To reflect the limited capacity of the crushing industry in Argentina, soybean production appears in the equation with a negative sign. In years of high production, the country's crushing capacity would be exceeded and soybeans would be exported whole rather than as meal and oil. In this scenario where production declines in response to lower price, the percentage of meal equivalent exported as meal jumps and causes a reversal of sign in the elasticity for meal (and oil). This change in supply by Argentina may be sufficient to change the composition of the residual pool of demand for U.S. soybean products and increase demand for U.S. soybean meal and oil. 76 5.3 SINGLE COMMODITY PRICE ELASTICITIES It is common to measure elasticities for individual commodities in isolation, where own price is changed while other prices are held constant. This situation is simulated for wheat with wheat price, revenue and loan rate each reduced by twenty percent from the baseline (B1) levels in each year between 1986 and 1996. The prices and loan rates of feedgrain and soybeans are fixed, but all other endogenous variables in the model are allowed to adjust. The purpose of a single-commodity run (number B3) is to provide a basis of comparison between the results generated by the Ag Model and those of other studies that have used the single-commodity approach. Further, it enables comparison of Ag Model results between the aggregate and the individual commodity situations. The same types of elasticities as in the aggregate case are calculated. Percent changes are measured from the baseline constant-price run (B1) for U.S. PEXD, competing exporter supply, harvested area and import demand. Changes in market shares are also calculated. As in the aggregate run (B2), exchange rates are constant and inflation is zero in each country. 5.31 Results The exporter elasticities are presented in Table 7. The volume of world trade in wheat rises by one percent in the first year as a result of a twenty percent reduction in wheat price (elasticity of -0.05) and increases by five percent over the baseline at the end of the period ( -.25 elasticity). Feedgrain trade declines very slightly and soybeans and soy products show essentially no change during the entire period. 77 TABLE 7: Price Elasticities for Exporters - Single Commodity Run Same Year Next Year Tenth Year Expected 1986 1987 1996 Sign Export Demand - United States Wheat -O.10 -0.37 -0.51 - Feedgrain +0.16 +0.36 +0.27 + Soybeans 0.00 +0.07 +0.50 + Soy Meal 0.00 -0.02 -O.95 + Soy 011 0.00 -0.07 -0.68 + Meal Equiv. 0.00 +0.05 +0.08 + 011 Equiv. 0.00 +0.03 +0.12 + Export Supply - Argentina Wheat +0.08 +0.97 +0.96 + Feedgrain 0.00 -0.39 -0.90 - Soybeans 0.00 -0.10 -1.92 - Soy Meal 0.00 +0.07 +2.13 - Soy Oil 0.00 +0.09 +2.21 - Meal Equiv. 0.00 -0.09 -O.17 - Oil Equiv. 0.00 -0.09 -0.17 - Export Supply - Australia Wheat 0.00 +0.08 +0.60 + Feedgrain +0.01 -0.82 -0.69 - Export Supply - Brazil Soybeans 0.0 0.00 0.00 - Soy Meal 0.0 0.00 0.00 - Soy Oil 0.0 0.00 0.00 - Meal Equiv. 0.0 0.00 0.00 - Oil Equiv. 0.0 0.00 0.00 - Export Supply - China Feedgrain -1.29 -1.37 -1.91 - Soybeans -0.09 -O.73 -O.51 - Soy Meal 0.00 -0.06 -0.01 - Meal Equiv. -0.01 -0.39 -0.21 - Export Supply - Developed Markets Wheat +0.39 +0.23 +0.10 + Soy Oil 0.00 0.00 0.00 - World Trade Wheat -0.05 -0.18 -O.25 - Feedgrain +0.07 +0.10 +0.09 + Soybeans 0.00 0.00 0.00 + Soy Meal 0.00 0.00 -0.01 + Soy Oil 0.00 -0.09 +0.01 + Meal Equiv. 0.00 0.00 0.00 + Oil Equiv. 0.00 0.00 -0.01 + 78 The PEXD for U.S. wheat is -.10 in the same year, -.37 one year later and -.55 in the long run, ten years later. Other U.S. exports decline as they become more expensive relative to wheat. However, for all commodities the U.S. PEXD is inelastic over the entire period. Wheat and feedgrain respond slightly more strongly than in the aggregate run but soybean products are less responsive. The signs are as expected. While soy meal and oil move in the opposited direction from expectations, the shift is not significant because meal and oil equivalents are signed as expected. Supply elasticities are largely inelastic except for feedgrain in China, which is elastic even in the short run. Soybean products, in general, show very little response to a change in wheat price, being a more distant substitute for wheat than feedgrain is. Harvested area responses appear in Table 8. All harvested areas either respond in the expected direction or negligibly in the opposite direction. Wheat in both the NIC's and Brazil shows elasticities greater than one. The elasticity of feedgrain harvested area in the NIC's also exceeds one in the long run. Wheat harvested area declines steadily in the NIC's in the baseline and reaches a minimum before 1996. Thus the elastic response to lower wheat revenue is eliminated by that year in the scenario run. Change in cropland base in this single-commodity run is dominated primarily by the relative areas of wheat and the other crops. Only in the NIC's, where the response is determined by an elastic feedgrain harvested area, is cropland base response moderate in the long run, though still inelastic. Short run elasticities for cropland base are not greatly different from zero in value. 79 TABLE 8: Revenue Elasticities of Supply at Farm Level - Single Commodity Run Next Year Tenth Year Expected 1987 1996 Sign United States Wheat +0.11 +0.17 + Feedgrain -0.19 -0.19 - Soybeans -0.01 -0.09 - Cropland Base -0.09 -0.02 - Argentina Wheat +0.23 +0.26 + Feedgrain -O.27 -O.33 - Soybeans -0.03 -0.16 - Cropland Base -0.03 -0.08 - Brazil Wheat +0.58 +2.95 + Feedgrain -0.32 -0.18 - Soybeans 0.00 0.00 - Cropland Base -0.10 +0.10 - Australia Wheat +0.06 +0.92 + Feedgrain -O.65 -0.60 - Cropland Base -0.19 +0.02 + Canada Wheat +0.03 +0.08 + Feedgrain -O.13 -0.98 - Cropland Base -0.03 -0.19 + Soviet Bloc Wheat -0.03 -0.03 + Feedgrain +0.09 +0.03 - Soybeans 0.00 0.00 - Cropland Base +0.01 +0.01 - Developed Markets Wheat 0.00 0.00 + Feedgrain 0.00 0.00 - Soybeans 0.00 0.00 - Cropland Base 0.00 0.00 - Newly Industrialized Countries Wheat +1.75 0.00 + Feedgrain -0.20 -1.05 - Soybeans 0.00 +0.01 - Cropland Base -0.06 -0.71 - TABLE 8 (cont'd.) Oil-Exporting Low Income Countries Wheat Feedgrain Soybeans CrOpland Base Low Income Countries Wheat Feedgrain Soybeans Cropland Base China Wheat Feedgrain Soybeans Cropland Base Next Year 1987 +0.06 +0.01 0.00 +0.02 +0.09 -0.06 0.00 -0.03 +0.15 -0.21 -O.93 -0.06 Tenth J. 1996 +0.06 0.00 0.00 +0.02 +0.30 -0.09 0.00 +0.09 +0.21 -0.53 -O.53 -0.17 ‘7- 81 Table 9 contains demand elasticities for the importers under this scenario. Price elasticities of demand for importers are extremely inelastic. Only Brazil, LDC's and China respond even moderately. Brazilian feedgrain net imports, one of the worst equations in the model in terms of magnitude of error (Shagam, 1987), shows very elastic responses in both the aggregate and the single commodity runs. However, Brazilian imports are quite inelastic to price and depend largely on variations in domestic production. The Brazilian government is willing to import corn, even when the world price exceeds the domestic price, in order to ensure stable domestic prices (ERS, 1986). A further complication results from the fact that the mean level of imports for Brazil is slightly negative, Brazil is technically a net exporter of feedgrains over this sample period. Therefore, the interpretation of the import elasticities measured for Brazil is questionable. Table 10 presents market shares for each exporting region. The U.S. and Canada gain slight increases over the baseline forecast in their shares of the wheat market at the expense of other exporters. The U.S. gain in the wheat market is offset by loss of market share in feedgrain over the period. This loss is distributed approximately evenly across the competing exporters. In the soybean market, the U.S. loses several percentage points, but gained as much in the soy meal and oil markets. As in the aggregate case, this shift mirrors the behavior in Argentina's market shares. China gains slightly in soybean product markets and Brazil is completely unaffected by the wheat price drop. 82 TABLE 9: Price Elasticities of Demand for Importers - Single Commodity Run Next Year Same Year Tenth Year Expected 1986 1987 1996 Sign Brazil Wheat -0.05 -O.26 -O.62 - Feedgrain -0.09 +9.22 -0.39 + Soviet Bloc Wheat 0.00 +0.03 +0.03 - Feedgrain +0.18 +0.16 +0.09 + Soybeans 0.00 0.00 0.00 + Soy Meal 0.00 0.00 0.00 + Soy Oil -0.02 -0.02 -0.01 Meal Equiv. 0.00 0.00 0.00 + Oil Equiv. 0.00 0.00 0.00 Developed Markets Feedgrain 0.00 0.00 0.00 + Soybeans 0.00 0.00 0.00 + Soy Meal 0.00 0.00 0.00 + Meal Equiv. 0.00 0.00 0.00 + Oil Equiv. 0.00 0.00 0.00 Newly Industrialized Countries Wheat -0.19 -0.15 0.00 - Feedgrain 0.00 0.00 0.00 + Soybeans 0.00 0.00 0.00 + Soy Meal 0.00 0.00 +0.09 + Soy Oil 0.00 0.00 +0.03 Meal Equiv. 0.00 0.00 0.00 + Oil Equiv. 0.00 0.00 0.00 Oil-Exporting Low Income Countries Wheat -0.02 -0.09 -0.03 - Feedgrain +0.13 +0.17 +0.19 + Soybeans +0.03 +0.09 0.00 + Soy Meal 0.00 -0.01 0.00 + Soy Oil 0.00 0.00 0.00 Meal Equiv. 0.00 0.00 0.00 + 011 Equiv. 0.00 0.00 0.00 Low Income Countries Wheat 0.00 -O.35 -O.52 - Feedgrain 0.00 +0.12 -0.09 + Soybeans 0.00 +0.02 +0.01 + Soy Meal 0.00 +0.02 +0.01 + Soy Oil 0.00 0.00 0.00 Meal Equiv. 0.00 +0.02 +0.01 + Oil Equiv. 0.00 0.00 0.00 TABLE 9: (cont'd) Same Year 1986 China Wheat -O.20 Soy Oil 0.00 Oil Equiv. +0.01 83 Next Year 1987 -0-35 -O.13 +0.05 Tenth Year 1996 -O.39 -0.90 -O 027 Expected Sign 89 TABLE 10: Market Shares (8) - Single Commodity Run Baseline Wheat Run 1986 1996 1986 1996 U.S. Wheat 50.9 99.9 50.9 52.9 Feedgrain 73.7 79.1 72.9 70.7 Soybeans 71.2 68.9 71.2 62.0 Soy Meal 35.8 31.9 35.8 37.3 Soy 011 38.3 91.7 38.3 97.3 Argentina Wheat 6.8 6.6 6.6 5.7 Feedgrain 12.9 12.9 12.6 13.5 Soybeans 15.7 22.9 15.7 28.8 Soy Meal 10.6 13.9 10.6 8.0 Soy 011 7.8 13.8 7.8 7.7 Australia Wheat 12.5 8.5 12.9 7.2 Feedgrain 5.0 5.6 5.0 6.9 Canada Wheat 25.5 29.5 25.6 25.0 Feedgrain 5.9 5. 5.9 6.9 Brazil Soybeans 9.0 9. 9.0 9.0 Soy Meal 97.5 96. 97.5 96.8 Soy 011 19.3 23. 19.3 23.8 Developed Markets Wheat 9.9 10.9 9.5 9.7 Soy 011 18.2 13.2 18.2 13.1 China Feedgrains 3.6 2.2 9.6 3.0 Soybeans 9.1 9.8 9.1 5.2 Soy Meal 6.1 7.8 6.1 7.9 Soy 011 16.9 7.5 16.9 8.1 85 5.32 Discussion The single-commodity run of the Ag Model can be compared to other studies of PEXD for wheat. In the short run, the PEXD of -.10 is smaller than any estimate reported in Gardiner and Dixit's (1987) review of the empirical literature (see Table 11). However, the Ag Model's long run estimate of -.51 falls within the range of estimates in Gardiner and Dixit's study, —.23 to -6.72, albeit at the lower end. Table 11 contains the wheat PEXD's reported in Gardiner and Dixit. The farm level revenue elasticities of this study that appear in Table 8 are similar to supply elasticities reported by the IIASA model (Seeley, 1985). Similar magnitudes appear for Ag Model changes in harvested area due to revenue changes and for adjustments in production due to changes in the world price for the IIASA model study. The two measures are not identical, as previously stressed, but are analagous. Table 12 compares IIASA model results to Ag Model results from Table 8. For example, Seeley reports elasticities of 0.19 in the second year and 0-29 in the tenth yearZ for Argentine wheat, while the Ag Model gives 0.23 in 1987 and .26 in 1996. ZThe IIASA study's forecast period was 1985 to 1999, close to the present study's 1986 to 1996 period. By IIASA methods of calculating years, Ag Model values for 1996 are actually the eleventh year. TABLE 11: Price Elasticities of Export Demand for U.S. Study Konandreas and Schmitz (1978) Taylor and Talpaz (1979) Baumes and Meyers (1980) Gallagher, Lancaster, Bredahl and Ryan (1981) Chambers and Just (1981) Gadson, Price, and Salathe (1982) Morton, Devadoss, and Heady (1989) Conway (1985) Johnson (1977) Miller and Washburn (1978) Bredahl, Meyers, and Collins (1979) Burt, Koo, and Dudley (1980) Webb and Blakely (1982) Paarlberg (1983) Dunmore and Longmire (1989) Honma and Heady (1989) Liu and Roningen (1985) Gardiner (1986) 86 Period 1955-72 1960-79 1951-76 1960-79 1969(1) 1977(II) 1963-78 1962-79 1969(1)- 1977(11) 1970 base 1976 base 1972/73- 1975/76 1960-75 1980/81- 1982/83 1969-78 1989 base 1967-80 Method Estimation (OLS) Estimation (SUR) Estimation (OLS) Estimation (OLS) Estimation (SSLS) -.17 Estimation (OLS) Estimation (3SLS) -.l9 Estimation (SC) Calculation Calculation Calculation Calculation Calculation Calculation Calculation Calculation Calculation Calculation Wheat Elasticity Short Long run run -3,13 -- -.15 -- -035 -- -.91 -- -0.23 -.21 -- -.26 -.93 -- -6.72 -- —5.00 -- -1067 -2.50 -- -1.05 -- "" '1082 -- -.89 -.99 -- -- -2.30 -- -.81 87 TABLE 11: Price Elasticities of Export Demand for U.S. Wheat--Continued Study Period Kost, Schwartz, and Burris (1979) 1960-75 Holland and Sharples (1989) 1979/80- 1981/82 Green and Price (1989) 1986 Seeley (1985) 1985 base Ray and Parvin (1978) Holland and Sharples (1981) -- Method Simulation Simulation Simulation Simulation Synthetic Synthetic Elasticity Short Long run run -0.35 -O.35 -070 "" “'05“ -- -.81 -1.99 -050 -1050 -050 -- Source: Gardiner and Dixit, 1987. Not available. 88 TABLE 12: Comparison of Price Elasticities of Supply - IIASA Model and Ag Model Country Commodity Elasticity with respect to Supplied world price of wheat IIASA Ag Model Year 2 Year 10 Year 2 Year 10 Argentina Wheat +0.19 +0.29 +0.23 +0.26 Australia Wheat +0.35 +0.31 +0.06 +0.92 Canada Wheat +0.50 +0.93 +0.03 +0.08 Feedgrain -0.28 -0.90 -0.13 -0.98 Source: Seeley (1985) In comparison to the aggregate run, U.S. PEXD's are predictably larger in the single commodity run because of substitution effects between wheat and feedgrain. There is little interaction between wheat price and soybean supply in this run, where wheat prices are changed independently. CHAPTER VI. SUMMARY AND CONCLUSIONS In this chapter the results presented in the preceding chapters are briefly summarized and conclusions are drawn from the findings of the study. The objectives of the study are to determine the short and long run price elasticities of supply implied by the Michigan State University Agriculture Model for wheat, feedgrain and soybeans and to examine the effects of alternative specifications of price. 6.1 SUMMARY OF RESULTS Chapters 3 and 4 address the second objective, that of comparing alternative specifications of price to determine whether the results are consistent or whether choice of specification alters the results. Specifically, do the low elasticities generated by the Ag Model result from the chosen specification of price? The relationship of cropland base,1 as an aggregate supply variable, and a weighted average of gross revenue based on border prices, as the expected price variable, is examined in Chapter 3. In no country studied is the relationship between cropland base and revenue strong. Among the possible alternative specifications of the price variable are revenue calculated on internal prices rather than on border prices in order to avoid problems of imperfect price transmission and deflated gross revenue to account for deviations in input costs from the general inflation rate. 1 Sum of wheat, feedgrain and soybean harvested areas. 89 90 Estimation may also be improved by inclusion of cross revenues relevant to the particular country. Some of the alternative specifications mentioned above are tested in Chapter 4. Internal prices, prices of substitutes and production costs are introduced in the harvested area equations for Australia, a country where internal prices are based on world prices. While the t- statistic is the primary criterion for Judging performance, other measures of fit, such as adjusted R-squared and F-statistics are also used. As might be expected in an open economy, internal prices do not contain more information than world prices. Imperfect price transmission is therefore ruled out as a cause for low price response in this country. In order to account for variation from general inflation in production costs, revenue is deflated by an index of prices paid by producers. The results show no improvement, however. A third trial is the introduction of competing prices. wool is an enterprise that competes for land in much of the wheat and feedgrain area of Australia, but wool price is not significant in either the wheat or the feedgrain harvested area equation. Similarly for Argentina, internal and competing prices are tested in the harvested area equations. Internal prices for wheat and feedgrains do not show a stronger relationship to harvested areas than do border prices in Argentina where there is considerable government intervention in agriculture. In the case of soybeans where market intervention is less strong, internal prices improve the fit somewhat. However, the internal variables do not present any evidence that harvested areas are in fact price elastic. Prices of competing enterprises, specifically sunflowers, flax and beef are introduced with 91 mixed results. Sunflowerseed revenue is statistically significant in the wheat equation and flaxseed revenue in the feedgrain equation. For soybeans however, none of these competing revenues are significant. The price elasticity of export demand (PEXD) for the U.S. and import demand and export supply elasticities for regions other than the U.S. are measured in Chapter 5 for each commodity. Aggregate elasticities are calculated for a simultaneous and prOportional price drop for wheat, feedgrain and soy products. Aggregate elasticities are low for both supply and demand. When PEXD is calculated for a single commodity price decline, elasticities are still quite low, but generally more elastic than in the aggregate run. 6.2 CONCLUSIONS Four types of elasticities are measured in the Ag Model simulations. The two supply elasticities are the elasticity of harvested area (supply) with respect to revenue (price), measured for each region, and price elasticity of export supply for regions that compete with the U.S. 0n the demand side, price elasticity for imports is calculated for importing regions and PEXD is measured for U.S. exports to the rest of the world. Most of these elasticities calculated from the Ag Model, both of supply and demand, are quite inelastic. In part, the low elasticities result from the Ag Model's specification. First, at the farm level, supply is approximated imperfectly by harvested area, an input to supply rather than the output itself. Further, the yield component of supply is exogenized for simplicity and it is the actual harvested area, rather than the planted 92 area that is measured. Planted area would more accurately represent farmers' production intentions. Second, though the pattern of change in revenue per hectare is similar to that in price, it is not identical. The partial adjustment framework is intended to account for time lags in hectarage adjustment, but by no means does it represent expected price perfectly. These simplifications in specification of the supply and price variables cause the model to measure the supply elasticity imperfectly and may have reduced the estimated value. Third, many regions are insulated from the world economy by government policies, as illustrated in Chapter 3. The estimated elasticities for these insulated regions are likely lower than they would be with no government interventions, but the responsiveness of farmers in insulated regions to the actual prices they face is masked by the use of world price as a basis. Insulation from the world market dampens the response that is measured by an elasticity based on world price. Finally, the specification of some well-insulated regions does not include world prices at all, but policy variables, because the production and import decisions in these regions are policy-driven. For example, the Soviet Bloc showed no supply response to price because price variables are not included in the harvested area equations. Rather, these equations are based on lagged ending stocks, which are intended to represent government policy. Ending stocks, in turn, are based on the price of the crop as well as domestic supply. Because ending stocks and prices are fixed at particular levels in this study, the indirect link between price and harvested areas is broken and 93 harvested areas show no response to price in these scenarios. If price were not fixed, it would have an indirect effect on harvested areas in the Soviet Bloc. The aggregate elasticities presented in this study are lower than those frequently seen primarily because there are few estimations in the literature measured for several commodities at once, rather than the usual ceteris paribus analysis. A more general reason for low elasticities in comparison to other studies is that these elasticities measure net effects under imperfect market conditions. Government intervention prevents the world price from having a direct impact on farmers' decisions in many parts of the world. Exchange controls, tariffs and subsidies and other barriers often lie between the two prices. Government programs to control production also restrict farmers' responses. Many studies in the literature assume neoclassical free trade conditions (Gardiner and Dixit, 1987). For example, Tweeten (1967) excluded certain types of exports specifically to adjust for barriers to free trade. In contrast, the present study measures supply and demand elasticities with these barriers intact. The evidence presented in Chapters 3 and A supports the low elasticity measurements in Chapter 5. Harvested area does not respond strongly to changes in revenue. Harvested area in the aggregate, or cropland base, responds even less. To the extent that hectarage responds to changes in revenue, it is through substitution between crops rather than an adjustment in total area. Attempts in this study to more accurately formulate the price variable do not demonstrate an elastic relationship. If supply is unresponsive at the farm level (hectarage), 94 it is not surprising to find it inelastic at the national level in countries that do not hold large stocks. The PEXD's generated by the Ag Model are clearly inelastic. This result may be seen to imply that the decision to reduce grain prices on the world market in order to expand U.S. exports has two important consequences. The policy fails to dispose of grain stock surpluses because exports do not increase substantially. More importantly, farm revenues would suffer substantially if the loss were not made up with deficiency payments and other forms of farm income subsidization. Larger subsidies to offset lower market prices cause government expenses to increase substantially and may exceed the cost of storing the surplus grain that accumulated with high prices. APPENDICES APPENDIX 1 TABLE 13: Ag Model Regional Groupings Region Acronym Countries United States US United States Canada CA Canada Australia AU Australia Argentina AR Argentina Brazil BR Brazil Developed Markets DM United Kingdom, Belgium, Denmark, Netherlands, Finland, Luxembourg, Portugal, Ireland, Greece, Iceland, Austria, France, West Germany, Italy, Switzerland, Sweden, Norway, Malta, Spain, Japan, South Africa Soviet Bloc SB Albania, Bulgaria, East Germany, Hungary, Poland, Romania, Yugoslavia, Czechoslovakia, USSR China CH China Oil-Exporting LDC's 0P Algeria, Ecuador, Indonesia, Iran, Iraq, Libya, Oman, Saudi Arabia, Venezuela, Nigeria, United Arab Emirates, Kuwait Newly Industrialized NC Hong Kong, Singapore, Malaysia, Countries Taiwan, South Korea LDC's LD all others 95 APPENDIX 2 TABLE 1A: Price Elasticities of Import Demand Brazil Wheat Feedgrain Developed Markets Wheat Feedgrain Meal Equivalent Oil Equivalent Low Income Countries Wheat Oil Equivalent Newly Industrialized Countries Wheat Feedgrain Meal Equivalent 011 Equivalent Oil-Exporting Low Income Countries Wheat Feedgrain Meal Equivalent Oil Equivalent Soviet Bloc Feedgrain 011 Equivalent China Wheat Feedgrain ' Soybean Price '* Soy Oil Price Own Cross Sample Price Price Period -0.07 - '61-'83 +u.uo - '61-'84 +0.19*'* - '61-'84 - +0.05"n '61-'83 -o.1u* - '6A-‘83 -0.16' - '63-'83 -0.56 +0.97(F) '61-'8A -0.50*' - '65-'83 -o.31 +0.35(F) '69-'83 -0.18 - '61-'83 -O.71’ - '65-'83 -0.3Au - '65-'83 -O.13 - '62-'83 -o.99 +1.33(w) '61-'83 -2.33 (lagged feedgrain price) -1.88* - '6A-'81 -0.81*"" - '6A-‘81 -1.25 +1.36(W) '61-'82 -0.98”* - '6A-'83 -0.37 +0.86(F) '61-'83 -7.56 +7.39(W) '61-'8A "* EEC Producers' Wheat Price minus WOrld Wheat Price '**' Weighted Average of Cassava and Soy Meal Prices '**" Crush Margin = (Soy Meal Price '.795 + Soy Oil Price “.175 - Soybean Price) 96 APPENDIX 3 Cropland Base Estimation, Statistical Results Cropland base is estimated for Argentina, Brazil, Australia, Canada, the U.S. and the Developed Markets. Cropland base (the sum of wheat, feedgrain and soybean harvested areas in a region) is regressed in a partial adjustment framework on cropland base lagged one year and lagged gross revenue. The form of the equations is: CLB = f(CLB(-1), REV(-1)). where: CLB(-1) = cropland base, lagged one year REV zi [ HM * h-yr avg. yield(i) * Pw’M ] i=1 CLB CPI(j) REV = average weighted revenue, HA(i) = harvested area for cr0p i, CLB = cropland base, sum of the harvested areas, yield(i) = metric tons per hectare of crop i, Pw = world price, XR(j) = exchange rate for country/region j, CPI(j) = consumer price index, country/region j. The statistical results of the regressions are presented in the remainder of this appendix. 97 ..A 98 APPENDIX 3 (cont’d) SHPL 1961 - 2‘ Observations . Ls // Dependent Variabl. 1. CLEAR Cropland Base, Argentina 1951 VARIABLE COEFFICIENT STD. ERROR T'STAT. Z‘TAIL S16. ocean...sons-nuns.cuteness-333.3:-a...seeIssssssaaluasaaasuaa-eases: C 3066.5911 1591.3336 1.6232163 0.083 CLEAR(‘1) 0.6097961 0.1630335 5.551559“ 0.000 REV1‘1) ‘50‘.29227 676.03539 '0.7459554 0.454 R-squared 0.606177 Mean of dependent var 11756.63 Adjusted R-squared 0.566670 5.0. of dependent var 1955.566 5.3. of regression 1255.332 Sun of squared resid 34639595 Durbin-Ustson stat 2.565515 F-ststistio 16.16172 Log likelihood -204.2540 Rel1dual Plot obs RESIDUAL ACTUAL FITTED O “ufluuuuunuee.eeeeeeeeeeoeeeeeeeeete .0 O. .0 O. O. .6 .0 O. .0 O. .0 O. . .6 6. O. 0. 06 ee 0. .6 .6 .6 C where: CLBAR(—l) 1961 1962 1963 1964 1965 1966 1967 1966 1969 1970 1971 1972 1973 1974 1975 1976 1977 1976 1979 1960 1961 1962 1963 196A '159.376 '2222.49 2066.07 -229.237 '1600.92 16.5197 453.335 -7.62116 '275.655 '1965.61 -1393.65 1296.96 -616.907 -609.060 512.626 1670.54 -1940.20 1207.61 -1005.07 1099.39 939.535- 1562.03 663.020 246.251 9322.00 7552.00 10925.0 11039.0 9703.00 10455.0 11515.0 11902.0 12022.0 10555.0 9901.00 12015.0 11157.0 10642.0 11312.0 13111.0 11055.0 12394.0 11421.0 12957.0 1a137.0 15775.0 15599.0 15141.0 9451.35 10101.5 5557.93 11255.2 11503.9 10A39.5 11061.7 11909.5 12297.7 12310.5 11294.5 10719.0 11953.9 11051.1 10799.2 11250.5 12995.2 11166.2 12429.1 11917.5 13197.5 11195.0 15015.0 1d59s.7 = Cropland base, Argentina, lagged one year REV(-1) = Weighted average revenue per hectare, lagged one year 99 APPENDIX 3 (cont'd) SHPL 1951 - 24 Observations LS // Dependent Variable 1s CLBBR Ckopland base, Brazil VARIABLE CD ?FICIENT STD. ERROR T-STAT. 2-TAIL SIG. 1964 C 961.95325 729.41355 1.3155445 0.201 CLBBR(-1) 0.9654743 0.0469419 20.567452 0.000 REVl-1) 2.5004696 7.5912935 0.3293592 0.745 ' 83......388IOIIICCICIOICIIIII...III'38.8.38888883333388888I...III... R-squared 0.972924 Mean of dependent var 16531.75 Adjusted R-squared 0.970346 5.0. of dependent var 6244.760 5.5. of regression 1075.371 Sun of squared resid 24254675 Durbin-Uatson stat 2.263969 F-statistic 377.3044 Log likelihood ~199.9523 Residual Plot obs RESIDUAL ACTUAL FITTED II...IUUIOIOCIlflflliflltiiflflflIIICIIICIIII88.838888888318888.8.8.888... : = . l = . 1961 -531.473 7551.00 5512.47 : : o : : i 1952 -497.029 5252.00 5779.03 : : . : : : 1963 -612.433 5370.00 9152.43 : : : e : : 1954 295.595 9520.00 9224.41 : z- : : l 1965 -579.522 9545.00 10427.5 : : .: : : 1966 ~125.125 10309.0 10437.1 : : . : : : 1967 -253.530 10556.0 11139.5 : : o: : l 1966 -190.954 11414.0 11605.0 : : : . : : 1959 475.490 12644.0 12165.5 : : : .: z 1970 924.156 14297.0 13372.6 : : : . : l 1971 551.204 15515.0 14965.5 : e: : : : 1972 -1212.54 15215.0 16430.5 : : : : . : 1973 2395.53 15410.0 15014.4 : z e : l 1974 49.3905 19245.0 19195.6 : z : . : : 1975 675.290 20672.0 19995.7 : : : : . l 1975 1440.55 22723.0 21262.4 : :I : : : 1977 -976.125 22310.0 23256.1 : z o: : : 1976 —134.015 22514.0 22745.0 : z : : o : 1979 1371.74 24452.0 23050.3 : : o : : l 1960 -295.307 24615.0 24914.3 : . : : : l 1951 -1369.99 23500.0 25170.0 : o : : : : 1952 -1622.13 22410.0 24232.1 : z : : a : 1953 1351.63 24256.0 22904.4 1 8 0 i 8 I 1964 -624.099 24267.0 24691.1 where: CLBBR(—l) = Cropland base, Brazil, lagged one year REV(-1) = Weighted average revenue per hectare, lagged one year 100 APPENDIX 3 (cont'd) SHPL 1961 - 1954 24 Observations LS ll Dependent Variable Is CLBAU’ Cropland base, Australia VARIABLE COEFFICIENT STD. ERROR T'STAT. Z'TAIL SIG. asa...stuns-81.88.Issuassssanssaasssas:sssssassssssssssstanzas-sues: C 3346.6059 1402.7519 2.3571591 0.026 CLBAUI’1) 0.9061672 0.0777019 11.667573 0.000 REV1'1) -903.65241 391.66737 ‘2.3050150 0.031 R-squared 0.557947 Mean of dependent var 12903.56 Adjusted R-squared 0.677275 5.D. of dependent var 2657.197 5.5. of regression 1000.936 Sun of squared resid 21039330 Durbin-Uatson stat 2.036945 F-statistio 53.20561 Log likelihood -195.2607 Residual Plot obs RESIDUAL ACTUAL FITTED IIUIIIIIIIIIIIICIIICII.I.IIII..I.8.IIIIIIIIIIIIOIIIIIIIIIIIIIIBIIOII : : o : : : 1951 -324.055 5442.00 5766.09 : : o: : : 1962 -192.616 9094.00 9256.52 : : - : : : 1963 -473.643 9115.00 9591.54 : : ;. : : 1954 254.762 9750.00 9465.24 : : .: : : 1955 -154.050 9623.00 10007.1 : : : . : : 1956 519.391 11472.0 10552.5 : : : . : : 1957 335.155 11795.0 11450.6 : : : :0 : 1965 1151.27 14055.0 12904.7 : . : : : l 1959 -1390.31 12566.0 14256.3 : . : : : : 1970 ~2956.71 10719.0 13555.7 : 3 2° : : 1971 215.751 11554.0 11437.2 : . : : : 1972 -971.559 11522.0 12493.7 : : :0 : : 1973 263.945 12550.0 12355.1 : z o: : : 1974 ~250.955 11515.0 11566.0 : : : o : 1975 1035.00 12444.0 11406.0 : : : o : : 1975 490.745 12579.0 12366.3 : : : o: : 1977 507.746 14323.0 13515.3 : z o: : l 1975 -254.423 14944.0 15226.4 : : : . : : 1979 559.305 15350.0 14790.7 : z : o : : 1950 342.096 15569.0 15226.9 : : : . : : 1951 515.542 16716.0 16102.4 : o: : : : 1962 -1096.56 15014.0 17110.7 : x : z a : 1963 1553.74 16659.0 17035.3 : : a: : : 1954 -253.395 15179.0 15442.4 I where: CLBAU(-1) = Cropland base, Australia, lagged one year REV(—l) = Weighted average revenue, lagged one year He ‘9! l.‘ 101 APPENDIX 3 (cont'd) SHPL 1961 ' 1984 24 Observations L5 // Dependent Variable is CLBCA Cropland base, Canada VARIABLE COEEFICIENT STD. ERROR T’STAT. Z‘TAIL SIG. sac-Incas...ass-asuassscslasussasassass:ass:ssssssasaassssass33:38.3 C 7543.9846 3597.4436 2.0970404 0.045 CLBCAC‘1) 0.5414351 0.1905878 2.8406599 0.010 REV('1) 321.60681 “13.24810 0.7782415 0.445 R-squared 0.309641 Mean of dependent var 15535.53 Adjusted R-squared 0.243592 5.0. of dependent var 1537.540 5.5. of regression 1595.053 Sun of squared resid 53631253 Durbin-Uatson stat 2.004140 F-statistio 4.709472 Log likelihood -209.4596 Residual Plot obs RESIDUAL ACTUAL FITTED I . I : I 1961 ‘1575.32 15102.0 17575.3 I : I o: I 1962 1324.24 15072.0 16747.5 : : . : : 1953 21.5759 15252.0 15230.3 I : o : I 1964 -142.570 15355.0 15500.5 I : o : I 1965 -51.5626 16250.0 15361.9 I : I o : I 1966 930.474 19254.0 15353.5 I : Io : I 1967 415.061 19497.0 19051.9 I : I . : I 1965 706.762 19533.0 16526.2 I : I : I 1969 -962.245 17950.0 15932.2 I o : I : I 1970 -4965.93 13133.0 15101.9 I : I : I 1971 2072.55 17679.0 15606.5 ' : I : I 1972 -395.215 17545.0 17940.2 I : Io : I 1973 321.127 15493.0 15171.9 ' o I : I 1974 -1750.19 17539.0 19259.2 I : o I : I 1975 -625.024 17739.0 15364.0 I : I . : I 1975 552.112 19366.0 16513.9 I : 0 I : I 1977 -773.136 15351.0 19124.1 I z a I : I 1975 -536.592 16060.0 15596.6 I :n I : I 1979 -1353.45 17242.0 16595.5 I : I o : I 1950 647.416 16764.0 16136.6 I : I : I 1951 '2561.43 21592.0 19030.6 I : I . : I 1962 1053.92 21405.0 20321.1 I : I o: I 1953 1245.65 21496.0 20247.3 I : I . : I 1954 1002.12 21150.0 20177.9 Canada, lagged one year where: CLBCA(-1) = Cropland base, REV(-l) = Weighted average revenue per hectare, lagged one year 102 APPENDIX 3 (cont'd) SHPL 1965 - 20 Observations LS // Dependent Variable is CLSUS Cropland base, United States 1954 VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG. IOIIIUIIIICSC'IIIIIIIICIOIIIIIIUIIIIII.It...SIIIIIIIIICIIISIOIIOIIIO C 16555.975 12217.035 1.5159423 0.147 CLSUS(-1) 0.6533245 0.1479205 4.6195374 0.000 R5V(-1) 1515.0697 1132.5011 1.5025552 0.127 ..IICCIO..IIIIIUCIIC.II...IIICOCO-IICI8I...888.888888......IOISIOCI. R-squared 0.652119 Mean of dependent var 56594.55 Adjusted R-squared 0.511191 5.0. of dependent var 9931.662 5.5. of regression 5192.965 Sum of squared resid 6.520¢05 Durbin-Uatson stat 2.551705 F-statistio 15.93362 Log likelihood -201.3770 Residual Plot obs RESIDUAL ACTUAL FITTED IIIIIOOIIIIIICIIICIOIIIIICIIIIIIIIlflifli.ICIIIIIIIIIIIIICIIIII 88.... I : . I : I 1965 ~2303.61 72915.0 75215.6 : I I : I 1966 -2145.15 74505.0 76654.1 I : I . : I 1957 2553.50 50534.0 77750.2 I : I I : I 1965 -2654.53 76315.0 50970.5 I :I I : I 1969 -5073.50 74452.0 79535.5 I : . I : I 1970 -2305.99 74907.0 77213.0 I : I I : I 1971 2165.40 79506.0 77319.6 I :n I : I 1972 '5049.95 75553.0 60532.9 I : I . : I 1973 2359.03 55553.0 53324.0 I . I : I 1974 -5509.66 57599.0 93205.9 I : Io : I 1975 1413.79 92154.0 90740.2 I : . I : I 1975 -1534.59 91551.0 93295.6 I : I I : I 1977 2515.55 94345.0 91525.4 I : I I : I 1975 -1919.22 91411.0 93330.2 I : I I : I 1979 3657.55 96227.0 92569.4 I 8 II I I 1950 1239.55 97259.0 96019.5 I : I x I 1961 7393.96 102570. 95176.0 I : I I I 1962 5424.77 102590. 97165.2 I I 8 I x I 1953 -14594.4 52636.0 97332.4 I z I : I I 1954 14044.6 .96923.0 62576.2 CDC-II...-III...IIIIIOIIIIOIIOIIISIIIIII8.8.‘IIIII.IIIIIIIIIOIOCCICC where: CLBUS(-1) CrOpland base, United States, lagged one year REV(;1) = Weighted average revenue per hectare, lagged one year 103 APPENDIX 3 (cont'd) SHPL 1961 - 1953 23 Observations LS // D0P0fldifl‘ V‘fI‘bIO 1' CLBDH Cropland base, Developed Markets VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG. .-IIIIOIOI-I...ICIII...IIIIIIIIIIIII.8838883888....8.883.838.8838... C 33955.156 9963.4101 3.4052564 0.003 CLBDHI-i) 0.2703127 0.2146573 1.2592756 0.222 REVI-i) 37.161754 123.15756 0.3017416 0.755 III-IIOIOI.IIIIIICCIIIIICIIICIIIIII.IIIIIIIIIIUIIIIIIIIIIOII.8888... R-squared 0.062655 Mean of dependent var 46750.43 Adjusted R-squared -0.005659 5.0. of dependent var 524.5445 5.5. of regression 525.9633 Sun of squared resid 5553507. Durbin-Uatson stat 2.025625 F-statistio 0.903404 Log likelihood ~175.1723 Residual Plot obs RESIDUAL ACTUAL FITTED where: I I I I I I 1951 -525.079 45739.0 46564.1 I I I . I I 1952 370.227 45572.0 46501.5 I I I I : I 1953 -255.951 46570.0 45527.0 I I I I I I 1954 ~375.709 45354.0. 46732.7 I I I I I I 1955 ~243.393 45439.0 45552.4 I I I I I I 1955 -355.405 45329.0 45594.4 I I II I I 1957 -117.937 45555.0 45573.9 I : I I I I 1955 353.719 47105.0 45721.3 I I I I : I 1959 246.532 47119.0 45572.4 I I I I: I 1970 457.291 47355.0 45570.7 I : I : I I 1971 555.751 47529.0 45943.2 I I I I : I 1972 ~705.345 45301.0 47007.3 I I I I I I 1973 323.543 45995.0 45572.4 I I I I : I 1974 355.052 47331.0 45942.9 I I I I I I 1975 -329.539 45575.0 47004.5. I I I I I 1975 19.1525 45775.0 45755.5 I I I I I I 1977 -1347.50 45399.0 45746.5 I I I I I I 1975 355.055 45753.0 45375.9 I I II I I 1979 109.905 45559.0 45749.1 I I I I I 1950 40.5274 45510.0 45759.4 I I I I I I 1951 310.375 47075.0 45757.5 I I I I I I 1952 155.041 45954.0 45509.0 I I I II I 1953 554.555 47455.0 45503.3 I CLBDM(-1) = Cropland base, Developed Markets, lagged one year REV(—1) = Weighted average revenue per hectare, lagged one year APPENDIX 4 Raw Data from Sources Other than the Ag Model Table 15: Argentine Price Data obs UHEATP FEEDGP SOYP SUNFLP FLAXP BEEF? 3388333833833333383388833338833388333:3333333333=3==8==g==3333333333: 1960 0.003000 0.003542 NA 0.006100 0.006270 NA 1961 0.003936 0.004169 NA 0.007990 0.008040 NA 1962 0.005142 0.005727 NA 0.007710 0.009720 NA 1963 0.007190 0.008495 NA 0.012820 0.011910 NA 1964 0.007819 0.008179 NA 0.016450 0.012520 NA 1965 0.007540 0.010020 NA 0.015380 0.013400 NA 1966 0.010660 0.010400 0.016860 0.018520 0.017810 NA 1967 0.015840 0.013970 0.021690 0.019980 0.023430 NA 1968 0.015520 0.013820 0.028120 0.023500 0.028720 NA 1969 0.017410 0.017260 0.029510 0.028900 0.033080 NA 1970 0.017780 0.018130 0.031810 0.034420 0.028370 0.000102 1971 0.021570 0.019740 0.044180 0.051770 0.031880 0.000188 1972 0.038100 0.035560 0.096460 0.093750 0.076720 0.000301 1973 0.058790 0.055270 0.141900 0.116240 0.145450 0.000439 1974 0.070850 0.064260 0.167810 0.136310 0.207350 0.000431 1975 0.151080 0.106830 0.362580 0.185330 0.465160 0.000904 1976 0.849000 0.957500 3.474080 3.371830 4.275160 0.006205 1977 3.497250 2.940410 7.793000 8.412911 7.221660 0.016854 1978 9.274830 7.226000 15.39633 18.49983 14.94633 0.037111 1979 17.93841 13.73108 28.82633 32.95408 36.05116 0.120919 1980 33.99183 26.62258 41.98833 40.74175 48.02508 0.185893 1981 87.57050 56.27233 99.61950 130.6613 122.4258 0.334800 1982 296.7571 185.8124 383.3923 419.4101 449.1897 1.539900 1983 1040.300 1089.200 2112.900 2321.600 1995.000 6.690000 1984 6225.000 6968.000 10402.00 15115.00 NA 43.25000 '=.a83'3883.388'I3.83.33.‘..=333833338=:3383a:==:3=33:33:3833:33:38: Source: Bolsa de Cereales, 1984 WHEATP = Nominal wheat price, Australes/SOO kg. FEEDGP = Nominal feedgrain price, Australes/SOO k0. SOYP = Nominal soybean price, Australes/SOO kg. SUNFLP = Nominal sunflowerseed price, Australes/SOO kg. FLAXP = Nominal flaxseed price, Australes/SOO kg. BEEF? = Nominal beef price, Australes/IOO kg. 104 105 APPENDIX 4 (cont'd) Table 16: Brazilian Price Data ==3883832:88==============================32 obs WHEAT? FEED? SOYP 1960 16.00000 6.000000 NA 1961 22.00000 8.000000 NA 1962 40.00000 15.00000 NA 1963 64.00000 17.00000 NA 1964 139.0000 40.00000 NA 1965 191.0000 52.00000 NA 1966 254.0000 71.00000 122.0000 1967 302.0000 93.00000 155.0000 1968 365.0000 106.0000 208.0000 1969 437.0000 136.0000 251.0000 1970 476.0000 155.0000 329.0000 1971 539.0000 184.0000 383.0000 1972 584.0000 253.0000 484.0000 1973 736.0000 363.0000 1110.000 1974 1160.000 540.0000 1100.000 1975 2130.000 750.0000 1130.000 1976 2090.000 1410.000 2320.000 1977 3170.000 1380.000 2620.000 1978 3969.000 1966.000 3312.000 1979 5161.000 2968.000 5044.000 1980 10810.00 5780.000 8751.000 1981 NA NA NA 1982 NA NA NA 1983 NA NA NA 1984 NA NA NA ===:=========8===========================8== Source: FAO tapes WHEATP = Nominal wheat price, Cruzeiros/MT FEEDP SOYP Nominal feedgrain price, Cruzeiros/MT Nominal soybean price, Cruzeiros/MT APPENDIX 4 (cont'd) Table 17: Australian Price Data 106 8888328338338:3333882338338333338883====3===3===3====:====:3======2: obs UHEATP HARLEY PINDEX UOOLP GMINPU 888888883333888333888338888883'33388:383883833232:38::3333:3333:8:33: 1980 50.08000 40.30000 23.40000 95.64000 55.74000 1961 53.06000 46.70000 23.60000 99.45000 57.87000 1962 51.22000 47.60000 23.90000 108.3100 58.17000 1963 50.45000 48.30000 23.90000 128.0400 52.98000 1964 49.57000 49.70000 24.60000 105.4500 53.57000 1965 51.81000 50.50000 26.00000 110.4100 55.74000 1966 52.07000 52.80000 27.00000 104.4500 56.95000 1967 54.09000 50.50000 27.90000 92.04000 60.26000 1968 45.46000 42.80000 28.40000 98.48000 52.28000 1969 43.85000 38.90000 28.60000 82.78000 53.61000 1970 48.30000 47.10000 29.80000 64.68000 54.20000 1971 42.40000 40.50000 31.40000 75.25000 55.78000 1972 43.51000 52.60000 33.80000 183.7700 57.61000 1973 97.41000 79.50000 39.00000 181.1600 58.79000 1974 96.84000 102.1000 50.80000 126.9900 73.49000 1975 86.50000 98.70000 59.30000 143.2500 76.55000 1976 69.25000 103.5000 66.00000 182.7300 76.29000 1977 77.57000 86.00000 73.00000 187.1400 80.94000 1978 107.3500 82.00000 78.00000 205.2400 91.96000 1979 132.7200 121.0000 87.00000 243.5700 114.7100 1980 124.6800 142.0000 100.0000 255.9700 131.9200 1981 122.1900 134.0000 111.0000 264.6900 141.5500 1982 144.4700 167.0000 123.0000 269.8500 141.3200 1983 121.8100 151.0000 133.0000 293.8400 150.0000 1984 133.3900 141.0000 141.0000 318.6400 145.6400 88.888883883888833833888883383888338388833388833883888338:383388388: Sources: BAE Quarterly Rev. of the Rural Econ. a) Trends in the Aust. Rural Sector b) Hist. Trends in Aus. Ag. Prodn, Expts, Fm Inc. & Indexes of Prices Rec'd & Pd by Fmrs c) Commodity Stat. Bulletin d) The Wool Outlook e) Wheat Sit. and Outlook f) & Outlook g) Coarse Gr. Sit. WHEATP = Nominal wheat price, ADS/MT d), f) BARLEY = Nominal barley price, AUS/MT c), d), g) PINDEX = Index of Prices Paid by Farmers, 1980 = 100 a), b), c) WOOLP = Nominal price of greasy wool, AUS/HT d), e) GMINPW = Guaranteed minimum wheat price, AUS/MT d), f) APPENDIX 5 TABLE 16: Year 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1996 * Dollars per bushel Nominal Loan Rates’ Wheat .37 .25 .25 -35 .50 .00 .20 WWWCAWUJNNNN—I I comm OOU‘I 2.40 2.180 Corn 1.10 1.50 .00 .00 .10 .25 .40 .55 .65 .55 .55 1.92 NNNNNNNNN 1.92 Sorghum 1.05 1.43 1.90 1.90 2.00 2.14 2.28 2.42 2.42 2.42 2.42 1.82 1.82 107 Barley 0.90 1.22 1.63 1.63 1.71 1.83 1.95 2.08 2.16 2.08 2.08 1.56 1.56 Oats 0.54 0.72 1.03 1.03 1.08 1.16 1.24 1.31 1.36 1.31 1.31 0.99 0.99 Soybeans ummmmmmxwwmm O. O O. O. .0 NOOOOOOU‘IU‘U‘IWM NNNNNNNOOOOkH I: q .4 APPENDIX 6 Argentine Harvested Area Estimation, Statistical Results SMPL 1964 - 1984 21 Observations LS // Dependent Variable is UHAAR VARIABLE COEFFICIENT STD. ERROR T-STAT. Z-TAIL SIG. 8888888888833!!!88888883838888.8838888.8833833338888888338:833388333 C -1680.6308 2069.6244 -0.8121429 0.429 UHAAR(-1) 0.4093181 0.1714918 2.3868082 0.030 URREAL1-1) 3891.9137 1326.3794 2.9342386 0.010 NRREALi-i) -3482.7557 871.74228 -3.9951666 0.001 TIME 81.901073 25.439874 2.4332304 0.027 3.....883838888283888838888838.888838888888338888888888888383823883: R-squared 0.852335 Mean of dependent var 5241.714 Adjusted R-squared 0.565419 5.0. of dependent var 987.3547 5.5. of regression 650.8918 Sum of squared resid 6778561. Durbin-Watson stat 2.314513 F-statistio 7.505324 Log likelihood -162.9876 8.8888 388888888882888888888888888388838888333=333::3:33:33:8:233::= uses dbmestic prices and prices of substitutes SMPL 1964 - 1984 21 Observations LS // Dependent Variable is UHAAR VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG. 88888.888.888888888838833!88888388888388388838888388888833838.838:3: C -3774.4461 2373.1437 -1.5904836 0.131 UHAARi-1) 0.3794200 0.1669022 2.2733076 0.037 WRAR(-1) 1771.0655 662.23182 2.6743890 0.017 SRARI-i) -1167.0536 330.16420 -3.5347672 0.003 TIME 100.82243 32.482998 3.0976953 0.007 888.838...IflflilflflllllillIIllafl8......8.83888888888888888888888888888 R-squared 0.578651 Mean of dependent var 5241.714 Adjusted R-squared 0.473314 5.0. of dependent var 987.3547 S.E. of regression 718.5542 Sum of squared resid 8215199. Durbin-Uatson stat 2.304511 F-statistic 5.493324 Log likelihood ~165.0059 8.8888. 888.8 '88::3333 .88 .88388.88888388888883888838883838838888: original equafion u31ng or er prices where: WHAAR(-l) = Wheat harvested area, lagged one year WRREAL(—1) WY*WP, lagged WRAR(-1) = WY*Pw, lagged NRREAL(-1) NY*NP, lagged SRAR(-1) = SY*Ps, lagged TIME = Time trend, 1964 = 64 ‘ WY, NY, SY = 4-year moving average yields, wheat, sunflower- seed, soybeans WP, NP Pw, PS Real domestic prices, wheat, sunflowerseed Real border prices, wheat, soybeans 108 109 APPENDIX 6 (cont'd) SHPL 1964 - 1984 21 Observations . LS // Dependent Variable is FHAAR Feedgrain harvested area, Argentina CUISIIIIIUIIIIIBIISSISSS8.33888388888838833383383383883383333::3883: VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG. 88888888888883.8888...88388333388388888888883888888I!88888:838:88=2: C 5404.2448 1052.9481 5.1324893 0.000 FHAAR(-1) 0.4697474 0.1386605 3.3877523 0.004 LRREAL(-1) -905.64834 403.79940 -2.2428174 0.039 URREALI-i) -2153.1459 668.43913 -3.2211548 0.005 DV7179 -1449.6922 307.0292? -4.7216743 0.000 8.8.8.38888888388888888.888.8388888838888888388388383888883388883833 R-squared 0.735075 Mean of dependent var 5972.429 Adjusted R-squared 0.668844 5.0. of dependent var 667.3186 5.5. of regression 384.0161 Sum of squared resid 2359494. Durbin-Uatson stat 1.510176 F-statistic 11.09863 Log likelihood -151.9068 8888.88.8883888888888888.833833888883888388338332:883:83:328:3:3::2: uses doemstic prices and prices of substitutes SMPL 1964 - 1984 21 Observations LS // Dependent Variable is FHAAR VARIABLE COEFFICIENT STD. ERROR T-STAT. Z-TAIL SIG. 38888883883838.88883888838888388883...88.888333833888888838338833888 C 4145.8615 981.77240 4.2228336 0.001 FHAAR(-1) 0.4288164 0.1751541 2.4482240 0.026 FRARi-i) 868.88783 534.12978 1.6263236 0.123 URARi-i) -1643.3118 603.44460 ~2.7232190 0.015 DV7179 -1595.1059 333.02915 -4.7896885 0.000 88.88.8888..-8.88.88.88.883883888888.8883388838888888.88883888888838 R-squared 0.890692 Mean of dependent var 5972.429 Adjusted R-squared 0.813368 5.0. of dependent var, 667.3186 S.E. of regression 414.9383 Sum of squared resid 2754780. Durbin-Uatson stat 1.518843 F-statistic 8.932114 Log likelihood ~153.5331 ' origina equation using border prices where: FHAAR(-1) = Feedgrain harveested area, lagged one year LRREAL(-1) = LY*LP, lagged FRAR(-1) = FY*Pf, lagged WRREAL(-1) = WY*WP, lagged WRAR(-l) = WY*Pw, lagged DV7179 = Dummy variable to account for adverse weather 1, 1971, 1979; otherwise, = 0 LY, FY, WY = 4-year moving average yields, flax, feedgrain, wheat LP, WP = Real domestic prices, flaxseed, wheat Pf, Pw = Real border prices, feedgrain, wheat 110 APPENDIX 6 (cont'd) SMPL 1967 - 1984 18 Observations LS // Dependent Variable is SHAAR Soybean harvested area, Argentina 8838833888888888838.88388833833833883883=3=====3===S===3::3338:8283: VARIABLE COEFFICIENT STD. ERROR T-STAT. Z-TAIL SIG. 3383383883838838388338333883338833=8338883833:383388338338223833:3:3 C -587.95523 259.67799 -2.2641704 0.041 SHAAR(-1) 0.9350342 0.0876551 10.667194 0.000 SRREAL(-1) 234.51682 90.949000 2.5785530 0.023 HRREAL(-1) 528.98212 372.95689 1.4163465 0.180 SPL76 3.6976303 1.9323527 1.9135380 0.078 8.88.38.8838833888833383838333883.383888838888333833382883328233233: R-squared 0.983362 Mean of dependent var 1060.278 Adjusted R-squared 0.978243 5.0. of dependent var 1074.041 S.E. of regression 158.4232 Sum of squared resid 326272.8 Durbin-Uatson stat 2.391077 F-statistic 192.0907 Log likelihood -113.7869 8333333328883838833338333383833383:8333332323:3::8333333223323:222:: uses domestic prices and prices of substitutes SMPL 1967 - 1984 18 Observations LS // Dependent Variable is SHAAR VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG. 33338883883.883888883888888888338882833338383383888888883382338:888: C -119.86270 130.54064 -0.9182022 0.375 SHAARI-i) 0.9584958 0.0818427 11.711437 0.000 SRAR(-1) 201.12661 75.460463 2.6653236 0.019 WRAR(-1) -203.10784 156.32800 -1.2992416 0.216 SPL76 3.4224289 1.8694793 1.8306857 0.090 I.88888.8.88338888.8888883888888888838888838383388838888883383!!!833 R-squared 0.982233 Mean of dependent var 1060.278 Adjusted R-squared 0.976766 5.0. of dependent var 1074.041 S.E. of regression 163.7128 Sum of squared resid 348424.2 Durbin-Uatson stat 2.289665 F-statistic 179.6717 Log likelihood -114.3781 original equation using border prices where: SHAAR(-1) = Soybean harvested area, lagged one year SRREAL(—1) SY*SP, lagged SRAR(-l) SY*Ps, lagged WRREAL(-1) WY*WP, lagged WRAR(-1) WY*Pw, lagged SPL76 = Splined time trend, 1967-75 = O, 1976 = 76 SY, WY = 4-year moving average yields, soybeans, wheat SP, WP = Real domestic priCes, soybeans, wheat Ps, Pw = Real border prices, soybeans, wheat APPENDIX 7 Usefulness of the Ag Model for Policy Analysis Econometric models are built for various purposes - forecasting, scenario analysis and as tools for understanding. Very few models as large as the Ag Model are built. Because the large models are built for different reasons with different structures, estimation techniques and types of data, there is little basis for comparison between existing models. Econometric models are frequently judged on how closely the results of ex post forecasts compare to actual values. In the case of this study, elasticities are calculated ex ante and can only be compared to other estimates and the general consensus of what the true values may be. As a basis for assessing the validity of the model, the structure of the model itself has been examined to search for obvious sources of error. The structure of the Ag Model tends to concentrate error in certain equations. The error accumulates, first because the equations within each region are solved recursively. Error in the estimate of harvested area is transferred to the production equation which depends upon harvested area. In a similar manner, ending stocks are estimated as a function of the estimated production value. Each equation builds upon the solution values of previous equations and is subject to the error of those previous estimates. Secondly, the fact that the U.S. is considered a residual supplier means that U.S. export and ending stock equations are functions of the net trade of all other regions. Error in each of the net export and import equations is accumulated in the 111 112 residual U.S. equations. As a result, the price equations, dependent on ending stocks, receive the accumulated error of all the equations in the model. The error in the price equations is then transmitted to the harvested area, import and consumption equations in the subsequent year. One advantage of using a structure with a residual equation (determined as the sum or difference of the other variables) is that it provides closure to the system - the quantities produced, traded, consumed or stored are internally consistent. However, the residual variable also contains the largest amount of potential error. Vhile multiple errors may cancel out, the amount of error that remains cannot be measured statistically. It is apparent that scenarios in which the variables of interest are residual variables would be subject to the largest error, but in the case of the Ag Model the simultaneous nature of residual quantities and price transmits the accumulated error throughout the model. Therefore, the accumulated error is distributed through all estimated quantities and prices in an unmeasurable way. Testing the model's robustness with respect to error in price estimates would indicate the severity of the impact of price error on other estimated quantities. Unfortunately, because this study exogenized price, robustness could not be tested by measuring elasticities. A more straightforward method of dealing with the unavoidable error in the Ag Model would be to estimate the residual variables - net exports, consumption or ending stocks - and report a statistical discrepancy without attributing the entire error to any one equation. However, the purpose of constructing multi-equation regions within the Ag Model is to incorporate more of the variables that affect domestic 113 supply and demand in each region. A net export equation that is estimated directly accounts for price, domestic supply and consumption, but fails to capture all of the shifters of domestic supply and demand which influence exports indirectly. A major difficulty in measuring PEXD is that direct estimations do not account for the many relevant forces affecting net exports or imports while complex models contain accumulated errors and other problems (Gardiner and Dixit, 1987). A model is necessarily a simplification of reality. The Ag Model simplifies global agricultural commodity trade by considering only three crop groups, by aggregating the world into eleven regions and by using a common world price to link these regions in trade. The economy is further simplified by disregarding differences in transportation costs and specific local conditions, such as internal prices or relevant competing crops. From the experimentation in Chapter 4 with more specific, local variables it appears that the inclusion of more detailed specifications within the present model structure would not improve its predictive ability appreciably. In order to significantly improve the model's forecasting capability, detailed modelling of each region with much more specific data would be necessary. On the other hand, in comparison to other models used to estimate PEXD, the Ag Model is very complex. Estimation of PEXD frequently uses a model with only a few equations that measure net excess supply or demand for the rest of the world or for only a few regions. When combined with a U.S. excess supply equation, these equations yield an estimate of PEXD. The additional equations in each region and the large number of regions in the Ag Model enable it to account for many more of 114 the factors that determine PEXD than a simple model is capable of. If the additional error from the extra equations does not overwhelm the estimate, the size and complexity of the Ag Model is an asset in obtaining estimate of PEXD and other relevant economic measures. As future research, it may be possible to devise some method of accounting for accumulated error in the residual equations and adjusting the variables or the interpretation accordingly. Uith corrections for error, the reliability of the Ag Model may be improved sufficiently to produce credible forecasts. Uhile indeterminate levels of error may leave the Ag Model's forecasting capabilities in question, a major purpose for constructiong the Ag Model is to gain a greater understanding of the world economy. The validity of the Ag Model for this purpose remains intact. Despite our inability to measure its error, the Ag Model remains a useful tool for policy analysis. Some mechanical limitations of the model do exist, however, and probably deserve further attention. The fact that the soy oil price equation is exogenous removes some of the Ag Model's power in determining the other prices since soybean price is directly calculated from soy meal and oil prices. An exogenous soy oil price is no limitation for the PEXD runs in this study because all prices are exogenized. However, it is impossible to accurately determine the effect of changing the loan rates because soybean price is based on soy meal price, which responded to lower grain prices, and on soy oil price, which did not. A potential problem exists with the percent meal (pmeal) equations. Imports and exports are estimated on an equivalent basis, then the pmeal 115 equations determine the breakdown between beans and oil and meal. In some cases, particularly where the equation contains a time trend, the estimated value may exceed one or be less than zero. The result is that exporters begin to import and importers may export soy products that they cannot produce. In other cases, a small error in the pmeal equation may lead to large shifts in the trade of the three products as seen in this study for Argentina. The problems with this equation do not seem significant enough to invalidate the entire model, but may warrant additional scrutiny of unusual shifts in the mix of soy products. BIBLIOGRAPI—I BIBLIOGRAPHY Abbott, Philip C. "Estimating U.S. Agricultural Export Demand Elasticities: Econometric and Economic Issues." paper presented at IATRC Symposium on Elasticities in International Agricultural Trade. Dearborn, Michigan. July 30-August 1, 1987. . "Modeling International Grain Trade with Government Controlled Markets." Amer. J. Aer. Econ. 81 (1979):22-31. Armstrong, J. Scott. Long-Range Forecgsting Frog; Crygtgl Ball to Computer. John Wiley & Sons. New York. 1985. Arnade, Carlos, and Cecil Davison. "Export Demand for U.S. Corn, Wheat, and Soybeans.” FAER Draft. USDA Econ. Res. Serv. 1986. Austin, Lynn A. ”Beef versus 9001 Production in Australia." FAER No.130. USDA, Econ. Res. Serv. 1977. Beattie, Bruce R. and Robert C. Taylor. The Economics of Production. John Uiley and Sons, New York. 1985. Binkley, J. K., and L. D. McKinzie. "A Monte Carlo Analysis of the Estimation of Export Demand.” Can. J. Econ. 29 (1981):187-202. Bolling, H. Christine. "Price and Exchange Rate Transmission Revisited: The Latin America Case.” paper presented at IATRC Symposium on Elasticities in International Agricultural Trade. Dearborn, Michigan. JulySO-Augusti, 1987. Bolsa de Cereales, Bolsa de Cereals, Numero Estadistico. Buenos Aires. 1984. Bredahl, Maury E., William H. Meyers, and Keith J. Collins. ”The Elasticity of Foreign Demand for U.S. Agricultural Products: The Importance of the Price Transmission Elasticity.” Aggr. J. Agr. Econ. 69 (1979):58-63. Bredahl, Maury E., and Leonardo Green. ”Residual Supplier Model of Coarse Grains Trade." Agar. J. Agr. Econ. 65 (1983):785-90. Brandow, G.E. "Interrelationships Among Demands for Farm Products and Implications for Control of Market Supply.” Pennsylvania Agricultural Experiment Station. Bulletin No. 680. University Park. August, 1961. 116 117 Bureau of Agricultural Economics, BAE Trends in the Australian rural sector. Canberra, Australia. Various issues. Commodity Statistical Bulletin. Canberra, Australia. 1986. . Historical Trends in Australian Agricglturgl Production, Exports, Farm Incomes and Indexes of Pricgs Received and Paid by Farmers. 1952-53 to 1978-79. Canberra, Australia. 1980. Quarterly Review of the Rural Economy. Canberra, Australia. Various issues. . Coarse Grains Situation_gnd Outlook. Canberra, Australia. Various issues. ”Uheat Marketing in Australia: An Economic Evaluation.” Occasional Paper No.86. Canberra, Australia. 1983. . Uheat Situation ,ggd Outlook. Canberra, Australia. Various issues. . _hg Vool Outlook. No. 31. Canberra, Australia. Dec. 1971. Burk, Marguerite. 'Ramifications of the Relationship Between Income and Food." Journal of Farm Economics. February, 1962. Chaing, Alpha C. Fundamental Methods of Mathematical Economics. 2nd ed. McGraw-Hill Book Co., New York. 1974. Chambers, Robert G. and R. E. Just. "A Critique of Exchange Rate Treatment in Agricultural Trade Models.” Aggr. J. Agr. Econ. 61 (1979):249-257. Cochrane, Hillard V. The City Man's Guide to the Farm Problem. University of Minnesota Press, Minneapolis. 1965. Farm Prices - Mythvggnd Reality. Univ. of Minnesota Press, Minneapolis. 1958. Daly, R. F. “Demand for Farm Products at Retail and the Farm Level.” Journal of the American Statistical Association. September, 1958. Economic Research Service. "The Agricultural Situation in the Far East and Oceania." FAER No.121. USDA, Econ. Res. Serv. 1976. "Agricultural Trade Responsiveness in Western Hemisphere Countries." ERS Staff Report No. AGE5860326. USDA, Econ. Res. Serv. April, 1988. "Latin America, Outlook and Situation Report." USDA, Econ. Res. Serv. Various issues. 118 Food and Agriculture Organization, United Nations. "Farm Price Data Base.” Unpublished data available on USDA TDAM computer system and computer tapes from FAO. Friend, Reed E., E. Wayne Denney, Mary E. Long, and Thomas A. Twomey. "Australia: Growth Potential of the Grain and Livestock Sectors." FAER No.60. USDA, Econ. Res. Serv. 1972. Gardiner, Walter H., and Praveen M. Dixit. "Price Elasticity of Export Demand: Concepts and Estimates." FAER No. 228. USDA, Econ. Res. Serv. February, 1987. Gilmour, Brad and Peter Fawcett. "The Relationship Between American and Canadian Wheat Prices.” Agriculture Canada. Working Paper. September, 1986. Goodloe, Carol A. "North America and Oceania: Outlook and Situation.” RS-84-1. USDA, Econ. Res. Serv. March, 1984. pp. 26-28. Huang, Kuo. "U.S. Demand for Food: A Complete System of Price and Income Effects.” ERS Technical Bulletin No. 1714. Econ. Res. Serv. December, 1965. International Wheat Council. World Whggt Statistics. London, England. Various issues. Jabara, Cathy L. and Alan S. Brigida. "Variable Levies: Barriers to Grain Imports in France, the Netherlands, Federal Republic of Germany, and United Kingdom." FAER No. 156. USDA, Econ. Res. Serv. March, 1980. Johnson, Paul R. "The Elasticity of Foreign Demand for U.S. Agricultural Products." Amer. J. Agr. Econ. 59 (1977):735-36. Judge, George G., W.E. Griffiths, R. Carter Hill Helmut Lutkepohl and Tsoung-Chao Lee. The Theory, and Prgctice of Econogetricg. 2nd edition. John Wiley and Sons. New York. 1985. Kmenta, Jan. Elements of Economgtrics. 2nd Edit. Macmillan Publishing Co. New York. 1986. Knutson, Ronald D., James V. Richardson, Danny A. Klinefelter, Mechel S. Paggi and Edward G. Smith. Policx Tools for U.S. Agricultggg. Agricultural and Food Policy Center, Texas A&M University. No. B- 1548. August, 1986. MacGregor, R.J. and Kulshreshtha, S.N. "Pricing in the International Coarse Grainse Market." Can. J. Agr. Econ. 28 (1980):110-118. Martin, Lee R., ed. A Survey of Agricglturgl Economics Literature. Vol. 1. Univ. of Minn. Press. Minneapolis, MN. 1977. 119 McCaIla, Alex F. ”A Duopoly Model of World Wheat Pricing." J. Farm Econ. 48 (1966):711-27. Meilke, Karl D. "A Comparison of the Simulation Results from Five International Trade Models.” paper presented at International Trade Consortium, Vancouver, B.C. Dec. 1985. Meilke, Karl D. and A.C. Zwart. ”Effects of Alternative Specifications on the Policy Implications of a Model: An Illustration.” Can, J. of Ag. Econ. 27 (1979):15-22. Mitchell, Donald 0. "Global Rice Model: Conceptualization and Design." Washington, DC. 1983. Myers, Robert. Personal communication. 1986. Nerlove, Marc. ”Distributed Legs and Estimation of Long Run Supply and Demand Elasticities: Theoretical Considerations." Journpl of Farm Economics. 40 (1958):301-311. Peterson, Willis L. "The Market as an Experiment: Problems of Data Generation." Staff Paper P81-17. Univ. of Minn., Dept. of Agr. and Applied Econ. 1981. Pindyke, Robert S. and Daniel L. Rubinfeld. Econometric Models and Economic Forecasts. 2nd edition. McGraw-Hill Book Co. New York. 1981. Roe, Terry, Mathew Shane, and De Huu Vo. "Price Responsiveness of World Grain Markets." ERS Technical Bull. No. 1720. USDA, Econ. Res. Serv. June, 1986. Schuh, G. Edward. "The Exchange Rate and U.S. Agriculture." Amgr. J. Agr. Econ. 58 (1974):1-13. Schuh, G. Edward. in U.S. Senate Committee on Agriculture, Nutrition and Forestry. ”Farm Policy Perspectives: Setting the Stage for 1985 Agricultural Legislation.” Committee Print 98-174. Washington, DC. April, 1984. pp. 47-61. Seeley, Ralph. "Price Elasticities from the IIASA World Agriculture Model.” ERS Staff Report No. AGE5850418. USDA Econ. Res. Serv. May, 1985. Shagam, Shayle D. Specification and Structure ofypn International Trade Model with Linkage Capability. M.S. Thesis. Michigan State Univ. 1986. Spriggs, John. "An Econometric Analysis of Export Supply of Grains in Australia." FAER No.150. USDA, Econ. Res. Serv. 1978. Tweeten, Luther. ”The Demand for United States Farm Output." Food Research Institute Studies. 7 (1967):343-69. 120 Wainio, John T. "Argentina: Farmers' Response to Grain Prices.” ERS Staff Report AGE5830726. USDA, Econ. Res. Serv. 1983. Waugh, Fred V. ”Demand and Price Analysis." USDA Technical Bulletin No. 1316. Washington, DC. 1964. Wilde, Linda Chase, Laurence D. Cornell, Vernon Sorenson, and J. Roy Black. "World Grain Trade: An Evaluation of Factors Affecting Net Import Demand for Wheat and Coarse Grains by Selected Countries.“ Ag. Econ. Report No. 487. Michigan State Univ. July, 1986. Williams, Gary W. and Robert Thompson. "The Brazilian Soybean Industry: Economic Structure and Policy Interventions.“ FAER No. 200. USDA Econ. Res. Serv. October, 1984.