“ " -,- .1 ~I A‘- --I u < . ~. A-.-I'< ‘I. IL‘HI - 'I «v' I' I L‘.‘,‘A'L~ I“ ~.,.‘v,. I27-.‘: ..~. F.1“.1._'FV,I_\.‘ f.‘ I I IMPLICATIONS 0F STRUCTURAL CHANGE IN U. S DEMAND" FoR MEATS 0N U. s. LIVESTOCK ANU GRAIN MARKETS j} Dissertation for the Degreeof :Ph. D} ' U , MICHIGAN STATE UNIVERSIFII ' LAURENCE DAVID coRNELL 1983 LIBRf‘RY MichigaI‘I State University “w .. -.. h.— .-—-——-, 'm PLACE IN RETURN BOX to TO AVOID FINES MAY BE RECALLED remove this checkou return on or before with earlier due date t from your record. date due If requested. 6/01 c:/CIRC/DateDue.p65—p.15 , U ”RUQZIU‘I. '1 '0 F09 «Liar "K ‘IE’ :kAI’G "1 I . | ' "-\-r A, .‘Of Iarl‘e‘. F I. . \. _:\‘ I try, 0 asthma-- - n. m u. ~.~;.~. . .{ .N {3.3, cg" .‘i I 11 on 8.3. mg: K. .3 t-- 10“ ‘. .v‘w -—-‘~< j. .n engages .F ABSTRACT IMPLICATIONS OF STRUCTURAL CHANGE IN U.S. DEMAND FOR MEATS ON U.S. LIVESTOCK AND GRAIN MARKETS by Laurence David Cornell The U.S. beef market and indeed the market for other major meats, pork and poultry, have been subject to substantial structural change over the past several decades. Within this background of considerable change in the U.S. market for meats, structural models of retail demand were estimated. The broad objectives of this study were (a) to test the assumption, underlying the classical model of consumer demand, of con- stancy of structural coefficients in the retail demand for major meats in the U.S.: (b) to investigate parameter variation in terms of sys- tematic behavior over time; (c) to relate systematic variation of estimated slope coefficients and estimated measures of responsiveness of demand to observed structural changes in retail meat demand; and (d) to incorporate estimated models of retail demand into a model of U.S. agri- culture and to simulate the long-term impacts of recent structural changes on U.S. agriculture and international feedgrain markets. and particularly on U.S. imports of beef. Structural changes in regression coefficients were identified and quantified using graphical analysis, qualitative shift and interaction variables, linear and cubic spline functions within a discontinuous time-varying switching regression model, and Legendre polynomials within )“po- ( ‘ .0 :aol “ {'I-‘ it "r" N". In- .D" . lul to O- V. o - «(II V. ....- .. n 3' u a». .t. I. I I'-- he 'l\ A...” _ I .v IOl'a- _, ‘I n," a. .,l’ f ‘ or .' ' " o W 3.2 q -. a .., n "o‘..\ 2' .n Q ~- w“ H._ I u e I. . . k 0. ' ' —— Laurence David Cornell a continuous time-varying parameter model. A major conclusion of this study, supported by considerable evidence, is that a constant parameter formulation for the retail demand functions of table beef, hamburger beef and broilers is not appropriate and is likely to result in mislead- ing structural coefficients of retail demand. A notable and consistent exception throughout this analysis was aggregate pork demand for which the null hypothesis of fixed coefficients was accepted in each parameter variation model tested. The hypothesis of irreversibility of demand for beef is one testable hypothesis explaining systematic changes in the slopes of retail demand curves and in the derived flexibilities. This hypothesis was accepted: direct flexibilities were higher during periods of beef cattle cycle or inventory upturns and lower in cycle downturns. Over the past several years direct flexibilities for table beef have risen while for hamburger they have declined. The lower the direct flexibility, the greater the extent to which increases in quantities available per perSon are absorbed by demand and hence, the smaller the price adjustment required to clear the market. This means that, recently, quantities of hamburger beef have been more readily absorbed by demand than have table beef or poultry meats. This result for chicken is consistent with the observations that the market for chicken is becoming saturated and that the impact of gains from technology in the broiler industry, which have helped keep real prices down, may be slowing. The implication is that the preference for hamburger beef has been flowing vis—a-vis other meats. ' \ ,c .9 . 'l ' a ‘ '“hoc. o II- * —_ Laurence David Cornell The inclusion of the preferred retail meat demand models into a model of U.S. agriculture permitted simulations of the impacts of iden— tified structural changes on the livestock and feedgrain markets of the U.S. For exanple, results of simulations run individually, of a declin- ing consumer preference for table beef, an increasing cross-effect on table beef from a growing hamburger and broiler demand, and a declining income flexibility of demand for table beef, each indicated substantial impacts on table beef demand and on the levels of production and prices of feedlot produced beef. In another analysis, the impacts indicated by a simulation of a proposed dairy herd reduction program on cow beef pro- duction and hence imports of processing quality beef were also shown to be considerable. . U . 1 _ 3 , “- mil v.3. DEMAND roa III-:A'rs 0N U.S. I" ‘ ~ . I LIVESTOCK AND GRAIN MARKETS by Laurence David Cornell A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics <' ‘ ‘ Copyright: by ‘ Laurence David Cornell 1983 ,"K 'b‘ b, I f . L ' 1 U!“ prowl-ion ~' I ‘ 'ocictlm the arm: .- . during ' sit" Aw" \ . It»: I I . f£wa to 2-1.1.“ ‘. w" my mwi‘ 7 To my parents - Unit‘s-min ‘. “WW1 c " for Jun-I M5 .inc 7-7.: H isblstinor 'l: ‘L hul ht. ‘ their para...:-..- ”Ih] 5"“ Luring this were: u.“ 1 . I 4"! Immts. rim-um. “it ' v A firfltftudd 3:334‘ 1- . ..‘~ '- ' . t. . . is: Newsweek's -'N': - » “HA ~93 .;-.- - '-' ‘4 air“! in the ’{ast tr amt-2‘ tnrslcdfii; m D "‘ r. 77:5 Io O. '. I: 35.2: l 25),: "n. ' .‘ol "" Il".. "" —— ACKNONLE DGEMENTS I wish to express my appreciation to Dr. Vernon Sorenson, Chairman of my lhesis Committee, who provided professional counsel and provided constructive criticism of this study. Alas, his sound advice to keep the size of this thesis within reaSOnable limits appears to have gone unneeded. Drs. Harold Riley, John Ferris and Norman Obst each made valuable contributions to the development and final drafting of this research. Special thanks are due to the Chairman of my Guidance Commit- tee, Dr. Donald Mitchell, for his encouragement and for his important contributions during the conceptualization and initial development of this study. I owe a large debt to Dr. Roy Black, whose significant con- tributions to this research made him an honorary, albeit very effective, member of my committees. Substantial assistance was also provided by other members of the Department of Agricultural Economics. I wish to acknowledge and thank John Ross and Tom Christensen for their considerable and essential assistance during the modelling phase of this research. Thanks are due to Paul Wolberg, Jim Reisen and Chris Wolf of Computer Services for their perseverance and valuable assistance. Many other individuals contributed in one capacity or another. During this research I consulted extensively with both faculty and fel- low students. Among this latter catagory, special appreciation is due ‘30 Michael Morris, David Trechter and Ismael Ouedraogo. It is with a deep sense of gratitude that I acknowledge the contribution of Susan Lucas. Her enduring understanding and support, professionally and per- ”WHY. have made student life in the 'fast track' uncalculably more 111 ."l as :v’ .e o‘c- :w O Q n u. ell .ao a 0'0. . on a to: ’- o ‘ A .:. “"’ Q "h v e ”H... '0. ' e a ... v‘ '—'— tolerable. Also my appreciation goes to Ed McLaughlin for his ready and valuable counsel and for his therapeutic ten and twenty mile runs through out my stay at M.S.U. For financial assistance received during my study at M.S.U., a number of organizations are gratefully acknowledged. First, my appreci- ation is extended to the Department of Agricultural Economics at M.S.U. which provided me with financial assistance including necessary computer funds. Second, I gratefully acknowledge the support of the Australian Heat Research Council provided during the early stages of my study. Third, I acknowledge the generous financial support of the Australian Government and in particular, the Bureau of Agricultural Economics and associated individuals for their personal support and encouragement. Finally, I am most grateful for the typing assistance which I received from Peggy Crawford, who typed most of the first draft, and from Debbie Andrews, Nancy Creed, Debbie Greer, and others who helped me out in a 'pre-defense' crunch. I also thank Pat Neumann for preparing the final manuscript. Errors and omissions remain my responsibility. .r». qu.l- e 3..-.- .v. one-”Ide- I a no.0 ’- ne- 0. - u w- n a .|-‘ '.- .' 5 _.. DEEKMTIO . . . . . . . .. ACKNOWLEDGEMENTS . . . . . . LIST OF TABLES . . . . . . . . LIST OF FIGURES. . . . . . . TABLE OF CONTENTS CHAPTER 1. INTRODUCTION. . . . 1.1 Focus of Problem . . . . . 1.2 Research Objectives. . 1.3 Orientation of Research. 1 4 Organization of Proposed Research. . STRUCTURE, INSTITUTIONAL SETTING AND THE CHANGING ENVIRONMENT OF THE BEEF MARKET. . . 1 The World Beef Market: An Overview. . .2 Institutional Constraints in WOrld Beef Trade. 3 Characteristics of the U.S. Beef Subsector . NNN o a 2.3.1 Beef Cattle Industry. . . . . 2.3.2 Meat Marketing. . . 2 3 3 Consumption of Beef, Pork, Poultry and Other Meats . . . .1 Trends in Consumption. 2 Price Competition from Pork, Poultry and Other Meats. . . . . 2.3.3.3 Disposable Income and Consumer Expenditures on Meat . . . 2.3.3.4 Some Cross-Sectional Aspects of Meat Consumption . . . . . . . 2.4 Current Forces for Change in Consumer Demand for Meats. . . . . . . . . . . . . . . . . . 2.4.1. Impact of Age Structure. and Size and and Composition of Consumer HouseholdS. 2.4.2 Changing Lifestyles and Consumer Health and Nutrition Attitudes to Meats. . . . Page ii iii xiii H U‘I-l-‘LAJH 12 18 21 21 29 30 34 40 43 51 52 56 Page CHAPTER 2.4.3 Consumer Preferences and Beef Grades . . . . 60 2.4.4 Changes in Family Eating Patterns and Growth in the Fast Food Industry . . . . . . 65 3. THEORETICAL FOUNDATIONS OF CONSUMER DEMAND AND APPROACHES TO THE ESTIMATION OF STRUCTURAL CHANGES . . . 73 3.1 Traditional Theory. . . . . . . . . . . . . . . . . 73 3.1.1 Consumer Preferences and Goods Characteristics. . . . . . . . . 78 3.1.2 Dynamic Models and Habit Formation . . . . . 84 3.1.3 Irreversibility of Demand. . . . . . . . 88 3.2 Structural Change in Demand Interrelationships. . . 95 3.2.1 Definition of Structural Change. . . . . . . 95 3.2 Nature of the Problem. . . . . . 96 3.3 Traditional Methods for Identifying and Estimating Structural Change . . . . . . . . . . . . . 98 3.3.1 Graphical Analysis . . . . . 98 3.3.2 Time Trend Variables (Gradual Structural Shifts). . . . 101 3.3 3 Dummy Variables (Abrupt Structural Shifts) . 105 3.3.4 Homogeneous Subperiods . . . . . . . . . . 108 3.3.5 Interaction Variables. . . . . . . . . . . . 110 3.4 More Advanced Methods for Identifying and Estimating Structural Change. . . . . . . . . . . 112 3.4.1 Grafted Polynomials and Spline Functions . . 112 3.4.2 Models of Quality Change . . . . . . . . . 121 3.4.3 Time-Varying Parameter Models. . . . . . . . 123 3.4.4 Cross-Sectional Time Series Model. . . . . . 127 3.4.5 Time-Varying Switching Regression Model. . _ 130 3.4.6 Continuous Time Varying Parameter Models . . 132 3.4.7 COMPLEX: A Non-Linear Optimization Search Technique. . . . . . . . . . . . . . . . . 141 3.4.8 Legendre Polynomials . . . . . . . . . . . . 148 3.5 A Time-Varying Model Using COMPLEX and Legendre Polynomials . . . . . . . . . . . . 152 vi CHAPTER 4. ESTIMATION OF RETAIL MEAT DEMAND AND IDENTIFICATION AND MEASUREMENT OF PARAMETER CHANGES. . . . . . . . . . 158 4.1 Specification of the Economic Model. . . . . . . . 160 4.1.1 Restrictions on a Price-Dependent Demand Function. . . . . . . . . . . . . . . . . . 170 4.1.2 Some Observations on Interpreting Consumer Responsiveness. . . . . . . . . . . . . . . 173 4.2 Statistical Model and Estimation Procedure . . . . 181 4.2.1 Functional Form of Equations. . . . . . . . 184 4.2.2 Method of Estimation. . . . . . . . . . . . 135 4.2.3 Estimation Period and Data Limitations. . . 187 4.3 Structural Model and Candidate Variables . . . . . 189 4.4 Results of the Base Model of Demand. . . . . . . . 197 4.4.1 Estimation of a Base Model. . . . . . . . . 197 4.4.2 Beef Price Adjustment Model . . . . . . . . 207 4.4.3 Effect of Age Composition and Expenditure Away From Home on U.S. Meat Consumption . . 212 4.5 Examination of Structural Shifts in Demand . . . . 218 4.5.1 Graphical Analysis. . . . . . . . . . . . . 218 4.5.2 Slope and Intercept Changes in Demand . . . 233 4.5.3 Measurement of Irreversibility of Demand. . 237 4.6 Time-Varying Switching Regression Model. . . . . . 246 4.6.1 Linear and Cubic Splines. . . . . . . . . . 247 4.7 Continuous Time-Vary Parameter Model . 260 4.7.1 Legendre Polynomials. . . . 261 4.8 Concluding Comments. 286 5° EMPIRICAL EVALUATION OF THE IMPACT OF STRUCTURAL CHANGES ON RETAIL MEAT DEMAND AND U.S. BEEF IMPORTS . . 292 5.1 Overview of the Michigan State University 294 Agricultural Model . . . . . . . . 5.1.1 Model Estimation Procedure. . . . . . . . . 299 5.1.2 Solution Algorithm . . . . . . . . . . . . 300 vii p \ a CHAPTER 5.2 U.S. Domestic Livestock Model . . 5 5. 5 5 5.7 Sensitivity Analyses. . . . Beef Supply. . . . . . . . . . . . . . . Dairy Supply and Demand. . . . . . . . . Pork Supply. . . . . . . . . . Poultry Supply . . . . . . . . . Livestock Feed Consumption . . UUIUIU'IUI C NNNNN o UI-waH 3 Model of U.S. Retail Meat Demand. . . 4 Projections of Key Exogenous Variables. 5 . Validation of Model . . .6 Base Projections for Endogenous Variables: Actual and Simulated Results. 5.7.1 Impact of a Short Livestock Feed Supply. 5.7.2 Effect of Changes in U.S. Beef Imports . 5.8 Economic and Policy Simulation Analyses . Sustained Increase in Real Disposable Income Per Person. . Structural Changes in Demand Parameters. 5.8.3 5.8.4 Impact of a Proposed Dairy Herd Reduction Program. . . 5.8.5 Effect of Changing Income Growth in World Markets. . . . 5.9 Summary of Main Points. 6. CONCLUSIONS. . . . . . . . APPENDICES A. Appendix Tables and Figures. . . . . . . 3. Selected MSUAM Subroutines . . C Listing of Data Used in Estimation of Retail Demand Models . . . . . . . . . . . . . . . . . . . . . . . LITERATURE CITED. . . . . . . . . . . viii Sustained Increase in Retail Meat Demand . Page 301 302 309 309 311 312 314 328 330 337 344 344 352 358 359 361 362 372 378 381 386 398 404 411 413 ..o 9 it. no- 2.1 2.2 2.3 2.4 2.5 4.2 4.3 4.4 4.5 4.7 4.8 4.9 LIST OF TABLES Page Beef Production, Consumption and Trade: Averages, 1971-73, 1976-78 and 1977-81 . . . . . . . . . . . . . 8 U.S. Distribution of Processing Table Beef in 1978 . . 25 Consumption Per Person and Annual Percentage Changes in Consumption Per Person of Red Meats, Poultry and Fish: 1960-82 . . . . . . . . . . . . . . . . . . . . 31 Shares of Beef, Pork and Poultry Meat, and Red Meat Consumption in Total Meat Consumption Per Person: 1960—1982 . . . . . . . . . . . . . . . . . . . . 35 At-Home Consumption and Expenditure in the U.S.: Spring 1965 and Spring 1977. . . . . . . . . . . . . . 50 Decomposition of Components of Change in Direct- Quantity Flexibility of Demand . . . . . . . . . . . . 176 Description of Candidate Variable Codes. . . . . . . . 190 Estimated Retail Demand Equations for Meat: Base Model Estimated Over the Period 1950-1982. . . . . . . 198 Estimates of Base Model Quantity and Income Flexibili- ties of Retail Meat Demand . . . . . . . . . . . . . . 202 Estimated Price Adjustment Equation of Retail Meat Demand: Estimated Over the Period 1950-1982 . . . . . 208 Short and Long—Run Quantity and Income Flexibilities of Retail Demand Price Adjustment Model: Estimated Over the Period 1950-1982 . . . . . . . . . . . . . . 209 Estimates of the Influence of Age Structure of Retail Beef Prices: Estimates Over the Period: 1950—1982. . 215 Effect of Time on Slope Coefficient of Retail Meat Demand: Time-Quantity Interaction Variables . . . . . 225 Expected Direction of Change in Direct Flexibilities Obtained From Decomposition of Actual Changes in Components of Direct Flexibilities . . . . . . . . . . 227 4.13 4.14 4.15 4.17 4.18 4.19 4.20 Page Average Retail Quantity and Income Flexibilities Over Two Periods: 1950-1966 and 1967-1982 . . . . . 229 Direct Flexibilities for Retail Meat Demand: Estimation Used T*Q Interaction Variable . . . . . . 231 Slope and Interaction Changes in Retail Demand for Meat . . . . . . . . . 236 Irreversibility in Retail Demand for Beef Based on Changes in Aggregate Beef Production . . Irreversibility in Retail Demand for Beef Based on . . . . 243 Changes in U.S. Cattle Cycles. . . . . . . . . 245 Procedure for Testing Contribution of Composite Spline Variables Relative to Base Models of Retail Meat Demand . . . . . . . . . . . . . . 250 Table Beef Price Equation: Spline Functions and Contribution of Linear and Cubic Splines . . . . . . 251 Hamburger Beef Price Equation: Spline Functions and Contribution of Linear and Cubic Splines . . . . 252 Pork Price Equation: Spline Functions and Contri- bution of Linear and Cubic Splines . . . . . . . . . 253 Broiler Price Equation: Spline Functions and Contribution of Linear and Cubic Splines . . . . . . 254 Direct Flexibilities of Table and Hamburger Beef Estimated from Linear and Cubic Splines . . . . . 258 Direct Flexibilities of Pork and Broilers Estimated from Linear and Cubic Splines. . . . . . . . . . . . 259 Calculated Values of Legendre Polynomials Up to the 5th Degree . . . . . . . . Time-Varying Parameter Model All Beef. . . . . . . . . Time-Varying Parameter Model Table Beef . . . . . . . Time-Varying Parameter Model Hamburger Beef . . . . Time-Varying Parameter Model Pork . . . . . . . . . . . . 265 of Retail Demand for . . . . . . . . . 267 of Retail Demand for . . . . . . . . . . . . 268 of Retail Demand for . . . . . . . . . . . . 269 of Retail Demand for . . . . ...... . . 270 ". 'Iu 'I Page 4.27 Time-Varying Parameter Mbdel of Retail Demand for Broilers. . . . . . . . . . . . . . . . . . . 271 4.28 Direct and Income Flexibilities for All Beef Estimated From Legendre Polynomial Model. . . . . . . 274 4.29 Direct and Income Flexibilities for Table Beef Estimated From Legendre Polynomial Model. . . . . . . 275 4.30 Direct and Income Flexibilities for Hamburger Beef Estimated From Legendre Polynomial Model. . . . . . 276 4.31 Direct and Income Flexibilities for Pork Estimated From Legendre Polynomial Model. . . . . . . . . . . 277 4.32 Direct and Income Flexibilities for Broilers Estimated From Legendre Polynomial Model. . . . . . . 278 4.33 Direct, Cross and Income Flexibilities for Broilers Estimated From Legendre Polynomial Model. . . . . . . 279 5.1 Operation of U. S. Meat Import Law on Imports of Beef: 1965—1983 a I o o o o o a o a a o o o c o a o 308 5.2 Variable Codes and Units of Measurement Used in Specifying the U.S. Retail Meat Demand Model. . . . . 317 5.3 Retail Meat Price Equations in the U.S. Meat Demand Model........................ 324 5.4 Estimated Farm-Retail Price Margins Equations: 1960- 1982. . . . . . . . . . . . . . . . . . . . . . . . . 325 5-5 Actual and Projected Values of U.S. Population, Inflation, Interest Rates and Disposable Income: 1975-1991 . . . . . . . . . . . . . . . . . . . . . . 329 5.6 International Income Growth Assumptions: 1975-1991 . 331 5.7 Population Estimates for 1982 with Projected Growth Rates for 1983-1991 . . . . . . . . . . . . . . . . . 331 5.8 Wheat, Coarse Grains, and Soybean Yield Assumptions: 1983—1991 . . . . . . . . . . . . . . . . . . . . 332 5.9 Actual and Projected Cropland Harvested Area of Wheat and Coarse Grains: 1975-1991 . . . . . . . . . 333 5.10 Model Validation: Theil's U and Root Mean Squared Error as a Percentage of the Mean: 1975-1982 . . . . 335 xi ,nu . .- . 4... . .. | « Mo- . I . ' no * . “H ' . ‘ c "u 5.11 5.12 5.13 5.14 5.15 5.18 5.19 5.20 Page Effect of a 20 percent Decrease in Feed Grain Yields: (Strong Feed Market): One Period Impact in 1983 O I O U I I O I I I l O Q 0 D U C I O O C O 9 24 7 Effect of a 700 Million Pound Increase in U. S. Beef Imports: One Period Impact in 1983 . . . . . . 354 Effect of a One Period and a Permanent Increase in U.S. Beef Imports . . . . . . . . . . . . . . . . . . 356 Effect of a Sustained Increase in the Retail Demand 360 for Meats . . . . . . . . . . . . . . . . . . . . . . Effect of a Continuous Increase in Own-Quantity Slope Coefficients in Retail Meat Demand Curves . . . . . . 364 Effect of a Continuous Decrease in Income Slope Coefficients in Retail Meat Demand Curves . . . . . . 368 Effect of a Continuous Increase in the Substitution Effect of Hamburger Beef on Other Red Meat Demand and of Broilers on All Other Meat Demand. . . . . . . . . 370 Production and Allocation of Dairy Cow Beef Under Alternative Dairy Cow Herd Reduction Policies . . . . 376 Potential Impact of Proposed U. S. Dairy Cow Herd Reduction Program . . . . . . . . . 377 Effect of Low Versus High Income Growth in Develop— ing Markets, Soviet Bloc and Mainland China . . . . . 380 .- '0 o ‘-- L. . \" .l on. . . N, " uh - . . u C u. . .I -. I I O 4'. ‘ ' l .. .1. ‘ D 2.2 2.3 2.4 3.6 3.7 3.8 4.1 LIST OF FIGURES Pattern of Whrld Trade in Beef. . . . . . . . . . . . The Political—Economic Structure of the U. S. Beef Subsector . . . . . . . . . . . . . . . . . . . . . Structure of Major Flows of Live Cattle and Beef Among Central Participants in the U.S. Beef and Beef Cattle System. . . . . . . . . . . . . . . . . . . Relationships Between Beef Consumption Per Person and Beef, Pork and Chicken Retail Prices. . . . . . . Retail Price Relativities of Table Beef, Hamburger Beef, Pork and Chicken. . . . . . . . o o o Income and Expenditures for Food at Home and Away From Home . . . . . . . . . . . . . . . . . . . . . Irreversibility in Price-Quantity Relationships . . . Shifts in the Structure of Demand . . . . . . . . . . An Illustration of Joined Segments of Different Polynomials . . . . . . . . . . . . . . . . . . . A 'Complex' of Three Points . . . . . . . . . . . . . Movement From Inferior Points in 'Complex' Optimization Procedure . . . . . . . . . . . . . . . . . . . . Representation of Continuous and Discrete Controllable Variables in an Optimal Control Problem . . . . . . . Block Diagram of Time-Varying Parameter Model . . . . Illustration of Continuous Time-Varying Elasticity Estimations . . . . . . . . . . . . . . . . . . . . Plot of Retail Prices and Domestic Consumption of Table Beef. . . . . . . . . . . . . . . . . . . . . . xiii Page 11 20 22 36 38 67 91 99 113 144 145 148 153 157 219 ———fi 5.3 5.4 5.5 5.6 5.7 5.8 Page Plot of Retail Prices and Domestic Consumption of Hamburger Beef . . . . . . . . . . . . . . . 220 Plot of Retail Prices and Domestic Consumption of Pork . . . . . . . . . . . . . . . . . . . 221 Plot of Retail Prices and Domestic Consumption of Broilers . . . . . . . . . . . . . . . 222 Flowchart of Major Components of MSU Agricultural Model . . . . . . . . . . . . . . 295 Flowchart of U. S. Beef Supply and Dairy Supply and Demand. . . . . . . 303 Flowchart of U. S. Pork and Broiler Supply and Turkey and Egg Supply and Demand. 310 Flowchart of Wheat and Feedgrains Consumption by U. S. Livestock Sector . . . . 313 Block Diagram of U.S. Retail Demand for Meats . 315 Base Level Projections: Actual and Simulated Results . . . . . . . . . . . . . . . . . . 333 Effect of a 20 percent Decrease in Feedgrain Yields: One Period Impact in 1983. 349 Simulated Increase in Table Beef Flexibility of Demand. . . . . . 355 .no= . .-¢- , P; I ' I \ .... f. v u..._' . .0. ~ I‘ ' A. “'9‘.- u ‘ - -_ z". —— CHAPTER 1 INTRODUCTION 1.1 Focus of the Problem (he of the important tasks of demand and price analysis has been to identify and to quantify the structure of consumer markets by estimating demand functions for various goods. The coefficients or parameters of the estimated demand functions are known as the structural coefficients or parameters. The quest for accurate and reliable estimates of these parameters arises from their use for public policy analysis and for forecasting. The journals of economics and econometrics are filled with attempts to overcome the many formidable obstacles arising during their estimation . Typically, economists have operated with some version of the clas— sical linear statistical model, a model which assumes that tastes and preferences, habits, expectations and various other sociological and psychological phenomena remain constant over time. In particular, the assumption is usually made in time series analysis that the structure of demand and the values of the true coefficients remain unchanged over the period under investigation. Despite this simplifying assumption, casual empirical observation will quickly show that the structure may change over time, gradually or abruptly. The difficulties which market analysts have experienced in understanding the recent developments in the retail markets for meats illustrate clearly the nature of the prob- lem. o .obl“ .vu :{ u. l- n b —— 2 Economists have long been plagued by the lack of a fundamental explanation for these shifts in price-quantity relationships, typically observed over time. To some extent this absence of an adequate explana- tion has been due to a preoccupation with the traditional theory. It was concluded by Tomek and Robinson (1977, p. 388) in their extensive review that in the literature of agricultural price analysis, as in much of the economic literature, own-price elasticities have received more emphasis than is justified by their economic importance. The large changes in consumption, production and price have occurred as a result of shifts in demand and supply functions rather than as a result of movements along a static ceteris paribus schedule. Preoccupation with price elastici- ties has in some cases led economists to ignore more critical variables. Alongside the preoccupation with price elasticities, the assumption of fixed coefficients stands equally as an obstacle to economic estima- tion and forecasting of key policy and outlook variables. With the increased economic and political interelationships among the various sectors of the national and international economies, the magnitude of structural parameters are less likely to remain constant through time. This has been true of agriculture (Fox, 1962; Schuh, 1976; and Gardner, 1981M Tomek and Robinson (1977. p. 388) cite the large underestimation of beef prices during the early 19708 using models based on pre-1969 data, Suggesting that a change in the structure of the demand for beef in the U.S.may have been responsible. More recently, the observed decrease in demand for red meats has been attributed to changing tastes, changing attitudes toward health, the rapid growth in new processed poultry p' w“ ": . "no." ‘ a? u ll! ll .I. so u. .... 0‘. -'w " 3 products and the associated increasing shift of consumer preferences towards chicken and away from red meats. These and other changes all point to an evolving and at times an abruptly changing structure in the demand for meats in the U.S. If account is not taken of these changes, future economic forecasts may well be inaccurate. Furthermore, if economic models are estimated with a constant parameter formulation and the paraneters change over time, especially for those markets where sig- nificant market adjustments have taken place, then economic policies based on these estimates are at best questionable and may be misleading. It is towards this matter that the present research is directed. 1.2 Research Objectives The broad objectives of this study are (a) to test the assumption, underlying the classical model of consumer demand, of constancy of structural coefficients in the retail demand for major meats in the U.S.; (b) to investigate parameter variation in terms of systematiC' behavior over time; (c) to relate systematic variation of estimated Slope coefficients and estimated measures of responsiveness of demand to observed structural changes in retail meat demand; and (d) to incor- Porate estimated models of retail demand into a model of U.S. agricul- ture and to simulate the long-term impacts of recent structural changes on U.S. agriculture and international feedgrain markets, and particu- larly on U.S. imports of beef. fi v‘. . .- 0,‘ 0 'DJ u. v“. I :I l‘.‘ —_ u 1.3 Orientation of Research In the U.S.. the beef subsector has had and continues to have a dominating effect on the retail market for meats. To be sure, poultry meat and certain non-meat foods have in recent years claimed an increas— ing share in consumers' diets. Beef still, however, commands the major portion of the consumers' budget and of the consumers' purchased quanti— ties of retail meats. It is also the major source of protein. The presence of the U.S. beef market is also felt internationally. Besides being the largest producer of beef, the U.S. is the world's largest importer of beef, mainly processing—quality beef and an impor- tant exporter of table cuts of beef. The U.S. market is of prime impor- tance to Australia, the world's major exporter of beef and largest sup- plier of beef to the U.S. U.S. beef imports, however, represent less than 10 percent of total U.S. beef consumption. In a world in which total beef trade is a small proportion of total world production and in which trade patterns and price formation in world markets are influenced considerably by institutional factors, changes in domestic markets of importers can have a pronounced effect on the returns to exporting nations. This is certainly the case in the U.S., where developments in the markets for meats can significantly influence the level of beef imPorts and consequently affect the returns to exporters who supply the U.S. market, chiefly Australia and New Zealand. Compounding the uncer- tainty of gaining entry into the U.S. beef market is the regulation of U.S. beef imports by a complex 'countercyclical' meat import law. Access for those exporters is further complicated by the fact that this law does not impose a binding constraint in all years. All h...‘ :241‘. x' . I: 'D-o- ‘ Q “A \- e . ‘n .I. . I . . -~‘ ———- 5 One source of uncertainty and a major determinant of the level of beef imports permitted entry in the U.S. is the U.S. retail demand for meats. Changes in the structure of demand for table beef, hamburger beef, pork and chicken are investigated in this research. Analyses of the impact of these and other structural changes on imports of beef into the U.S. are reported. 1.” Organization of Proposed Research In the following chapter, the U.S. beef market is discussed in the context of world trade patterns. The structure of the U.S. market and key changes in consumption and marketing are also discussed. In Chapter 3 the theoretical underpinnings of traditional consumer demand are crit- ically reviewed with respect to the measurement of structural changes in demand parameters. Given this theory and its noted limitations, an extensive array of methods for identifying and estimating structural changes in demand are reviewed. Chapter ll reports the results of empir- ical analysis of retail meat demand models. Emphasis is given to relat- ing observed changes in retail meat markets to estimates of time-varying structural coefficients of demand for table beef, hamburger beef, pork and chicken. In Chapter 5 preferred models of retail demand for each of these four major meat categories are incorporated into a larger econometric model of U.S. agriculture and world grain trade. Simula- tions of this model over the longer term provide scenarios of impacts of structural changes on U.S. retail meat demand and in particular, on the level of beef imports permitted entry under the countercyclical meat import law. The final Chapter 6 contains concluding comments and the 90116}! implications from the previous analyses. I... . ,, I. n ‘ I a '~..‘ I‘ u e. H |.' ‘ e . .“ ‘l .l f”~. ‘J n . n A v'.,l .- . '1 n ‘e l.- . , n u A 1 In - . CHAPTER 2 STRUCTURE, INSTITUTIONAL SETTING AND THE CHANGING ENVIRONMENT OF THE BEEF MARKET 2.1 The World Beef Market: An Overview The world beef market can be viewed in terms of the following broad dimensions: cattle population, aggregate production, consumption per person, and imports and exports. World beef cattle numbered over 1.2 billion in 1982.1 Two—thirds were located in developing countries with the greatest concentrations being in Asia (30 percent) and South America (18 percent). The majority of beef cattle in developed countries are found in Europe (11 percent) and the U.S. (9 percent). Centrally planned economies which include parts of the developed and developing worlds accounted for 17 percent of all beef cattle in 1981; USSR con- tained 8 percent of the total. Over the past two decades (1960-1981), growth in cattle numbers has been greatest in the USSR, Oceania and Africa; in these countries growth has been 50 percent or more. Slower growth has been experienced in the U.S. and Europe, at less than 20 per- cent. Developed countries carried 35 percent of the cattle but accounted for 67 percent of total beef production. This disparity may be eXplained by the subsistence livestock practices and low productivity in 1 Data are cited from various issues of FAO, Production Yearbook, and from Simpson and Farris (1982). A proportion of the animals reported as beef cattle are dual purpose breeds. e ll '3 :- -’ "' - l \ F a "all r o I” .l « Ole. . .. '9 I ‘ I U I .‘* q .- O "O fi 7 developing countries. The U.S. produced 22 percent of the world's beef from only 12 percent of the world's cattle. Europe had 11 percent of the world's cattle and 22 percent of production. Growth in production has been relatively slow over the 1960-1981 period: 43 percent in the U.S., 1111 percent in the USSR and 53 percent in North and Central Amer- ica. By contrast production growth, though from a smaller base, has been much faster in Asia and Oceania. About 211 percent of world beef was consumed in the U.S., 22 percent in the whole of Europe, 15 percent in the USSR and 12 percent in South America between 1961 and 1981. Growth in consumption has been fastest in the USSR . The international market for beef is characterized by a few major exporting and importing countries. The beef trade has represented between 8 percent and 10 percent of world production (OECD, 1980). Currently, four major regions dominate beef imports, namely, the U.S., the EEC. and to a lesser extent Canada and Japan. These key developed country markets jointly account for over two-thirds of total beef imports. The U.S. and the EEC are also leading producing regions in which imports comprise only a small proportion of total beef consump— tion. Oceania (mainly Australia and New Zealand) and South America (mainly Argentina, Brazil and Uruguay) provide the bulk of the world's beef exports, although these countries account for a relatively small Proportion of total world production. Details are presented in Table 2'1. There exists a large number of other countries which trade in beef. Collectively, they represent a significant proportion of total trade | ~... o .:g Table 2.1 Beef Production, Consumption and Trade: Averages 1971-73, 1976-78 and 1977-81 (Carcass Weight Equivalent) Production Consumption Imports (Exports);i Z of Pounds 74 of As a Z of '000 World '000 Per ‘000 World Consumption/ Tonnes Total Tonnes Person Tonnes Total Production A. Major Importing Countries 066 1971-73 10.115 29.5 10,741 117 373 39.4 8.1 1976-78 11.735 27.7 12,619 130 953 45.7 7.6 1979-81 10.092 24.3 10,979 106 949 40.6 8.6 mm (9) 1971-73 5,337 15.7 6,399 55 672‘“c 30.4 10.5 1976-78 6,428 15.2 6,529 62 205b 10.0 3.1 1979-81 6.864 16.9 6,667 57 324b 13.9 4.9 Japan 1971-73 279 0.8 393 9 111 5.0 23.2 1976-78 343 0.3 467 9 123 6.1 27.4 1979-81 430 1.1 594 25 175 7.5 29.5 Canada 1971-73 390 2.6 933 95 94 4.2 10.1 1976-78 1,148 2.7 1.209 115 115 5.5 9.5 1979-81 979 2.4 997 93 32 3.5 3.2 Total 1971-73 16,671 48.6 13,466 -- 1.750a 79.1 -- 1976-78 19.659 46.4 20.824 -- 1.4063 67.2 -- 1979-31 18,365 45.1 19,237 -- 1.5303 65.4 - B. Major ExportinLCountries Anittalia 1971-73 1,306 3.3 566 97 716 26.2 54.8 1976-78 1.954 4.6 955 139 932 34.5 50.3 1979-31 1,573 3.9 730 110 373 27.5 55.5 Argentina 1971-73 2,117 6.2 1.540 101 562 20.6 26.5 1976-73 3,892 6.8 2,302 121 591 20.7 15.2 1979-31 2,989 7.3 2,436 132 551 17.4 13.4 New 2611666 1971-73 407 1.2 134 143 274 10.0 67.3 1976-78 544 1.3 173 194 330 13.3 70.0 1979-31 502 1.2 137 198 341 10.7 67.9 Brazil 1971-73 2.100 6.1 1,383 42 135 6.3 8.3 1976-73 2.333 5.5 2,137 42 171 6.0 7.3 1979-81 2,167 5.3 2.017 35 183 5.9 3.7 Total 1971-73 5.930 17.3 4,123 -. 1,737: 63.6 1976-73 7,723 13.2 5,622 -- 2,124b 71.7 - 1979-81 7.231 17.8 5,370 -- 1.953 61.5 Source: USDA. Foreign Agriculture Circular: Livestock and Meat (various issues). aFor importers these data relate to gross imports; for exporters data relate to gross export b Excludes intro-EEC trade. cExcludes live cattle trade. U ., q ‘10." .l' nul '71.: ' u “ .r-U ,.. I ‘9 8"!“ L'- ,.. .a... up up A... '. hhinoio‘. :Il ‘AOI . ' occa. W21; A I '"~~\ Ml‘hdividually. however. they exert very little influence on the seam: and account for a very small proportion of world beef production ' and oohmption . The present structure of world trade in beef has arisen partly because of comparative advantages in beef exporting and partly due to institutional factors in response to, for example, the existence of foot and mouth disease in certain regions. Exporting regions such as South America and Oceania have a comparative advantage in beef production because of natural conditions, such as their large pasture bases which favors cattle raising and their low human population pressures arising from a comparative endowment of land. Areas such as the U.K. and Japan have a comparative disadvantage in beef production and, hence, are importers of beef. The U.S. and Canada have a relatively small area of permanent pasture and are major importors of grass-fed beef. However, these countries are also major exporters of livestock products, sup- Ported by their extensive feedgrain systems. Also impacting on the overall pattern of trade and effectively seg- menting the market into an Atlantic market and a Pacific market has been the regulations relating to foot and mouth disease. The beef trade con- sists predominantly of fresh, chilled and frozen beef, with frozen boxed manufacturing quality beef2 being the main type. The flow of trade is Predominantly between Oceania and North America on the one hand and between South America and the EEC on the other. This pattern emerged 2Manufacturing beef, processing beef, ground beef and hamburger beef are used interchangeable in the literature. This practice is adopted in this research. At various points in this research some differences anong these terms will be highlighted. Il~ll 0".I a "C I :'--v'< I "'I on . I I e . 1.7. a a .I .6 . ‘ “‘ID- e. \ ." u I a“ \ i 1 '1 1 e . . ~ fi 10 following prohibition by Japan and the U.S. of meat imports from foot and mouth infected regions such as South America. Because these regions cannot ship their fresh. frozen or chilled beef to the U.S. they now depend almost exclusively on Europe and the USSR.3 Hence, Australia and New Zealand, countries free of foot and mouth disease, are the only major suppliers to the North American and Japanese markets. Australia accounts for around 50 percent and New Zealand for some 20 percent of U.S. beef imports. An additional factor which led to the current structure of the market was the entry of the U.K. into the EEC-9 in 1973. Up to the 1960's the U.K. dominated international trade in beef and was the major market for Oceania beef. mring this time North American and Japanese imports were negligible. The U.K. had no prohibitive restrictions on trade in meat except for health and sanitation and Australian beef trade flourished under the Fifteen Year Meat Agreement with the U.K. By the early 1960's this agreement expired, fortuitously for Australia, just when the U.S. emerged as a significant importer in the world beef market. The U.S. was experiencing rapidly rising income per person, a Brewing ‘fast food' industry and had few trade restrictions. These developments consolidated the trade pattern illustrated in Figure 2.1, i.e., the establishment of separate Pacific and Atlantic trading markets Which were only weakly linked by Oceania's European trade, now greatly diminished by U.K. entrenchment within the EEC. 3Argentina is permitted to export cooked canned beef to the U.S., although these exports represent less than 10 percent of total beef imports. 11 Pacific Market Atlantic Market North Japan America / mporters t 7 * Expor ers I, P“ [\4\‘ / \ ‘ l \ / . \ - / Argent1na Austra11a (Brazil \\ I New Egaland / Uraguay) l / \ / ._ r_— —— —— -—- r" "" I Figure 2.1 Pattern of Wor1d Trade in Beef mu.- ..uvv-uu we ’ari‘l. 1.! as 1.1 H v'ovv . ....1 b‘.‘ .‘ I I ‘ "be... -«..;P . a . '. 6" ' l t... '1‘! 'o.‘ 1.-" O 40- :. —— 12 2.2 Institutional Constraints in World Beef Trade Various policies restricting trade have been adopted by all major beef importing countries, the objective being to maintain prices and/or incomes to domestic producers. Major beef exporters have responded with other institutional measures which attempt to counter importing country restrictions and hence achieve maximum access and returns for exporting country producers. As a consequence, world beef trade is dependent not only upon economic considerations in these countries, but also upon their political environments. Institutional constraints operating in 1' world beef markets will now be discussed. Institutions governing U.S. imports and to a lesser degree Australian exports will be outlined in greater detail . In 1964 the U.S. Congress implemented a Meat Import Act for the purpose of regulating. through quotas, imports of manufacturing beef into the U.S. Beef was permitted unrestricted entry up to a quantity 'trigger' level, which was set 10 percent above the quota level. In any given year, if imports were expected to exceed this trigger level of 110 percent of the base meat quota then the quota was invoked. The President had the power to relax or suspend the quota in the national interest, e.g., when internal beef prices have been high as in 1973. However, because of certain destabilizing effects of this law, the 1979 Meat Import Law was enacted. The latter law has a countercyclical com- ponent which, in effect, permits increased imports when domestic beef suPplies are low without involving quotas but limits imports when sup- plies are high. Further details of this quota will be discussed in Chapter 5. M L“; an s ' a a. 0‘... a no. . ‘ a. .H I l 1 ' - a. m o 11‘. ' I v" .A ' l 33' a I." U - .3 . In. .Ie . . '. '. 6". ‘- ‘. 1 .I. ‘e fi 13 Access to the Canadian beef market has been subject to a quota res- traint since 1976. Previously, restrictions on imports were less for- mally imposed, often being negotiated with major supplying countries. The level of the restraint has been determined by the Minister for Agri— culture in response to general market conditions. However, moves to replace these rather ad hoc quota arrangements gained momentum with the passing of the countercyclical Meat Import Law in the USA in December 1979. In November 1980, a Meat Import Act (Bill C416) was introduced by the Minister for Agriculture. The government felt that without such import controls Canada would be vulnerable to market distortions caused '1‘ by the U.S. beef import controls. This Act, also countercyclical in effect, became law at the end of 1981 and was applied to beef imports in 1982”. In practice, Canadian beef trade policy is very similar to U.S. beef trade policy and its meat import laws. This is because the Cana- dian cattle industry is similar in many reSpects to the U.S. cattle industry. Like the U.S., a fed-beef production system predominates which is also subject to a pronounced and regular cattle cycle that closely follows the U.S. cycle. Internal stabilization objectives of domestic livestock industries largely determine Japanese beef trade policy. These objectives are to maintain prices to producers at stable and high levels by world stan- dards. The Livestock Industry Promotion Corporation (LIPC) is the con- trolling agency, responsible for establishing global quotas on beef A Week; and Turner (1981) provide a detailed review of the poten- tial effects of this law. .'. any-1 ”.1? uo—e "" . ‘ up ' . .0... . .\ at ei'avk “I a.. I D- ' 0.. '- I. a I —w 111 imports, stockpiling beef, and various other activitiess. The LIPC administers the quota system by issuing licenses to Japanese importing firms which then negotiate with overseas suppliers. Levies are charged on all imports and effectively raise world prices up to internal whole- sale prices. Proceeds from these levies are used to promote Japan's livestock industries. A commitment to greater self-sufficiency in livestock products underlies Japan's policies. As a result, consumer beef prices are main- tained at very high levels relative to other countries, and imports are virtually unrelated to world trade prices. Japanese beef imports are influenced by the domestic market, in particular by the level of domes- tic beef production. For instance, since 1979. the Japanese government has been subsidizing Japanese dairy farmers to slaughter low-yielding milk cattle. The aim is not to increase milk production, but to increase dairy industry productivity. These subsidies have led to higher dairy cattle slaughterings and a consequent increase in domestic meat production. In most years, Australia has captured 80-90 percent of this market. However, future global beef quotas continue to depend upon both economic and political considerations (Bureau of Agricultural Economics, 1981). For example, the currently depressed dairy-beef market induced the LIPC to reduce the chilled beef quota, since these chilled beef imports compete directly with Japanese dairy-steer beef. Furthermore, Japan's current trade surplus with the U.S. has resulted in pressure from the U.S. on Japan to relax restrictions on imports, 5 Details of Japanese beef import policies are contained in Bureau Of Agricultural Economics (1975) and Longworth (1978). 1.. a' F owafir 1m"? ‘ I‘le"" ' we!» 0 l a a. u n , 1 :19" “-.. I 0...... .- L. a 'II 9. .1. S". :"eoh "v. 4:, I I fi 15 including U.S. table quality beef. For this reason Japan can be expected to give preferential treatment to U.S. beef imports. Intervention in the beef market of the EEC comes under the Common Agricultural Policy (CAP) and its extensive and seemingly plethoric array of rules and regulations. Since 1980, the EEC has been a net exporter of beef and veal. A system of variable levies and duties pro- vide the mechanism for controlling imports and supporting internal prices. This mechanism comes into effect when internal prices fall below 'guide' prices. The market is supported also by intervention buy- ing, stockpiling of beef, and 'restitution' payments (subsidies) to EEC exporters in times of surplus production or stocks. On occasions when their measures have been ineffective in maintaining internal prices at guide price levels, virtually all imports have been prohibited as occurred in the mid-1970's. Daring 1981, imports by the EEC operated under three quota schemes: first, the GATT levy-free quota including boneless frozen beef and veal and high—quality cuts plus an Australian quota of buffalo meat; second, the ACP imports from African, Caribbean, and Pacific nations; and third, the balance sheet arrangements quota (mreau of Agricultural Economics, 1981, pp. 25—26). This turnabout of the EEC in 1980 from net imperter to a net exporter of beef and veal reflects to a very large degree the extensive POlicy mechanisms of internal support and associated barriers to cheaper world imports which have protected beef producers in the EEC. EEC beef 'surpluses' which are finding their way increasingly into world trade and third country markets are directly related to the high internal Price supports given to the perennially surplusing production of the EEC eh ‘guu ~10" "; i no pl "aue: .- to» I: II . 0 a .I I. g);- u n... i 0‘ t: 'E' "N- ehl . n.. .. - I ' " ‘aeae . t e O 11 u. ' 'nu e :0, e. ‘o a. :‘W 3' I.”‘ 0-. fi 16 dairy industry. EEC dairy cattle are predominantly dual purpose. As dairy cow nunbers grow. so does beef production. With a mere 8-10 percent of total world beef production entering world trade, the use of government measures has allowed major importing regions to transfer much of their potential price instability to exter- nal markets. As a consequence, the instability in major exporting coun— tries is greatly magnified and their access to world markets is substan- tially reduced . In view of the above policies, major exporting countries have implemented their own institutions and policy measures to maximize ‘ access to world markets. Among these, price discrimination is often administered by way of a state monopoly marketing board. In 1968, the year U.S. quotas first became binding, Australia implemented an Export Diversification Scheme to regulate shipments of beef to the U.S.6 This Scheme is operated by the Australian Meat and Livestock Corporation (AMLC) which issues licenses or 'entitlements' to exporting firms who must first earn the right to export to the U.S. by making sales to non- U.S. markets. This price discrimination has the effect of directing some Australian beef to lower priced export markets (sometimes at a loss to Australian exporters) and of raising domestic consumer and producer prices. New Zealand also operates a national beef marketing board which, like the AMLC, allows private firms to handle exports. Also like the 6 (An analysis of this scheme is given by Freebairn and Gruen 1977). *| r‘- ' I": 0'; z a .A "‘I O U .l. u u “'01-. l I l I. It 3. . "s, o c -5 fi 17 AMI-C, the New Zealand Meat Board has considerable market power. Since 1955 the Board has used its intervention powers to maintain a price sta- bilization scheme for producers and maintain minimum prices for meat in New Zealand under the Export Meat Prices Act7. The majority of cattle slaughtered are destined for the export market. Argentina's once dominant position as a beef exporter weakened, partly because of foot and mouth disease in that country. Australia and New Zealand, both free of the disease, prospered as a result. Ensuing sanitary restrictions limited Argentina's access to the 'Pacific' market of North America and Japan to canned beef. Argentina exacts substan— tial, though variable, taxes on exports of beef primarily to generate government revenue. To some extent these taxes militate against national goals to augment foreign exchange earnings. Other policies attempt to restrain the high level of domestic meat consumption, which is the highest in the world. Overall, trade patterns and price formation in the world beef market are influenced in a pronounced way by institutional factors. U.S. and Canadian beef imports are mostly unresponsive to prices except during tight domestic supply conditions. That is, normal fluctuations in domestic prices will not affect import volumes. Similarly, Japan's i“IPOrts, which are strictly controlled by quotas, are determined pri- marily by domestic needs and not by changes in the world market. Hence, in these three major Pacific markets which annually accept around 60-70 percent of Australia's beef exports, exporters face a demand which is 7 The Scheme is described by Johnson (1978). L‘.‘ Igl’ '€ Dv u .- . '0'. Irv A. .' u"... ‘ ... '0 I.- .l' . . ,‘ _ 'Iel n 'v-ea ,. I ‘. I 'u, n n U ‘. 'n e c e v i.. fi 18 virtually unresponsive to price. However, trade with the Atlantic seg- ment is highly price responsive. In this market Oceania competes directly with South America. From the foregoing, therefore, it is clear that although the U.S. is an important market for beef exporters, exports to the U.S. have represented a relatively small proportion of total U.S. consumption of beef; less than 10 percent. However, beef imports represent around 20 percent of U.S. consumption of processing quality beef. Exports of beef from the U.S. are even less significant. Various studies conducted over the last decade or so have concluded to a greater or lesser degree that U.S. beef imports have had a small impact on domestic prices of beef and, consequently, on the profitability of domestic beef produc- tion8. Therefore, from the U.S. standpoint, the aggregate U.S. beef market, especially for high quality table beef cuts, may reasonably be considered predominantly as a domestic system. 2.3 Characteristics of the U.S. Beef Subsector Except for quotas on imports, the beef market unlike several other U.S. agricultural industries, is relatively unfettered by government intervention policies. Nevertheless, the beef subsector is an integral Part of the rural economy, and has important linkages with, in particu- lar, the feedgrains subsector and the pork and chicken subsectors. Direct government intervention policies affect the price of grain. And M 8 The magnitude of this impact was the subject of a major investi— Bation by the U.S. International Trade Commission (1977). The Commission concurred with the results of earlier analyses. u ,- 0‘ t" I "’ u ‘ u e . -.a's; at. . h. :1..." n. _ n»: e, “a.- ,g I”, -l' r ' fi 19 since feed costs comprise two-thirds of total production costs in feedlot Operations, these grains policies, significantly, though indirectly influence the beef market. The study of the beef subsector9 or the beef market is the study of a political economy. Consistent with this notion is the subsector representation in Figure 2. 2. Albeit simplistic, the principle partici- pants in the system, the subsector, are identified. Figure 2.2 goes beyond the traditional industrial organization framework of structure, conduct and performance, and includes the structure of a broader environment. The political-economic environment of the beef 'market' m’ ' includes the many overlapping opportunity sets of participants in the system . The political-economic system structures relationships among participants, thus structuring the Opportunity sets for indi- viduals and groups by defining rules for access to resources and pay-offs from the aggregate opportunity set. (Shaffer, 1979. p. 2) This is but one possible configuration of the interactions among parti- cipants and clearly, no attempt is made to detail every component of this system. Nevertheless the important linkages between participants and the commodities produced are emphasized. In this way the operating environment, i.e., the underlying structure of the beef system may be visualized. 9FOr treatments of the beef subsector from slightly different per- spectives see Campbell and Hayenga, ed (1978); McCoy (1979); Shepherd and Futrell (1982); or Simpson and Farris (1982). Ma- terial in this section also draws from U.S. International Trade Commission (1977) and Harris (1980 and 1981). 20 enema-IS Consumers nsmers Organizations «Mainly as «as Voters I :i' e Supermarkets 0 Super Stores 0 Combination Stores e e Fast Food Outlets Catering Institutions Military Restaurants Discount Houses warehouse Stores Table Beef Reta i l ers Food Service Industry 1 lihol esal ers Processors Other Beef Exports ‘ Austral ian Exporters Australian Beef Exports Cattle Non-Farm Environment Aus tral in Government and Aqenci es - lnput Supplies - Financial Sector - Transportation. etc. lion-Farm Organizations Organizations Farm Environment --0ther Farm Subsectors lnsti tutions and Agencies Bureaucracy u .S . Goverment I r '\ J< Figure 2.2 The Political-Economic Structure of the U.S. Beef Subsector In" -:.' bv‘.'“‘ Ju' w I- gnu-Lu“ i ‘0 'be. . I 2.3.1 Beef Cattle Industry 1110 production of beef may be segmented into three major stages: cow-eel! operations, feedlot operations, and meat-packing operations inolming slaughtering and processing. These stages are illustrated in Finn-e 2.3V The central system of production involves the rearing of feeder cattle on pastures and rangelands followed by the fattening of these feeders on high energy feedgrain based rations. Currently about 85-90 percent of all steers and heifers slaughtered in the U.S. have been grain fed; in 1960, 65 percent were grain fed (Appendix Table A2). Not all cattle enter feedlots nor are they slaughtered as veal calves or fed cattle. These so-called 'non-fed' cattle are raised primarily on range grasses or on a combination of grasses and other roughages (that have a low alternative value) and limited grain before being slaughtered. They yield a lower grade of beef than the carcasses of grain fed animals when slaughtered. When feed costs are high, a greater proportion of slaughterings are of non-fed cattle. 2. 3. 2 Heat Marketing Particular changes in the structure of the beef subsector have taken place in an attempt to improve the efficiency of the marketing process (Hard, Henderson and Hayenga, 1978). Coordination systems are very market dependent, and at times, supplies do not match demands. Contracts between major feedlot operators and major meat packers facili- tate dependable supplies with rigid specifications. The importance 0f 4‘ - 5 n &E “ _'.‘ .nld‘od Int II.’ -3. III]... Ilinl ‘ |l..ot|n'.'. .\.I~ ll o‘ 22 x—’ sasmsm oppose comm ecu comm .m.= on» =_ mucma_u_staa $.55... 98:2 moon ecu 238 2,3 .3 9.8: 8.32 we 9.32:3 m.~ ae=m_m .uonuao Humane u on as needs unoe noguoea on ucammuo -oum wow nonnanm on xos an no uemgn menxuamueoa sea as commouOkm xuaam we Adasfiuuam on Awe unsuuao ogb . :uel 69300.3... Indus: . use: 1393 . ye.- nuts-unused: aeammonnd ”annex awwmom ecuusnnwumao unamo~onz acaueunJmmm unannouous neaxmoun acquousuaaum meannwemm caucusuonm 9.3 We exu .§.e 0“. EL} -~ b gm .‘H ....I‘h-' _ a 1'. r. 6 .B! .m 0' 'rv no" . 0‘. ii. .4 .’ e P" L h t. 23 With more exact specifications lies in the emerging trend in consumer attitudes towards greater healthfulness in food with respect to packag- ing ad leaness of meat. This and other consumer trends are discussed 1a.? 0 Beef is processed into different forms at the packer and processor level, ranging from chilled carcasses, primal cuts, boxed beef, boneless beef and ground beef, according to the distributor's needs. Ground beef is one form that has been steadily increasing in importance. The trend towards more processing, such as deboning of carcasses at the packing plant level accelerated with the introduction and sale of boxed beef to retail outlets and institutions. Boxed beef involves the division of the carcass into primal and sub-primal cuts at the meat packing plant and the packaging of these cuts into vacuum-sealed plastic-lined card- board boxes . Returning to Figure 2.3. it can be seen that following the finish- ing phase of production in feedlots, fed cattle are slaughtered to pro- duce fed beef. Fed beef provides table meat cuts, which include veal. Table cuts consist of steaks and roasts, and account for a large part of the beef consuned in the U.S. This meat is usually graded choice or better by the USDA.10 Some imported cattle, mostly from Canada, go directly to slaughter as fed cattle. Non-fed beef cattle are slaughtered to produce some table beef but aainly to produce manufacturing quality beef. Manufacturing beef 1 0The role of beef grades and the implications beef grades have had for U.S. beef consumption will be discussed later. 'u‘ ‘6‘ hwoL" 3' n “2"" '. ".-.I- I‘ll;s ‘96 "u... I -e a “I; ‘ .gu on. Q 1‘ '5‘ 3" flea-e '| . M4 ta a" u.‘ ' l 0, fl ' Va .‘ r- ..g. p: I u 2“ . "a. I ‘ I ‘III‘ a “‘ I ‘ 2” includes all beef requiring further processing (ground, chopped, cooked or ended) and includes ground beef, sausages, and various types of cooked beef. The use of beef for manufactured products depends largely upon the quality of the meat and the demands of the market. As noted above, the bulk of the commercial slaughter is made up of grain fed steers and heifers, and these are the primary source of table beef. Most of the remainder consists of cows and non—fed steers and heifers. Host manufacturing beef derives from cull dairy and beef cows and bulls and meat from non-fed steers and heifers. A further and major source of manufacturing quality beef is imports of Australian and New Zealand non-fed boneless beef. Because of its leaness, U.S. processors often mix imported beef with otherwise 'waste' fat and trimmings from domestically produced fed beef, in their production of hamburger beef, thereby lowering their plant operating costs per unit of production. The average proportion of processing beef obtained from each car- cass type and the respective amounts of processing and table beef derived from each source is presented in Table 2.2 (Agnew, 1979). A feature of these data which is quite often overlooked is that grain fed cattle are a significant source of processing beef. The average fed beef carcass provided 20 percent of its weight as trimmings and other low value cuts for use in processing beef production (including fresh ground beef). This represented absent a third of all processing beef in 1978. Almost 90 percent of imports which enter the U.S. primarily as frozen boneless beef, are ground for use in processed and ground beef Products. Imports accounted for approximately 9 percent of all beef .. _. m Vt n-a'u iiaaanaaiau halal:- uaauavi- l. use-d..- aI-.-aa‘-. .o-.e.u¢- Inn-Islet.un‘ana I||I.‘1al.|. ‘‘II III-.. .lla.‘ . "Cll Iva-II than! i , ota.alla z‘,- II‘ u.-|‘ -‘l.'n.u I a! \ 25 cause .hao>«uoummuu .uuun unannououn Hausa no assumed as can unwound mm mum nuwuuucmsv .mecscc :ouuuun om.c voowuaaoo muoaooun vommouou n new unused acuaaan nu.n nomaunaou moon ecsouov .moon ucscum smunu mousaocHo .oammn unwed: mucuumu a co .uoon wcummuuoun mu uosu ouuaom sumo uo ammucuouon vouaauumu may ma uaouh omnuu>m scan .m:wu==0u ou use muouumm .monmav 3u=m< «cousom ooH n.o~ ooa a.mH cod vq.oa cs "deuce a «4 a m. 3 ed 3 mean seduce: Hm o.c~ mm n.mH om n.m mm "uuumoaon amuoa N o. c I e m. 00H madam uumuaoz 0H n.¢ c c. on m.n om ozoo pumicoz h w.H w N.H o 0. mm enemas: vac museum mohlccz co n.nn mm w.na cm n.n on «usage: vac mucoum com N .mAA N .mAA N .mnA n ouusom downaan sowaHun cowaaam awaammououm cowuusvoum couscoum cowuusuoum vousvonm nowuusuonm couscoum new moon mo annoy no >uwueeso deuce no huauaasc Hausa mo huuucsso mama» owcno>< eowuuomoum cowuuomoum cowuuonowm wean deuce Moon names uuoon . h IQhOd fl.“ WI.“ OHAIH final NGHIOQUOHE HO GOfiua—Aflhuflflfi .m.: N . N CHAIR wt“? ‘I' a: at!“ .' '9 r.» ‘5' J"’ 3" "i m' it" "" i. than aIE’ g... await" “ a 1.0;, u-J'..v. or“ ‘0 loan ‘. a a “'9': "II-I... .~" an: I" ~,' .3. ‘ ‘H J l 3‘“ n .'A . ‘s chained in the U. 3. Australian beef exports have commonly been around 26 50-55 percent of total U.S. beef imports. If it is assumed that all “.3. beef imports from Australia were lean beef of manufacturing qual- 1 _ ity. then these imports account generally for around 10 percent of all processing beef consuned in the U.S. The proportions presented in Table 2.2 will vary somewhat over the cattle cycle as the composition of slaughterings of fed and non-fed cattle varies in response to relative price levels. During herd liquidation, non-fed cattle slaughterings (especially of cows) increases markedly. The supply of processing beef .- increases subsequently relative to table beef production, which tends to i vary less over the cattle cycle. hiring the herd build—up phase sup- plies of processing beef is greatly reduced. Some 98 percent of total U.S. meat production is consumed domesti— cally; the remaining 2 percent is exported. The majority of domesti- cally consumed meat is channelled through retail outlets, particularly retail food stores.11 However, one of the most visable structural changes that has occurred in meat distribution has been the rapid growth in the food service industry (HRI). HRI trade handles an estimated 110 percent of final meat sales and this share has been increasing (McCoy, 1979). One-quarter of HRI purchases are supplied by retail stores. To the retailer, meat is a very important item, generating on average some 25 percent of total gross retail food store sales (McCoy, 1979). The 1 1For a discussion of wholesale pricing in the livestock industry :28. for example, Hayenga (ed.) (1978) and McCoy (1979. pp. 220- 1). i .na’ 5' ‘ A ’17. w' .' N Op A ,l b V‘ {pp A"..', 'M. push. 1‘ .IA «pi. maria: .‘5 u! an N‘ "Isl. llaa' I‘ a ‘a' I ‘ line. u ":00. "“.l' til; .‘. 27 impactof family eating patterns and the growth of the fast food indus- try on constaner demand are discussed in more detail towards the end of this chapter . Although food retailing has been subject to substantial change and develoment many of these changes and developments have not been recent ones. The development of the system of retail chains gained its greatest momentum in the 1930's and 19u0's. Their initial raison detre was in attaining some management control over supplies. With chain growth and mass purchasing economies, came gains in bargaining power. The second major development was the supermarket, also a product of the 1930's and 1940's. The emphasis was on economies of size in terms of volune. By the late 1950's and early 1960's, food retailing was highly concentrated, dominated by a few national and several regional chains. The growth in fast food outlets has been a most prominent develop- ment in meat retailing in the U.S. over the last two decades (Schmelzer, 1981). Hamburger restaurant growth has been evident but so has growth in fast food steakhouses. Much of the growth in these franchised, chain operated, ready-to-eat food establishments has been associated with an increasing preference of consumers to eat more meals away from home. “HEY-Wom-home eating has increased both in terms of the proportion of the food budget spent on such meals and the proportion of meals eaten away from home. Beef, especially ground beef used in hamburger patties, forms a major component of fast food meals and forms a sizeable propor- tion of total sales value. Vertical integration by retailers is not a major factor in the structure of meat marketing. A possible innovation for backward (4| -H “—..—#— l D I w": 1:. ‘5'": " 33.1.?! 3 Whey aneu- in. I";- I'll I‘- a... I up an- on f. .v I“ 'D. a. la‘._' :. “g Q "‘N as a 1 .‘fii’. it a. Ni- a. u ”in... .'"~ I c "J l a 4- I. O :1 .. .a. c»- integration by retailing firms is to establish a brand name and hence to 28 achieve some product differentiation. While such differentiation is possible with processed meats (mainly sausages) the applications appear ‘ limited for fresh meat. For instance, distinctive packaging has not - ' been a major factor in differentiating fresh meats‘z. Pricing of meat at the retail level is a relatively simple matter for processed products (similar to non—food items), since the product bought at wholesale is identifiable with the product sold at retail. This is not so in the case of fresh meat where wholesale cuts must be broken down to retail cuts. Shrinkage costs (losses due, for example, to trimming. spoilage or moisture losses) must be accounted for in the pricing of each cut. Perhaps even more critical is how the wholesale cut will be divided into saleable retail cuts. This will vary among carcasses depending upon conformation of the animal. degree of finish, carcass weight, degree of trim, sex and method of cutting (McCoy, 1979, P- 237). The total value of the carcass to the retailer is dependent, not only upon the retail price per pound but also upon the cut-out (i.e., saleable retail meat obtained). The characteristic that two car- casses of identical weight and grade can yield significantly different Quantities of saleable meat, prompted the adoption of USDA yield grades in 1965. If a retailer can purchase beef according to rigid specifica- tions then he is able to achieve a consistent cut—out range and hence he can more accurately match those supplies to the effective demand for various retail cuts in the store. This level of market coordination, 1 250m success in product differentiation seems to have achieved in retailing chicken. a . a. I .'.e e. Q .~.. ‘1', . '».-.- y.» no. \ ""1 ie'l I a "Ivla J ’ ”f on. ‘zr n. . 1, .‘I'eu ,MJ: \.‘ a VI . ‘u" u '. I 29 though desirable, is difficult to achieve in a cyclical production environment. Not only are product specifications variable, but variable product supplies cause inefficiences at the retail level. The per- sistence of these inefficiencies undoubtedly encouraged such contractual arrangements as exist among Cargill, Keystone, MBPXL and McDonald's which coordinate feedlot operations with fast food distributions. To the meat retailer, the optimal cut-out of a carcass or wholesale cuts, given the prevailing consumer demand for retail cuts, is the basis of a profitable business. However, with no detailed information on the consumer demand for individual cuts, retailers have to use considerable Judgment on the prices they set for the various retail cuts. The prices of cuts which move slowly may need to be reduced before spoilage occurs. If table cuts are not moving and some deterioration occurs then they may be converted into ground beef and sold at a discounted price. Prices may be raised on cuts which are moving rapidly. 2.3.3 Consumption of Beef, Pork, Poultry,_and Other Meats A concern widely held by many participants in the U.S. beef subsec- tor has been the strong downturn, since 1976 record levels, in beef con- sunption per person. This downturn stands sharply against a long-term upward trend in consumption per person for most of the post-war period. Many attempts by economists and industry analysts have been made to identify and quantify those factors determining the demand for beef and in particular, causing the recent decline in beef demand. Factors usu- ally considered in traditional demand analyses are (a) the unit price of beef, (b) unit prices of competing meat products, namely, chicken, pork Q rite}! , f 3e .35 :0 “‘ “. upaAC‘ . i ’ db' ...w~-- rte em: ‘01:“. q'n one '.' cad ‘:Oa' Ilee 00"' ”I. Ce ‘1" 3'... l‘. n“ ‘l e?‘ln .. u e 5 ‘3 e... A, . "" I. I n“. . a. 1‘. ‘ U . a La. I one 30 and to a lesser extent turkey, lanb, mutton and fish, and (c) some meas- ure of income available for purchasing these products, usually dispos- able income per person. Supply factors are also important since, with the exception of relatively small amounts of net imports and beef stocks, production will usually approximate consumption. More will be said later on how supply factors effect quantities of beef demanded. However, some would argue that other factors, less easy to identify and even harder to quantify, are the key to explaining the apparent decline in the demand for beef. These other factors include longer term influences such as changing consumer tastes and preferences, changing eating habits and life-styles, and changes in the demographic composi- tion of the 0.8. population. The intention in this section is to review briefly the main characteristics of and trends in the consumption of beef, other red meats, poultry and fish, and then to analyze some of the underlying forces which are currently shaping consumer demand for beef. 2.3.3.1 Trends in Consumption Beef and veal have accounted for between 50 and 65 percent of all red meat consumed in the U.S. As a proportion of all meat (red, poul- try, and fish meats) beef usually accounted for less than “0 percent (Table 2.3). For many years, until 1976, beef consumption per person generally followed an upward path, increasing from 6“ pounds per person in 1960 to a record peak of 911 pounds in 1976. The exceptional decline in beef consumption in 1973 followed a 28 percent increase in livestock prices received by farmers, the subsequent decision by cattlemen to withhold cattle for herd expansion, and the associated boycott of beef N: Aunt-wen nan-II .I Dwne‘ \l Inn-anew. .nlelq.l!l I‘llII I Pie..- I'D. i.-.- OIA.U‘ UPIEII'fIIIMV III I‘!ll‘o. V 0"! lull..o l!!!- I 'elfllee< .Iel‘ allele .0... II'eI IIIII IIQIIIII'IIllyi I 1 NH I'IeIII I1 Table 2.3 Consumption Per Person and Annual Percentage Chan ges in Consumption Per Person 1960-82 of Red Meats. Poultry and Fish Percents e Annual Chan e: Total Red Meat Lamb Red A11 Heat Edible and Pork Mutton All Heat Heat Pork Beef Fish Poultry Offals Veal Beef .Poultry Year Pounds Per Person, Retail we: ht 31 NHHNIfiOr-INOOMOOGGQOHOOOO NMNQBQIGHmGOnONOONGQOMQ HflonOHQMOOQMQGOfiOnQO¢O I II I M¢NOMHMMMMONQHOHOOHHQN QNC'IOO ONHN£€QQQOFIOZN a: I H?H I II TI 0 O O I. O O O I. 0,... 0.. CO. I I IHI PIHHHHHHHHHFIFI OM”\D“N#G\OOQ~DROI’S~ODNQV\¢QO NONQ mceomcooacoxacannogun gnnnmocqeeeenccsnnn i000 NHHMV‘OMNONO‘OOQQNN‘OHBO‘O‘O OOOOOOOOOOOHOOOOOOOGOGQ ONIOQGOQU‘OMMQQIQNI‘OMOO‘DNQ O O I O O O O. O O O O O O. O 00.... O O‘NQOQQHONHHMMNOtha HHHHHHHHHHHHHHFIHHHFIHFIHH nnoehnonm—aammeowonenncn {QQQMMG‘MMMNNNNNHHHFIHHHH o e e e e e e e s e e e o H HQNNH OM!“ nodes-nos NNOflMMQNON€NQnOQMN€NIflON nfi'CQQ§MMMNNNHHHMMMNv-IHHH NflNO‘OOOONOHVnfi‘OO‘QQNr-IIANN NQHNGAIHO“ hidhfl‘bhh o-aaox adolruq-Ms¢>h-¢:a\c>—Ioa asasosasasa~a~a~a~a~o~o~a\a\a\a\a\0\a\a\a\axes FIFIFIF‘FdFIFeFIFiriFIF4FiFIFiFiFIF194Fiflirird Source: USDA data published in American Meat Institute (1982); and USDA (1983), Livestock and Poultry, LPS-lo, Hay 1983. Sxprnz'g . In I " .L. .4 a II- I. H. on I lur eas' dob: u.... 1‘“. k '64- s '. lr‘..‘ Isl... “ {It as J me... ‘ ‘6‘! I _ .557 'no..‘ . . I ‘z'”' ""» «I. 7 I '5‘ 32'. c hm. e."'. -b 4.! 32 by consuners; retail beef prices rose about 20 percent that year. Surprisingly, consmption per person of the major meats, pork and chicken also declined in that year. Consumption of beef then fell by more than 18 percent to 77 pounds per person in 1982.13 By comparison, pork has shown considerably greater yearly variabil- ity in consunption per person. During the recent years of decline in beef consumption (1976—80), consumption of pork tended to increase, but since beef consumption has stabilized, pork consunption has begun to decline. Veal consmption has trended downward over the long term, except during 19711-77 when cattlemen marketed young calves for slaughter in response to sluggish cattle prices. Generally, it has been more pro- fitable to mature calves for slaughter as steers or heifers than to slaughter calves. Lamb and mutton consumption has steadily fallen over the period. Red meat consumption and total meat consumption per person reflected much of the annual variation in individual meat consmnption. Notably, red meat consunption has varied since 1967 within a relatively small range of 140-155 pounds per person. This, to a large extent, reflects the relatively high degree of substitutability between beef and pork over much of this period. In some respects it would seem that pork has played largely a residual role for consuners of red meat. By contrast, there has been a pronounced upward trend since 1960 in total meat consumption reflecting the steady rise in poultry consump- tion, and to a lesser extent, fish consunption per person. Between 1960 and 1982 poultry consunption per person rose by 88 percent from 311 1 3Retail weights . ,...‘ on ”IS V~ ‘ In, ‘. ' LU‘ ‘ I si nee :W. "‘- " i w 0:" .s V .e: I". 33 pounds to 611 pounds; and fish consumption per person rose by 26 percent from 10 pounds to 13 pounds. Total meat (including red meat) rose 22 percent between 1960 and 1981 from 189 pounds to 231 pounds, but in 1982 it fell 2 percent to 126 lbs. Despite the recent decline in beef consumption per person, this commodity remains for consumers the major source of animal protein. The shares of beef, pork and poultry meat in total meat consumed have varied over the past 20 years or so, however beef still commands the largest share (Table 2.11). For the first time, in 1982, the share of poultry meat exceeded the share of pork. Increases during the 1960's and 1970's in the share of poultry meat in total meat consumption were largely in response to reductions in its relative prices. Variations in the shares of beef and pork consumption in total meat consumption per person over the 20 years, also reflect changes in their respective production cycles. For exanple, because the production of beef closely approxi- mates beef consumption each year, beef consumption patterns will follow changes in the beef production cycle, with prices the adjusting mechan- ism. Thus, the production cycle itself is a reflection of supply and demand conditions in previous years.” Despite the high degree of substitutability between beef and pork, the total share of red meat in total meat consumption per person has 1 “The 'demand-shifting force' of livestock supplies was discussed by Uvacek (1968) and will be considered later. . e. Aw" .I.. II.“ . . u , .'-- I t l'.‘ O. I. —‘ . I \ 'l'. Ines? a Jew- e ., .I'. : ‘6' "r; I" fr I N I 1. .l . ' el.‘ L.‘ Sr. .’. I'lb : :3 ‘e.‘ n . . e's , 311 tended steadily downward. Apart from a small increase in fish consump- tion over the period since 1960, the decline in importance of red meat is largely due to inroads made by chicken into meat consumption in the U.S. Another way of viewing changes in meat consumption is in terms of animal protein consumption. Since 1970 there has been relatively little change in total meat consumption per person (i.e., beef, pork, veal, lanb, and poultry) (Table 2.11). Despite the considerable variation among consumption levels for individual meats, total meat consumption per person since 1975 has been around 226-230 pounds. Protein from fluid milk and eggs has fallen steadily while fish and particularly cheese consumption per person has steadily risen over the past '20 years (National Cattlemen's Association, 1981). The total consumption per person of all these protein foods trended upward to 1970 but since the mid-1970's it has stabilized somewhat around 1135 pounds equivalent per person. The significant point to be made from these data is that future growth in overall meat consumption may be limited, with the levels of consumption of individual meats being determined largely by their rela- tive prices. 2.3.3.2 Price Competition from Pork,_Poult_rl and Other Meats The previous discussion on relative shares of meats suggested that there exists competitive interrelationships among the price of beef and the consumption of other meats, especially pork and poultry. Figure 2.1; illustrate such substitutability between beef and pork and between beef and poultry meat. Simply, in periods of relatively high beef price | u I" av. ‘ a seat. I u a O- . 15.6. .VD‘ 35 'Retail weight. blncludes veal, 1m and mutton, edible offals, and fish. cIncludes beef, veal, pork and lamb and mutton. Table 2.4 Shares of Beef, Pork and Poultry Meat, and Red Meat Consumption in Total Heat Consumption Per Person: 1960-82 Product‘ 1960 1965 1970 1975 1980 1981 1982 Percentages Beef 34.1 37.3 37.8 40.9 33.2 33.5 34.2 Pork 32.0 27.7 27.8 22.6 29.6 28.2 26.1 Poultry heat 18.0 20.6 21.8 21.8 26.2 27.1 28.3 ~ 06m" . 15.9 14.4 12.6 14.7 11.0 11.2 11.4 Total: 100.0 100.0 100.0 100.0 100.0 100.0 100.0 and km" 71.1 68.8 68.0 64.3 64.1 63.1 60.3 Other Heat 28.9 31.2 32.0 35.7 35.9 36.9 39.7 Total: 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Pounds Per Person Total: 188.5 197.4 222.4 214.7 230.4 230.5 225.9 Source: Derived from Table 2.3 36 I10 'OUIO. ". Out..." a a n n w m a ”" “If WI!" .u .n J... .. r» in 1u nu 1“ In In In ru r... I.“ Ct". '35. 'ou'o PIUIDI PIC P2280. "um WNW 1:01 “OJ ma.’ "1 an ‘n C(II. ". Poona mus lacuna, "‘ ’6‘...“ — ICU "ICE "“° mutate cumulus :u rn rn 1” 1” ta 5 ru In ‘ rs 130 no. u t + o m n n ‘ - ‘ tutu" 't' Ducal“ Figure 2.4 ption Per Beef Consum Person and Beef, Pork and Chicken Prices Relationship Between ,.' see: [.251 'V'” we ' I- 85: In. '9 ' 9' O A. we p ‘0‘- v‘g b one 37 rises, consumption of beef will decline and demand for pork and poultry will rise. In terms of quantities consumed per person beef has been preferred over other meats by U.S. consumers. Pork is the next most important meat consumed, followed closely by chicken. However, it is expected that consumers of beef are more responsive to competing meat prices dur- ing periods of tight budgetary constraints. This would seem to explain a large part of the reduced consumer demand for beef over the past few years. Real disposable incomes per person declined and rises in retail prices of beef made beef less competitive with the relatively lower Priced pork and poultry meats. Over the last 10 years to 1982 the undeflated retail price for beef (choice) increased by 70 percent compared with 117 percent for the CPI (all items), 99 percent for the CPI (all foods only), and 61 percent and 20 percent for pork and poultry (broilers), respectively. However, the real danage to beef's competitiveness occurred over the last three to four years. Between 1977 and 1981 retail beef prices rose by 63 percent compared with 59 percent for the CPI (all items) and 119 percent for the CPI (all foods only). Over these years retail pork prices rose by 110 percent, and retail poultry prices rose by only 19 percent. As a conse- quence, meat price relativities have moved sharply against beef. These trends can be seen from Figure 2.5. In the early 1950's the retail price of chicken was around 80 percent of the beef price compared to 30 percent in recent years. Large supplies of relatively cheap pork and Ponitry meat substantially eroded the demand for beef at the retail level. Another factor which has contributed until recent years to the suxowzo use xuom .msom nowusnsmm .mocn canoe mo moaufi>fiusacm ouwum Hamucm 00 a. as, 88:33:88838883338 p p h r P h p n — r P plp b on .8. 8.9.2.3231: r 3. to. 83.823. 9533 II 8 n.N mhswfim in» 2.2.2.3: b p . b «noon-.2. Fun- .0 «e on - D - ecu-conic. pp. sun. 38.8895 .3! sun. uu—axuage I a. r a. 3 e .. u, . J a A .. 8. u .2 .3. U .00 ee\\ §e\ecro once a m u a s. N . 3:...) .. . .3 m 3. m .3... / .... M a w .o: x 3... IL as on. tn» «0 a. 2. 2. oh 2. O0 0. 0. v. u. a. on .0 cu «I on s P . p c s r s p r . p . . p e rs. tutu—:0 II see. u... r hum. dues; I... an {I :8 38» I I’l‘[" ‘II/ I. a o . ()6... 33.1.1...(). 1 3 - III \e‘ld‘e r u .1.......u..I lo... .I on C 6”) .36 1.! 6...la.%.\. I/ J“ I O. u \ /e\e's \e\ sIO‘I” ‘0 ) 3. /\ . u. , . 2 I 8 § 8 8- 2. cu. tit-Oh. fl‘h—o UUIF. ‘8. $0339 ebz' ta: you. I n. u e' ’3‘...u ... 1 1' I 1‘... ‘Op?‘ suede-1‘ ' w-«au ‘ 'p. .ai-l I 9" new; 9“ ‘3. v .‘fli ape lee.- u... ‘9‘! as; - '“‘ snob e Oh so .. 4': 3 a Ute-.1" . 399.13 :1 «u (g... 84““. .' ‘91s r,£ c 9‘07. a ._l‘ ' ‘ I, ‘1 v.. N. u 9|. ”I .A ' d e e I. 0‘. 5s 'e ‘s I: ‘4. e e.‘ 39 widening spread between beef prices and particularly chicken prices has been that production and distribution efficiency have increased more rapidly in the poultry industry (and, to a lesser extent in the pork industry). This has made possible a steady decrease in real chicken prices. Notably, for chicken this price spread with beef has levelled off more recently at 30 percent, possibly suggesting that these produc- tion and distribution efficiency gains achieved in a highly industrial- ized chicken industry have begun to diminish. The potential of the com- paratively traditional beef industry to achieve such gains for itself seems quite limited, at least over the short term. However, the beef industry can be expected not to lose further ground it terms of its price competitiveness with chicken (and pork, whose position in this regard lies somewhere between that of beef and chicken). Later in Chapter 11 an attempt is made to estimate the responsive- ness of demand for beef to changes in the prices of competing individual meats. Much of the past research on the 0.8. demand for beef provides estimates only for the combined effect of competing prices on the demand of beef. A recent exception by Haidacher et al. (1982) indicate that on average the 0.3. consumer decreases consumption of red meat by almost 7 percent for every 10 percent increase in the price of red meat, ceteris Paribus. Also, consumption of red meat per person increases by one per- cent per every 10 percent increase in the price of poultry meat, ceteris Paribus. More specifically, the results of this study indicated that beef and veal consumption per person in the 11.8. decreases on average by 6.5 percent in response to a 10 percent increase in the retail price of beef and veal, ceteris paribus. Moreover, beef and veal consumption per Person is estimated to increase by one percent and 0.11 percent for every e 49 “'A.I .1 u- .no I! . “.:"‘ Ra .gpv. e. r the: area'" 5 I51. u: ' w .95. a e u: e..- IOAA ‘5 l 1 I In "6.9:: '40 10 percent rise in the retail price of pork and chicken, respectively, ceter is par ibus . The cross-effects of the broader meat categories are substantially greater. For exanple, a 10 percent increase in the price of red meats will result in a 5.6 percent increase in the consumption of poultry meat, ceteris paribus. Similarly, the price effect of poultry meat on red meat consumption is greater than the price effect of the individual product effect of chicken on beef and veal or of chicken on pork. This would suggest that for consumers, the major meat purchasing decision is between red meat and poultry meat. Thereafter the decision is which red meat to purchase. This purchasing behavior by consumers would tend to explain the relatively higher cross-price elasticities between beef and pork and between pork and beef than exists between these products and chicken . 2- 3.3.3 Disposable Income and Consumer Expenditures on Meat Retail prices, of course, are not the only determinants of consu- mers' decisions to buy. Changes in the level of disposable incomes also have a significant impact on the demand for beef and, indeed, on the demand for all other items within a consumer's preference set. Consumer purchases are subject to a budgetary constraint on a budget that must be allocated among durable and non-durable goods, services and food. Demand theorists suggest that this allocation process occurs in two stages: in the first stage the consumer allocates his budget among the broad categories of expenditure, for exanple, food versus non-food. In the second stage the consumer allocates the food sub-budget among say perk, stick! sue at .uc: ‘se 11 infigen‘ o use “g per-- ha about- 111 groceries, produce, frozen foods, meats and beverages.”- Arguably, a third stage might involve the allocation of the meat budget among beef, pork, chicken, and other meats. A related decision is also made to con- sume at home versus away from home. The level of real disposable income per person is influenced by developments in the U.S. economy. Factors include the rate of growth in the economy (GNP growth), the inflation rate (the effect on consumer purchasing power), the level and rate of unemployment, interest rates, the level of government disbursements, taxation rates, and changes in the distribution of disposable incomes. There is often a direct rela- tionship between the level of real disposable income per person and many of these factors. The impact of these factors on food expenditures and in particular, on beef consumption, can be substantial. Interest rate levels, for example, appear to have an influence on consumer purchasing patterns, although their impact is difficult to measure. The effect of high interest rates on consumers will be favor- able when consumers act as lenders: conversely, the impact will be unfavorable when consumers act as borrowers or as purchases of products in which interest charges are a substantial component of costs. Either way, it may be argued, that the high interest rates between 1979 and 1982 are likely to have had a negative effect on consumer demand for beef. 0n the one hand, high interest rates will tend to attract consumer's disposable funds into saving and investment schemes and draw funds away from consumption expenditures. This will result in at least 1 l‘SDetails of the theory of consumer demand are Chapter 3. 20 provided in nu tradifii 1.1.. mi *- sue ca: .J u'q In: 5: “A h! w ‘3. A. I la! 11" we no: I 5 Ms of net ‘fI‘."~*upg IDIII slldb I s 3853358 3.3.15 3"“ A PI. ‘I. Sta. I In, ‘v. n" .' I... d 9" 'e..~' 112 some trading-down (i.e., purchasing of cheaper versions of a food item, e.g., moving from choice beef to hamburger or eating more at home) and in some cases in trading-out (i.e., no purchases of the item, e.g., mov- ing from beef to chicken or meat to non-meat foods). On the other hand, higher direct costs such as home mortgage charges, and higher induced costs of food due to higher interest charges to producers, wholesalers, distributors and retailers, may result in a similar reduction in consu- mer demand for beef and other food products. Real personal consumption expenditures per person on beef and other meats increased strongly during the early 1970's to a large extent in response to increases in real disposable incomes per person. The recent declines in beef and pork expenditures have been due to rising retail prices of beef and pork and a strong downturn in disposable incomes. The overall result of these recent movements in real disposable incomes and meat expenditures has been a large reduction in the proportion of consumer income allocated to beef expenditure (Harris, 1982, p. 88). From its peak level of nearly 2.7 percent in 1975. the proportion of U.S. disposable income per person spent on beef has fallen to two per- cent. The proportion spent on pork declined to slightly over one per- cent. Interestingly, the proportion spent on poultry meat has also declined, though slightly, and is at about .5 percent of real disposable income per person . During periods of declining real disposable incomes, therefore, consumers can be expected to reduce real personal consumption expendi- tures, which includes expenditures on food items, especially 'luxury' food items. Most table cuts of beef (steaks and roasts) are generally uttered 11:}: me u "91 “q :bafibs- {3.- II?“ “0' 2? be? so “W:- "'a uhvmv on us' ”a debt. Mn- III ea- 9 : 1!i, — 4'- "Wu 8' tube}. ‘. ‘ 9 135. I! 132313: 'Dse '4' CC? 0" A. I 40'. '.~ (i:- .,' u 11" ‘e u,_ .w: ‘13 considered to be of higher quality than the relatively cheaper pork and poultry meats. Therefore, demand for beef is probably more responsive to changes in real incomes than is demand for either pork or poultry meat. Similarly, demand for the higher quality and more expensive cuts of beef such as fillet steak, is more responsive to changes in real income than is demand for the lower quality beef products such as ground beef. The responsiveness of demand for beef to changes in income is measured by the income elasticity of demand. Time-series estimates of income elasticities of demand will be discussed in Chapter 4. One aspect of the retail demand for beef which warrants further analysis is the interrelationship between retail prices and income lev- els over time. That is, the relatively low consumption of beef over the past few years may have been influenced by the interaction of of income constraints and relative retail meat prices. It may be hypothesized that consumers were more responsive to changes in the price of competing meats during periods of relatively tight consumer spending, as occurred in 1980 through 1982, than during periods of steady income growth, as occurred during the 1960's and early 1970's. 2. 3. 3.11 Some Cross-Sectional Aspects of Heat Consumption Certain insights into the nature of demand for beef, which may not be obtained from time-series analyses, may be gained by examining cross-sectional data. Many of these insights derive from being able to keep certain factors constant in the analysis which may vary in time series analyses. These factors include income distribution, family 3128. age and other socioeconomic and demographic characteristics of the muster. master 1 math; the: exis rgregete 3rcs «pg ‘01 "web’s. In,“ A»: an 5.1.. eev‘u‘vd g ”(3'34 ‘51 l "u, 1.. ”55% Sum. 11‘! consumer. Time-series data will be used in Chapter 4 to examine the consumer demand for beef and other meats in the U.S. The purpose of examining cross-sectional data is to highlight the substantial variation that exists among consumers of meats: a variation that is masked when aggregate market data used in time series analyses.16 Cross-sectional data, used primarily to investigate income- consumption relationships, have been collected during various household food consumption surveys conducted in 1955, 1965-66 and 1977-78 and pub- lished by the USDA (1956, 1972, and 1982).17 The analyzed results of these surveys have been presented elsewhere.18 In each of these studies, estimates of 'income' elasticities were obtained. These estimates may differ depending on the form of the income measure used. The most recent analysis of household food consumption surveys is by Haidacher, et a1. (1982) who examined data from the 1977-78 survey. These data covered 32 categories of meat which included fresh and pro- cessed. Heat categories were analyzed by income quintiles, household size, consumer age group, race, region, seasonality of consumption and type of urbanization. Each of these items were in turn analyzed by 16Certainly, time-series analysis would be more revealing if cross-sectional data were available continuously through time. Such data, unfortunately is not available in the U.S. 17The USDA has also conducted food consumption surveys on a na- tional scale in 1936, 19112, and 19118 (urban only). Only the 1955-’65 and 1977-78 surveys included all four seasons of the year. 18For“ an analysis of the 1955 and the 1965-66 surveys, see Rockwell (1959) and Rizek and Rochell (1970), respectively. For a comparison of the 1955 and the 1965-66 surveys see George and King (1971). For a comparison of the 1965-66 and the 1977-78 sur- veys, see Haidacher, et a1. (1982). Wu a .:.1I"‘ 1 here. F 111., '08} ’52:. .. 1'? ”ES“ "‘11 ‘-\‘m 9’ a‘l ‘ "a... Mil iI‘;e.£ "u. 3351 :1 ‘1 ass 9". a In“; x: ’91:: 1:1» In. .3: ‘b 115 quantity and value of meat consumed. These researchers then compared the results of the spring quarter of this survey with the spring quarter of a similar 1965 survey. Relationships among selected meat categories only will be reviewed here. For instance, consumption patterns of the major meat categories, viz., beef, pork and chicken will be analyzed. Also important to this research are the patterns of consumption of table cuts of beef (steaks and roasts) vis-a—vis manufacturing quality beef products (especially ground beef). Ground beef19 was the largest category of beef consumed per person in the 1977-78 survey, accounting for an average 37 percent of all beef eaten at home. The level of household income, as discussed earlier, is a major determinant of meat consumption patterns in the U.S. However, there are many diverse aspects of this determinant which are simply averaged out in aggregated annual income data. For instance, consumer households in the South had after-tax incomes more than 20 percent lower than other regions in the U.S. Black households had incomes nearly 60 percent lower than non-black households. Low income households consumed at home, more meat in total, although they paid less for it per person than 19Substantial confusion and misunderstanding surrounds the label- ling of ground beef. The term hamburger may be misleading. Although fresh ground beef is mostly sold as ground round, ham- burger, or ground chuck, it is all ground beef with different per- centages of fat. ‘lhe lean-to-fat ratio of the meat is the most important classificatory standard of ground beef and not from where it comes on the overall carcass. Ground round with 10-15 percent fat is the leanest category and ground sirloin is usually the fattest category. Additional fat may be added to the beef up 1:377: legal maximum of 30 percent fat (or 70 percent lean) (Bloch, :tscers tiis cor purpersc: ueats 3': its: pcrk a! cine . . ll‘v} a. .44 was! 1' M, 1‘ Is EIIJ fur In 923 hn‘ "saw. :I {5 M1 ers ." “VIA h“ bum 53': 5;: Iii”; rm. “Hie 'fi{ “ u > ‘~I\. ‘ I I" .1: i‘ I. ‘ 515$ 116 consumers in high income households.20 However, these low income house- holds consumed less red meat, less beef, and fewer table cuts of beef per person. Higher income consumers, consumed more of the high priced meats such as lab, veal, and rib and loin cuts of beef and consumed less pork, poultry (especially whole chickens) and chuck, round, ground and other beef. In terms of quantities, low income households consumed only slightly more than high income households. However, the proportion of ground beef consumed in total beef consumed ranged from 40 percent for low income households to 3'4 percent for high income households. Regional differences were found to be significant for most meat items. Host of the variation was in the type of cut of meat and, hence, in its price rather than in the quantity consumed. The type of urbanization of the household residence also influences the choice and anount of meat consumed. Central city residents consumed and spent more in nearly every meat category than did suburban or non- metropolitan residents. City residents consumed over 90 percent more chicken parts per person than did non-metropolitan residents. This result would seem to reflect (a) the greater anount of home production in non-metropolitan areas and, hence, fewer store purchases, and (b) the greater demand for convenience in at-home food preparation by the larger number of two-wage earning households in central city areas. *— 20A“ consumption data are on a per PEN"n basis ““1933 otherwise stated. Also, consumption data in this section refers to food consumed at home. Measures of meat -consumed from non-household supplied, such as schools, cafeterias and restaurants were not available and, therefore, were not included in this analysis. 'Jw‘u; .1. av. . Ivi! '0”. u p:;'s 9‘ ... run. run :‘r cud. li‘ tut-gm A ,. 1”,;9. VI» 1". '9: an. I ‘Ib. ”up“ ,, I‘ I‘D: .b, ‘3‘ l '~u. Its: a.“- :ulF’ - c‘ ..-‘ n,‘ N7 Household size was a factor which influenced meat consumption in the U.S. Small size households consumed more and spent more per person on meats than larger households in virtually every meat category.21 Rea- sons for this characteristic may be that (a) small size households are usually of adults only who consume and spend more per person than house- holds with children, which are more common in large households, and (b) two-person households are more likely to have two income earners (except for retirees), with a higher income elasticity for food and a propensity to eat more expensive foods. Age of household occupants is therefore important. In fact, differences in the consumption levels among the various age groups were significant in nearly every meat category both in terms of expenditures and quantity consumed. The heaviest meat eaters in the majority of meat categories came from the 140-61! age group. A notable exception was the 13-19 age group wherein 418 percent of total beef was eaten in the form of ground beef, compared to no percent in the 20-39 age group and 31 percent in the 130-61: age group. Smaller quanti- ties, but similarly high proportions of ground beef were consumed by the youngest age groups (0 < 5, 5-12). Possibly the heavy promotion and advertising associated with the rapid growth in ground beef-based fast food outlets (e.g., hamburgers, tacos) have influenced at-home consump- tion patterns among the younger meat consumers, in particular, the teenage group. Certainly the more expensive table cuts of beef are con- sumed largely by the older age groups. Other notable exceptions, con- sistent with the consumption pattern of ground meat are chicken parts, Some of these differences, however, were not statistically sig- nificant, for exanple, as was the case for ground beef consumption differences. ‘.~vi mwuup‘ r" u ‘A . p - ‘ 3.5%. 4v . 0n cl . “An 5:“ push” u b M .c "and u ELI: E.’ 1'.‘ Ah. ‘"V In.‘ “up. I»@_ 'fl’ I4. I '0)” a: (. '9 , l').. .4" - "fa '| N w i us processed chicken and franks, which are eaten at home mainly by those under 20 years. A further factor providing a considerable influence on meat con- sumption patterns in the U.S. is seasonality. The seasonal variation tends to be associated with weather conditions and holidays. For exam- ple, meats such as ground beef, steaks, chicken parts, franks, and fish, which are frequently cooked on outdoor grills, are consumed more heavily during the spring and summer. Conversely, meats such as whole turkeys and chickens and beef and pork roasts are more heavily consumed in the fall and winter, reflecting in particular the Thanksgiving and Christmas holiday periods. Overall, however, these seasonal variations tend to counter each other such that total meat consumption, red meat consump- tion and also beef consumption remained relatively stable throughout the year. Comparison of 1965 and 1977-78 Surveys: Again drawing from the Haidacher et al. (1982) study, a snapshot of household meat consumption and expenditure patterns in 1977-78 is compared with a similar snapshot in 1965. As before, this review will focus on the major meat categories and on the within-beef categories, especially between table beef cuts and ground beef. Data on changes in at-home and away-from-home eating patterns reveal substantial between-survey increases in the percentages of meals eaten away from home for every meal occasion, regardless of sub-grouping considered. In both surveys, meals eaten away from home increased with increasing income levels, although between 1965 and 1977 an increased percentage of meals were eaten away by lower income groups. It is likely th; fans ma lunar inc '5‘. F 9": I 56% f is 3:21 33.5. F! {Erna-e It's e .A-b ‘ 'V':.'.S 51:»:- .m— “ I‘u‘ dif'Au (”a u... ' mg i" ‘19 likely that the relative cheapness and convenience of fast food restau- rants may have contributed to this increase in away-from-home eating by lower income groups. Changes in consumer lifestyles and the increased proportion of young adults may also have been factors. Comparative details of consumption of individual meat items are available only for at-home consumption. Perhaps the strongest 'trend' between survey periods was towards the more convenience-type foods such as ground beef and chicken parts, and away from the heavier table beef cuts, roasts and steaks, and away from whole chickens and turkeys (except during specific holidays) (Table 5). A further tendency was towards increased fresh pork consumption per person and away from con- sumption of processed pork products. Ground beef and chicken parts con- sumption per person rose by 51 percent and 75 percent, respectively, while consumption of table beef cuts showed a tendency towards no change or a decrease between survey periods. The proportion of the total meat budget allocated to red meat declined from 70 percent to 66 percent. As would be expected consumption of total meats, red meats, and beef increased with increases in income level per person. Consumers in low income households increased their consumption levels of total red meats between the two surveys. There were substantial increases in ground beef consumption per person in each income group, with the larg- est gains occurring in the lowest and highest income groups. A similar Phenomenon occurred for chicken parts, except that the largest growth was in low income groups. The negative relationship between income and pork consumption per person in 1965 was more pronounced in 1977-78. . 9.. . I .34. His: 52:12. 1:,- . 1 its. “5 I v . all. 3:; r“:- x .5. ‘4 50 Table 2.5 At-fiome Meat Consumption and Expenditure in the U.S.. Spring 1965 and Spring 1977 1965 1975 1965 1975 Allocation of weekly At-Home Meat Budget Consumption Per Person $ 3 lbs. lbs. Total Heats: Nominal 124.28 252.77 4.42 4.39 Real (1967 - 100) 131.51 139.27 2 Z Z Z Total Meats 100.0 100.0 100.0 100.0 Red Meats 70.1 66.4 62.0 61.5 Beef 42.1 41.0 36.0 39.4 Steaks 6 Roasts 31.5 26.6 23.5 22.8 Ground 7.7 11.7 9.3 14.1 Other 3.0 2.5 2.9 2.5 Pork 25.1 23.1 24.0 20.5 Fresh 8.1 8.5 7.7 8.0 Processed 8.9 7.4 7.9 5.9 Bacon & Sausage 8.1 7.0 8.4 6.6 Veal 1.5 1.2 .9 .7 Lamb, Mutton, Goat 1.5 1.2 .9 .9 Poultry 11.4 12.4 18.6 19.4 Chicken 10.3 10.9 17.4 17.3 Whole 8.1 6.2 14.3 11.8 Parts 1.8 3.7 2.7 4.8 Processed .4 1.0 .4 .9 Turkey .7 1.6 1.0 1.8 Whole .3 .6 .9 .9 Parts .4 1.0 .2 .9 Fish 6 Shellfish 7.7 10.5 8.1 8.4 Fish 6.3 8.5 7.2 7.5 Shellfish 1.5 1.9 .9 .9 Miscellaneous 11.1 10.9 11.5 10.7 Franks 3.3 3.1 3.6 3.4 Luncheon Meats 6.3 6.4 5.7 5.5 Variety Meats 1.5 1.4 2.3 1.8 M Source: Errors due to rounding. Derived from Haidacher, et a1. (1982, pp. 56-57) ks: eu'hiztu: ”0"ch ‘0'. 4 II“:A:.“ ..-wavdulw an II - Must?‘ I “{ .lv h... _c| o J bi: I. g... ‘ u .9. 1;. Mind“, U Us: -‘ n...‘ I u ..‘!‘I~ "Oiw .. fl" thi" . «up. “1(9- 'lF‘a. ‘." 1x- a D 51» Most notably, for every major meat category, except chicken, the expenditure and quantity elasticities were smaller in absolute value in 1977-78 than in 1965. As concluded by Haidacher et.a1. (1982), this may indicate a declining preference in the demand for meat and meat items consumed at home . Some important problems and inconsistencies in these results and in the elasticities estimated in the Haidacher et a1. study were evident. This may be in part due to the omission of away-from-home data in much of the study. For instance, away-from-home consumption of meat is thought to be considerably more income responsive than is at-home meat consumption. Nevertheless, it appears from the study that consumers are switching increasingly to the more processed, convenience-type meat items, in particular, ground beef and processed poultry items such as chicken parts, processed poultry and turkey parts. Also, most of the various socioeconomic and demographic variables analyzed were found to have had a significant influence on meat consumption, especially on an individual meat item basis. 2.4 Current Forces for Changewin Consumer Demand for Meats In the previous section it was concluded that a distinct shift seems to have taken place in the consumption per person of the major meats in the 0.5. A major shift has been away from red meats, particu- larly beef, and toward white meats. More particularly, the transition seems to have been away from the traditional table cuts of beef, i.e., steaks and roasts, and from whole chicken and whole turkey, toward pro- cessed ground beef, chicken parts and to a lesser extent fresh pork. Clearly, H “Menu! .:I'l ‘."c‘ «4 ., ,- u'r i.ele. 5b a 4 men. .63 A e 4' h rescue" in ‘009 r‘ ‘ "any hula. wn‘ ‘te r alt o In 3r per “I E: ‘ar hen... .- 52 Clearly, changes in relative meat prices and income levels are major determinants of the market demand for meats at the retail level. When viewed in aggregate, these two factors explain a substantial part of the recent decline in beef consumption in the U.S. This conclusion has been reached by a number of researchers?2 Notwithstanding this observation, some analysts of the U.S. meat, markets argue that prices and incomes are not the only factors explaining this downturn in total beef consumption per person since the mid-1970's. Certain non-price influences are also considered to have had a significant impact on the retail demand for beef and other meats. For instance, changes in the demographic charac- ter of the U.S. population, changes in the distribution of incomes, changing consumer habits and tastes, shifting attitudes to health, changing preferences for convenience foods, increased frequency of eat- ing away from home and changes in the overall beef marketing system have been credited with some of the changes in beef consumption. Indeed, some of these developments in food retailing and consumer responses may have been price and income induced. Therefore, it is within the context of changing meat prices and changing income levels that some of these other factors influencing beef demand are now considered. 2.14.1 Impact of Age Structure and Size and Composition of Consumer Households Changes in the structure of the U. S. population appears to have had a profound influence on the consumption patterns of particular food items. There exists considerable evidence that differences in economic 2 2For example, see Haidacher, et a1. (1982) who concluded that 95 Percent of the variation in quantity demanded of meat products is e"Plained by changes in prices and income. A Opu- Eu. 5* U- 0"" all 0'». ul u ’waz" un- .01 V new "i .b bfllu the .5.) ‘Duu. ‘ V ‘ I. ”c. h '. .II 53 behavior are systematically associated with differences in age.23 The post-war baby boom has had and will have a continuing impact on the size and structure of age cohorts as the babies in that boom grow up and move through those cohorts. That boom began in 1946 and lasted until 1965. Thereafter a substantial drop-off in births occurred, which in time will have its own impact on consumer behavior. Therefore, people born during the baby boom years are now aged between 18 and 36 years. During the mid-1960's through the 1970's the bulk of this group would have impacted primarily on teenage-oriented markets. The group's impact on the labor force began when the group's first members reached working age in the mid-1960's. As this group swelled, the number of teenagers and young adults available for work increased and rapid growth in the labor force followed, causing unemployment rates to climb (Russell, 1979).2“ Over the next five years as the baby boom group grows older and the smaller cohorts of people ‘born since 1965 begin to enter the work force, the character of unemployment will change further. This will not, however, be the end of the baby boom group. Its impact will manifest itself in different ways in older age cohorts. Changes in the age structure may have important effects on both the amount and composition of consumer spending in the future. For example, it may be inferred from the previous section that an increase in the luhéfl age group relative to the 20-39 age group will cause a 20-30 per- cent increase in the consumption per person of all meats, red meats, 23For a discussion of this literature, see Russell (1979). 24 Under normal circumstances this particular group has higher unemployment rates than older workers. ‘3: Abs 5 \ “v... '0' m ft“ A'm‘ I" Leia we I ll ‘ Snug], i 'ra‘ni 5. .m'u he”; a d .1 I I?! ' h aces i.‘ in .3556 ‘au. 0... .-‘ 'I O L": 0“ '4‘ (x) 52:: 54 beef, chicken and fish. Similarly, greater percentage increases would occur for pork, turkey and high priced veal and lamb. Consumption of ground beef and chicken parts, highest among the 13—19 and 20—39 age group, would tend to fall as the baby boom group grows older and moves beyond their cohorts?5 However, the logic of these conclusions rests heavily upon the assumption that as the younger consumers age and move into older age cohorts. they change from being, for example, heavy ground beef consumers to being heavy table beef consumers. This would depend greatly upon the degree to which they carry tastes and prefer- ences into older age. The composition and size of households also change over time. These changes affect people's spending and food consumption patterns. The decline in birth rates since 1965, the tendency towards postponement of marriage, and the rising number of divorces all point to proportion- ally more single people and childless couples among the young households of the 1970's and 1980's than were among those of the 1950's and 1960's. Three major demographic trends have developed concomitantly. First, there has been an increase in the number of single people, especially in the 2040 age group and an increased number of single person households, some of which would have been occupied by old people and/or widows. Mitchell et al. (1980) estimated that on average single person house- holds have higher per person annual incomes. Spending of food by single Person households has-been higher on a per person basis than multiple Person households (Sexauer and Mann, 1979). They also spent nearly 100 __ 25 A similar conclusion was reached by Mitchell, et a1. (1980, p. 6). an 1’ v. vi". 6 91:8? C ..J .3 than t A Oh; {an inn A'uv' .4” 21:59 In . “I 4qu I" . Ui- ‘Ffi! 1‘“ :‘a duly“... ‘5' gap tvu i=1 . . ":u‘- saau5,l q ":3! I O *-I la» ‘ J ' “HI.‘ '4\J:‘ A e.- 'b 3., 55 percent more on meals away from home. Second, there is an increasing number of two-income households. Third, perhaps the most significant trend of the 1970's and obviously related to the above two points, is the increasing number of working women. Between 1960 and 1980 the per- centage of women who were working rose from 33 percent to 48 percent (Beef Industry Council, 1982). The last of these three development has been due to a number of factors. The baby boom contributed to a natural increase in the percentage of women eligible to participate in the work force. The recessionary economy, high unemployment and the inflationary squeeze on incomes during the 1970's and early 1980's made it necessary for many women to supplement the household income. Also, the changing lifestyles of and attitudes toward working women encouraged or at least made it possible for women to develop careers. The effect of this trend is not straightforward since there exist many types of working women, i.e., with or without a family, profes- sional or non-professional, or single partner and no family. The age SPOUP 45-54 traditionally has supplied the largest labor force partici- pation rates among women. However, relative increases in numbers of working women have been greatest among those under 35 years of age (3433911. 1979). A result of these changes is that men are now doing more of the shopping (Mitchell, et a1., 1980; BIC, 1982). To some extent there have been changes in the traditional roles played by men and women. This may in turn change the overall complexion of consumer behavior as it was once known. There may now be a need to know more about the shopping demands of men. For example, it is alleged that, in Seneral, men Shop faster than women. ‘v 'uaih' .3‘~.»I\ a a u u —: 1 i “a bow tau-nu lequ-v :‘N" nu the r. 23:. a such as :IAI‘ H, on. m . Trig. in” 'n- 1 b. .- '- 56 2.4.2 Changing Lifestyles and Consumer Health and Nutrition Attitudes Toward Meats An overall effect of the above demographic changes has been a change in consumers' lifestyles. With more two-income earners per household, free time has become scarce and leisure time has become more valuable. Hith less time available for shOpping purposes, eating out becomes more common or, if eating at home, convenience in shopping and food preparation becomes more important (Stafford and Hills, 1979). In turn, a derived demand has developed for time and labor saving equipment such as microwave ovens, slow cookers, hot—dog and hamburger cookers and food processors. The demand for convenience in shopping has resulted in the growth of supermarkets, superstores, and hypermarkets, which emphasize one-stop shopping. Convenience in food preparation is usually achieved by a greater degree of food processing, as reflected in the increased demand for chicken parts, processed chicken, and ground beef in contrast to whole chickens and beef roasts. A 1978-79 survey of 166 convenience foods showed that 59 percent of these fOods cost more per serving than the ingredients for their fresh or home prepared counterparts; 13 percent cost the same and 28 percent cost less (Traub and Odland, 1980).26 On average beef convenience foods were 45 percent more expensive than their home prepared counterparts; all—beef patties were 81 percent more expen- sive. Pork foods were 35 percent more expensive, chicken fbods were 64 percent more expensive, and chicken meat was 154 percent more expensive than when prepared at home. Although many of these processed 6 It was not clear from this article how labor and energy costs were allocated. I" F‘fi' :c.,le“o\-I wall nf' 5a.. .1 . .‘1‘ pm: Jude-“3 0 u: sax. ipgpu' A] I‘Iiu .Du 1n 5h; ml at.» ' As .52! u 57 convenience foods are, in general. more expensive, any premium paid for such processing would usually be of little'concern to a two-income household. That is, for many. the convenience outweighs any shortfall in quality or additional cost. A typical response of consumers to stagflationary pressures27 and deteriorating real incomes such as have existed in varying degrees dur- ing the latter part of the 1970's and early 1980's, is to 'trade down' to a cheaper form of the product, to buy less of the same product, or to 'trade-out' of that particular product and buy a different, though sub- stitutable product. Consumers (except perhaps for the very wealthy) have made this response in various ways over the past several years. As noted earlier, high beef prices or reduced consumer budgets caused con- sumers to buy less or trade-down from expensive steaks and roasts to cheaper table beef cuts or to ground beef. Some consumers traded out towards chicken and even towards non-meat products. For slightly dif- ferent reasons, for many consumers enduring a recessionary squeeze on their incomes, the greater consumption of convenience and/or processed foods, and the increased patronizing of fast food eating places was an obligatory trading-down. For others, many of whom were up-scale shoppers, such behavior was in response to a desire for greater shopping value. The rapid growth in generic foods was manifest evidence of this desire .(Anon, 1981a). Upward spiralling beef prices imposed substantial 2‘IStagflation, a term coined in the 1970's, describes the situa- tion when an economy is experiencing inflation during an economic recession, i.e., when economic activity is said to be stagnant but inflationary. I?! 33:13, O . a. :u. u in ‘ ’ In at: fr! ‘37. IN 1.. “”3" "Wm. “have ‘ 'vnl. 58 cost pressures on restaurants specializing in beef menus. As a conse- quence, many restaurants, especially fast food places, introduced non- beef meals and sandwiches, namely chicken and fish, as a protection against the rising cost of their beef inputs (Anon, 1981b, Anon, 1981c). Yet, clearly, no one lifestyle has been adopted by all consumers”. For some, priority is given to more holistic preferences and a concern with freshness, nutritional value and healthfulness. This holistic con- cern, itself a result of a more educated consumer, has resulted in, for example, a greater demand for product labelling. Health and nutritional concerns have become much more important to American consumers. A recent USDA survey indicated that nearly two-thirds of those surveyed said they had adjusted household diets in the past several years for health or nutritional reasons (Jones and Heimer, 1980). Weight control was stated as the main factor influencing food choice, and reducing fat intake was the major reason cited. The move away from red meats and toward white meats, some people have contended, has been in response to this concern. Recommendations by various health organizations and government agencies that meals should contain more fish and poultry (which have less saturated fat than red meats) and more lean or trimmed meat cuts, may have furthered this move away from red meats (LeBovit and Boehm, 1979; Longen and Stucker, 1980; Jones and Weimer, 1980)”. 28For a discussion of various consumer lifestyles which have helped shape food retailing in the U.S., see Mitchell, et al. (1980) and Hamm (1983). 29111 1977 a U.S. Senate Select Subcommittee issued 'dietary goals' which specifically recommended that consumers eat less fat, espe- cially animal fat, to prevent cholesterol related health problems, among others. The American Medical Association has made similar recommend ations . e 5. a1 1:. 6h ‘3 p wan-I .92....2... u. “ .4": JCS v5 r‘ us 1" m9. 0“ unit}. ‘054 0.1 owl-‘9. ‘ Q "r :u‘ 'le-qy' 59 Nevertheless, there exists considerable controversy surrounding these recommendations. The tenor of these recommendations is probably not in question, but rather the appropriateness of these very general recommendations for healthy people. For instance, the Food and Nutri- tion Board of the National Academy of Sciences recommended in a recent report, Towards Healthful Diets (1980) that only sedentary peeple attempting to achieve weight control should be advised to reduce fat intake. The Board made no specific recommendations about dietary cholesterol for healthy persons. Another institutional concern held in some quarters is the feeding of' antibiotics to farm animals, especially to beef cattle (Burbee, 1980). Almost one-half of all antibiotics produced in the U.S. in 1978 were administered through farm animal feeds. They are used to control animal diseases and ostensibly to promote growth. Opponents of their use state that the offending organisms develop resistance to the antibi- otics, raising the possibility of organisms, now safe to humans, transferring their resistance to harmful organisms that could infect humans and/or farm or wild animals (Roberts, 1979). A ban on antibiotics would reduce meat supplies and raise prices to consumers (Burbee, 1978). Meat consumption would fall, albeit slightly, although with relative price inelasticity of meat demand, gross revenue to farmers would rise. Higher farm income from the ban may increase meat supplies but it may take some time befbre supplies reached their ! Av'1"3 u. s‘t'" . :- I In- In :15 t' m. " ”I '5 AI: ..,*_ D: I“. a; ‘I 0" - 60 original levels30. 0n the other hand, current adverse attitudes towards the healthfulness of meat may already be reducing meat demand. Such drugs are the most frequently detected chemical contaminant in meat (Roberta. 1979). However, to conclude that drugs (antibiotics and sul— fur drugs) administered to beef cattle account, at least in part, for the recent decline in beef consumption may be premature. These drugs are as frequently detected in poultry meat, and poultry meat consumption has risen markedly. 2u4.3 Consumer Preferences and Beef Grades Despite reservations about certain attributes of meat, in particu- lm' beef, consumers have indicated a strong preference for beef. A study conducted in 1975 on consumers‘ perceptions of benefits of meat items scored beef highest in nutrition, ease of preparation (convenience factor) and family taste (Sun, 1979). On price, beef scored worse than poultry and fish but fared better than pork and shellfish. 0n nutri- tion, poultry also ranked behind fish, second in ease of preparation and 30A proposal is presently before the Food Safety and Inspection Service, USDA to require that countries exporting meat to the U.S. submit their product to a residue testing program. The proposal requires that the standards applied to imported meat and meat pro- ducts be at least equal to those applied to the domestic 0.3. pro- duct. However, the potential for this proposal to be essentially a non-tariff barrier to meat imports is reflected in the catch-all sentence that ‘testing would be required only for those substances known to be in use in the production of meat and meat food pro- ducts in the particular exporting country or otherwise known to be present in the environment of such country' (Office of the Federal ReSister (1982)). The arbitrariness of the interpretation and hence the impact of this proposal is obvious. For instance, since most antibiotics are used in feedlot cattle and most beef imports are range fed, antibiotic levels are not in question, but of course, any other substance may be questioned. ; Os . ,i‘ .35.? ”a: a I a", Ila-‘1'? ; y". . in p. ‘1'... -e‘)$. dial L4“! a whim in" 3 NR. 1‘ V! us‘ ‘uu‘r: “we. . ‘Il Au sub u. ‘ :Gi‘ I 61 in taste, but best in price. These attitudes may have changed since 1975. a year of high beef prices. Some researchers have argued that this change may have in fact occurred as reflected by the change required in the role of the meat department in supermarket stores (Allen and Pierson, 1980). There has been a substantial change in the format of stores retailing food in the U.S. The move has been away from the traditional supermarket which has relied on the meat department to generate store profits. A prolifera- tion of new products, new packaging concepts and marketing formats has caused fresh meat sales to lag behind other food product sales. Today's consumer is oriented toward improved value, greater convenience, a desire for more information and change. This change demands a rethink- ing of product development, packaging and merchandising of meat. There is a need for consistency of trim, thickness and eating quality, and for better information on preparation. There also exists a need to develop new packaging systems. Boxed beef marketing has been a major development in this direction. Other trends which have not yet run their course include the growth in boneless cuts, the expanded variety of meat cuts, family packs, 'chub' packs for ground meats, and microwave-ready meats. UPC (Universal Product Code) scanning has not yet made a major impact on beef retailing but can be expected in the future to provide detailed information of demand for individual meat cuts and to facilitate a closer matching of supply and demand for meat at the retail leve1 (Walsh, 1977). Given this information, retailers will be able to more accurately determine the most profitable carcass cut-out, and provide consumers with a product which more closely meets their particular demands. ff uh MAP '0' w'vn a-»- m- .« 1-3 3.1L any:fi'1 15.? nursing: Vvlwfi—uu on r. due 58 9!: up nu v y o—‘ :51 '- 62 Another, perhaps more direct way of'meeting the specific demands of consumers for meat products is to market a product more closely tailored to consumer requirements. For a number of years evidence has emerged in various forums by various participants in the beef marketing system that consumers want leaner and less expensive beef. Literature supporting this observation has already been cited. Allen and Pierson (1982) quote the experience of Safeway, the nation's largest retailer. Safeway feels it successfully shifted from the USDA choice grade of beef, which for decades was the focus of its merchandising program, and now sell mainly leaner, no-roll beef products. The means of providing this differen- tiated beef product is to develop meaningful beef grades. The issue of beef grades in the U.S. has had a relatively long and tempestuous history which will not be entered into here31. The central aim of grading is to improve pricing efficiency, the accuracy of iden- tification of product value, and overall production efficiency. Net all beef is graded. In 1980, 56 percent of the commercial beef supply was graded; 75 percent of beef cattle were federally graded (Burbee, 1982, P- 17). Of that beef graded, 5.9 percent was graded Prime (the top beef grade). 89 percent was graded Choice, and 4.3 percent was graded Good. Most beef that might have been graded U.S. Good or Standard was not graded. Grades designate differences in color. tenderness, juiciness and flavor. Therefore, grading would seem to be closely tied to consu- mer preferences. Grade standards must be adjusted from time to time to accommodate new technologies for raising cattle, new techniques for M 1 3 For some of this background see Nelson (1977). MCCOY (1979) and references contained therein. 64w ”algavb, urcmef l ' Aq-M was... 1 In! an in F: vi “v a : m... “N‘f‘r: ‘V'lii 1.. 72‘".- l 1“ hr" ~- ...‘I . I‘fl€‘ a. IHHL v 63 evaluating beef quality, changes in economic conditions, and shifting consumer tastes and preferences. Changes in grades have implications throughout the entire beef marketing system from producers to consumers. If grade changes achieve their objectives and truly reflect consu- mer preferences with respect to palatability (eating quality) and value, then a structural change in the demand for beef will occur. Theoreti- cally, the increased utility from the consumption of beef will shift the demand curve for beef to the right, ceteris paribus; otherwise no signi- ficant change will occur. Researchers have attempted to test for this response following the 1976 Federal grade changes (Purcell and Nelson, 1976; Nelson, 1977). Virtually every study on consumer preferences for beef has shown a clear preference for leaness in meat appearance. Partly in response to this preference, the 1976 grade changes were designed to decrease the degree of marbling to qualify a carcass for Prime and Choice grades and hence, increase the proportion of carcasses grading Choice and Prime. Nelson (1977) examined monthly data on beef consumption by grade covering a period before and after the introduction of grade changes. Using a zero-one dummy variable for before and after grade changes, no significant response was found; that is, no shift in the demand for beef resulted from the grade changes. The study's results indicate that the grade changes achieved the objective cf Aden: 5:.an l ...uv’ I Reverie * h spen. . “A! a u IA'h ’v t . u us: ' OI I‘IIEI VI 'mh'... ‘Vanuu ‘uu‘u‘ 'Ivvg‘l 64 of widening price differentials between quality-yield grade combina— tions, suggesting that between grade price spreads had increased32. Nevertheless, there was little evidence found in that study that time spent by cattle in feedlots had been reduced and hence reducing produc- tion costs or that today‘s beef cattle are leaner than those graded under the previous standards (Burbee, 1982)”. The dilemma which faces those charged with designing the optimal system of beef grades is, as many studies have shown, the large amount of confusion and unfamiliarity which consumers generally have with USDA beef grades (Smith and Heimer, 1980)“. The main confusion or contrad- iction exists between preferences as revealed in sensory tests (eating tests) and preferences as revealed in visual tests. Visual tests indi- cate a preference for leaner grades at equal prices or even with price differentials against beef35. 0n the other hand, eating preferences were very rarely revealed for the leaner grades. Despite the apparent inconsistency, this result does show some identifiable characteristics, 3zlwith more beef of grade Good being graded as Choice beef, consu- mers became concerned at the time of the grade changes, that they would be paying the same Choice prices for beef formerly graded lower. It was a question of whether producers‘ savings in produc- ing less costly beef would be passed on to consumers. Price dif- ferentials widened, i.e., implying product differentiation, but beef demand did not increase. 33More beef has been graded Choice since 1976, however, since Shading is voluntary, a simple comparison of the grade distribu- tion of the officially graded carcasses grossly misrepresent the grade distribution of all beef carcasses. 3“See also for example, McCoy (1979). Reidy (1980), and Purcell and Nelson (1976). 3‘S’I‘his result was obtained almost 30 years ago by Kiehl and Rhodes (1956). The same conclusion was found in survey results published in 1980 (Smith and weimer, 1980). we ‘r' 3:. dos"b a”: ‘C‘ 3‘ um ku n, . ‘Av I n.4,. W‘s me ’n. -u 65 such as preference for leaner cuts and a positive relationship between marbling and palatability. Certainly, some of the misconceptions and confusion could be alleviated by improved consumer knowledge and market- ing information, as well as further refinements in the beef grading sys- tem36. Part of the problem may be that consumers have begun to lose confidence in the grading system (Burbee, 1982). Although consumers have been infbrmed fer years that Prime is superior to Choice which is superior to Good, many consumers, concerned with their nutrition and health, want tender, juicy meat which is nearer in leaness to the ‘lower‘ grades of Good and Standard. Essentially, this means that pro- ducers, packers and retailers have yet to market a product which satis- fies the demands of the final consumer in the marketplace37. 2.4.4 Changes in Family Eating Patterns and Growth in the Fast Food Industry In analyzing the U.S. meat market, it is worth remembering that there is no such person as an average consumer. Part of the failure of beef marketing has been the absence of any targeting of promotion and advertising to specific consumer segments which will provide the greatest return for the marketing effort expended38. The diversity 36See veblen (1977) for an overview of these and other problems existent in the U.S. beef system. 37Currently, alternative beef grading standards are being examined by the Agricultural Marketing Service of the USDA. Proposals have been submitted by, among others, the National Cattlemen‘s Associa- tion and by the Community Nutrition Institute. For details of the main alternatives see Burbee (1982). 38This specific issue was addressed in 1981 by the Beef Industry Council of the National Livestock and Meat Board. The Council as- signed a marketing consultant the task of developing a marketing strategy, including national advertising and promotion for the Purpose of ‘returning the beef industry to profitability.‘ This strategy is contained in Beef Industry Council (1982). man“ "N‘s-s to CE. EX ' 'u 15 :31. 4.!‘0'! .hlkg: U ‘94 an; Iu.‘bd, :czseq ‘Ilzn‘ "‘wa I . u an , rule} o u,” ’ Hm , rt“; 511.. D .M: ..._ A I.. U A 'I'.‘ ”N. 'y'.: :Is sq. 66 among consumers of beef and other meats is evident from consumer atti- tudes towards fresh versus processed foods. For instance, for many con- sumers fresh foods are more attractive from a nutritional and health 'standpoint. However, for the most part, the trend toward convenience is essentially a trend toward processed and/or frozen foods. Hence, one can expect demand by consumers to move toward a variety in foods which is tailored more towards a range of diets and lifestyles. Many of the demographic and socioeconomic changes among U.S. consumers reflect this tendency. In particular, developments in family eating patterns are a consequence of the collective impact of many of these changes. This impact on the demand for beef and other meats is likely to be of a longer term nature. For example, in many dual income households, break- fasts as a weekday meal have all but disappeared -- hence, the consump- tion of breakfast meats like bacon and sausages has declined. Moreover, with a family of four or more, it is usually cheaper to prepare meals at home than it is to eat out at a restaurant. However, with the trend toward smaller families and working parents, eating out becomes more attractive. A parallel development of these trends is a decline in the traditional family (i.e., housewife, employed husband, and children) and the traditional sit-down family meal. This particular trend explains much of the decline in per person consumption of beef and pork roasts, and whole chickens. A major characteristic of the change in eating patterns in the U.S. has been the increasing prOportion of the consumers dollar spent on meals away from the home (Figure 2u6). Over the period 1970-80, with | Ell..sal I'AIDIIHH L ”Univ. ZOOWN‘U ”I“? «soar-Poo 3'33"!” zomxm-o aim-u macaw-Foo r-mr'uzu 67 3800 3400 . ""DISP0398LE INCOME PER PERSON 3200 3000 2800 2800 2400 2200 2000 1800 . uaoo1 1400 I I 1 I T I I I I 7 I T I I j 80 52 54 58 58 80 82 84 88 88 70 72 74 78 78 80 82 YEAR 800 5m). 5m) 4mJJ 400 - . ................. .u‘“ an). ' 3m)- --ToraL F000 sxrrnouruas 260 . ----- rooo EXPENDITURE 91 “0 E ---Fooo exrsuourunss nwav FROM none 2mJJ uso _ .11... mo .. _____ _‘__,__-_-.—----— """"" " so I I I r T I I I I I I I I I I so 52 54 56 so so 62 a, as so 70 72 74 76 78 so 32 vans Figure 2.6 Income and Expenditures for Food at Home and Away From Home "‘4 6139] ”A A It! CCES purse. c .. Ho ex,€:u5u 15" ,. :gvl oust. 1. .1964 I wavy. . :AAA ~vhe . rm It.‘ 1 a5#;. v‘.‘ 68 the exceptions of 1974 and 1980, expenditures per person on food con- sumed away from home increased at a rate faster than expenditure on food for consumption at home (Gallo, 1981). As a result, expenditures per person on food consumed away from home increased its share of total food expenditures per person from 23.2 percent in 1970 to 26.5 percent in 1980. Some of the reasons for this result have already been mentioned; i.e., increased purchasing power in dual income households, greater value placed on leisure time and relative cheapness in feeding a smaller sized family. With an increasing number of consumers eating away from home the food service industry experienced, especially since the early 19603, a rapid growth in sales (Schmelzer, 1981). Much of this expansion was due to the rapid growth of firms in the fast food industry (i.e., McDonald‘s, Burger King, Hendy‘s, Kentucky Fried Chicken). Business census data show that between 1958 and 1978 real sales in fecd stores and eating places39 increased 44 percent and 83 percent, respectively (VanDress, 1980). Real sales in fast fecd places increased by 305 per- cent over this period. Moreover, fast fbod restaurants currently account for around 45 percent of the total number of eating places. The growth of the fast food industry in the late 1960‘s and 1970‘s brought about a rapid rise in consumption of ground beef and has been largely responsible for the generally strong demand for imported beef over the past 20 years. Most of the lean beef imported from Australia 39Eating places include restaurants, cafeterias and fast food Places which together account for about three-fifths of eXpendi- tures for food consumed away from home. Fast food outlets are the fastest growing segment of the away-from-home market. us ccusu u 1 cf a1. . 4- ' I $30.2 mud n ’ A. .59: x ’11qu N swim i I” “la: uu Abia' hush. n‘uupsih "has ‘ ARA w', JVV ti u" ’m» v.‘ 69 is consumed through the fast food industry. Commonly, 30 to 40 percent of all beef consumed in the U.S. is in the form of manufacturing quality beef (American Meat Institute, 1982, p. 17). Ground beef is the major component of manufacturing quality beef and has comprised regularly around 64 percent of all manufacturing quality beef. Ground beef is used to manufacture hamburger patties and is sold primarily through fast food outlets, although ground beef is sold also at retail fOod stores for meals prepared at home. A factor having a major influence on the share of ground beef and other processed beef products in total beef consumption is the beef pro- duction cycle. In addition to a weakening of demand for beef during 1980 through 1982, the current relatively low level of beef consumption per person is a partial reflection of the cattle industry being near the low point of its production cycle. Moreover, the prOportion of ground beef in total beef consumption is generally highest when ground beef prices are lowest relative to table beef prices, as occurred for example during 1975-77. These years coincided with peak beef cattle slaughter and beef production and a period of low producer prices. The combina- tion of an excess supply of beef cattle (a peak in the cattle cycle) and low producer prices, resulted in the liquidation of many breeding cows, reduced grain feeding of cattle in feed lots, and the related diversion of large numbers of cattle as non-fed. lower quality beef. With this influx of manufacturing quality beef, ground beef prices tend to fall relative to choice beef prices, and hence, ground beef consumption rises. ‘1 era..y I acusser b e‘. .c “”1 u “qu u its at: A . Alma ql S:.. ... ticn h ‘ PAW“. e Vb in i ‘ifi’a :R 'uiu '.. :e: du .g‘ .3?) . at: 'V 70 Satisfaction among consumers with ground beef has shown to be gen- erally high (Smith, 1978). It does, however, face competition for the consumer‘s dollar. There appears to be two, though not unrelated, lev- els to this competition for the consumer. For away-from-home consump- tion, ground beef predominantly used in hamburger-type sandwiches, faces its strongest competition from fast foods like chicken and fish sandwiches. Sun (1979) argues that much of the growth in fish consump- tion has been due to growth in fast food places, which have recently accounted fer some 60 percent of total fish consumption. It would seem, therefore, that a large part of the increased consumption of fish has been due to increased eating away from home plus an increased conveni- ence in fish products purchased fer home cooking. Hamburgers, per se, compete also with a number of products which contain, in varying degrees, ground beef, for example, tacos, pizzas. For home consumption, ground beef competes more directly with chicken (whole and parts), tur- key and beef and pork cuts of meat. The increasing number of salad bars in many fast food places is likely to compete with hamburgers and other meat sandwiches. However, at home salad vegetables are more likely to have a complementary relationship. A development in meat processing which may have contributed to the decline in hamburger meat and beef sausage/frankfurters consumption per person since the mid-1970's was the successful introduction of minced chicken and turkey in frankfurters. The inroads new processed chicken and turkey products have made into traditional domains of beef (and to a lesser extent pork) are reflected in the increased market share of poul- try franks, from 5.2 percent to 10.1 percent of the frankfurter market between 1978 and 1980 (Sun, 1982). Beef‘s share declined over this ‘, In! Cur.a6 I price 9. 337.539 113C '1' E'. u; 352'; u,.': \ "You 864., cy‘ ., 71 period from 38.8 percent to 28.9 percent of that market. Price changes during this time contributed to much of the shift in market shares; the price of beef franks rose on average by 5.8 percent while chicken frank prices fell by 2.2 percent. In fact, it was high beef prices, falling consumer demand and an attempt to product differentiate which led to innovations such as the textured vegetable protein blended hamburger during the mid-1970's (Weimer, 1976). This product failed primarily because of adverse consumer perceptions about blended meat products. Over the last few years the rate of growth in the fecd service industry and to a lesser extent in the fast food segment of that indus- try, has slowed appreciably (National Restaurant Association, 1980). This slower growth has been due mainly to the decline, in real terms, in expenditures per person on food for consumption away from home of 5.6 percent in 1980 compared to a 3.5 percent reduction in expenditures per person on food for home consumption. This is consistent with cross- sectional data that indicate that expenditures on meals away from home increase with increases in income per person. Therefore, reduced economic growth which reduces growth in real disposable incomes has had an adverse effect on sales in the fecd service industry. Interestingly, these periods of slower economic growth coincide with periods of reduced demand for imported beef. Hence, it would seem that sales of food for consumption away from home are affected before sales for home consump- tion are affecteduo. ___m 4 . 0Apart from the lagged relationship implied here, expenditure elasticities of demand for food eaten away from home, are likely to be greater than those elasticities for food eaten at home. '6 qt 31.21,.5 122A ‘4'. 72 There are other reasons to believe that the trend toward eating away from home may steady or even decline in the future. In recent years growth in the fast food industry has tended to plateau despite attempts to vary the menu offered. The apparent maturing of this indus- try'may reflect the aging of the baby boom group and its 'move‘ beyond the teenage years, years of greatest per person consumption of ground beef, chicken parts and processed chicken. Moreover, the present teenage cohort reflects the post-boom decline. Another factor slowing the growth in away—from-home meals is that the children of the baby boom are having families of their own. This, in many ways belated baby boom will tend to decrease the number of single and dual income earning households and can be expected to lead to more meals being eaten at home. Moreover, with the increased sale of food products with conveni- ence in home preparation, there will be less incentive to eat meals away from home. Fast food and other restaurants have already begun to move against some of these adverse trends. They have expanded menu selec- tion, increased promotion, changed services and redesigned their prem- ises with an aim to attract families and the older customer (VanDress, 1980). ‘1 Au. 8 s A A H i'gcns 59“ I.‘q 'AoAI: 0.0; I 5.5... ‘ u we. a . me my ."141 'Cuow :ettra .':'.e 1 \Pa.‘ ‘U'QI I?“ .1..‘ I'M i“- VI 5 CHAPTER 3 THEORETICAL FOUNDATIONS OF CONSUMER DEMAND AND APPROACHES TO THE ESTIMATION OF STRUCTURAL CHANGES The study of consumer behavior generally is undertaken within the neo-classical paradigm, the foundation of which is the theory of the utility-maximizing behavior of individuals. In empirical work the static theory of consumer demand often forms the basis for specifying market demand. In this section this theory will be reviewed briefly for the purpose of providing a basis for subsequent discussions of the iden- tification and measurement of‘structural changes in consumer demand. Also, since changes in consumer preferences, tastes and habits appear central to an understanding of the nature of structural changes, their role in demand theory will be discussed. 3.1 Traditional Theory1 The theory attempts to explain the problem of choices faced by an individual consumer with given prices, income, tastes, preferences and habits. The consumer will choose the most preferred combination of goods (and services) in such a manner that utility derived from their consumption or a combination of them is maximized. The behavior follows from the assumption of consumer rationality. Utility associated with the consumption of each good requires that the underlying preference 1Brown and Deaton (1972, p. 1148) consider that by 1939 most of the strengths and weakness of what we may call classical demand analysis had been probed and most of the techniques still in use had been discovered. 73 . , AD'AF.‘ re.:séb -“ «b I! ' ”3.3 user is CCSCE‘IE £3.75th I25": 'Als, '\ n ..l: I «Wang ‘.A V' . )5 7:: relationships of the individual satisfies certain behavioral actions (George and King, 1971, pp. 4-5). More specifically, an individual con- sumer is assumed to possess a twice differentiable continuous quasi- concave utility function defined over a finite commodity space: u - u(q1) i = 1,2,...,n (3.1) That is, a consumer makes a choice of quantities of various goods, q1, 92, . . . q“, from a goods space with n elements. This utility function is assumed to be maximized subject to a fixed or known budget or income constraint in each time period. The consumer‘s budget constraint is: n >3 no _<_y i=1,2,... 1-1 11 ,n (3.2) where unit prices of these goods are p1, p2, "'pn’ and y is the consumer‘s income or budget constraint.2’ This maximization process yields a set of demand functions of the fern: <11 ' q1(p1. y) i = 1.2..--.n (3.3) From these demand fUnctions a number of important relationships or Properties may be obtained on differentiation of the first order condi- tions (George and King, 1971, pp. 8-10). These properties which are in fact restrictions on the nature of the demand relationship, include the 2The inequality sign is a more general notation, however, for the present purposes an equality sign will suffice. p’cqbgflF ...vv.b- 1:3 'in o,“ I.'H~ IDba. mm,“ L”. I..“ 9.,‘u 75 homogeneity condition, the symmetry restriction and Engel aggregation or adding-up restrictions.3 The easiest way to impose these general res— trictions simultaneously is to derive the system of demand equations from a specified utility function. This will provide demand equations for which all general restrictions are automatically satisfied (Phlips, 19D“ p. 55). The cost of this approach, however, is a loss of general- ity' implied by the choice of the particular utility function. The alternative is to begin with the desired specification of the demand equation and impose the restrictions as constraints in the regressions. However, the cost in this case is the large computational burden requir- ing very sophisticated econometric methods. A particular advantage of imposing restrictions on consumer demand is that this reduces the number of parameters to be estimated. As George and King (1971, p. 21) pointed out, the number of parameters to be estimated independently still remains large when the system of equa- tions involves a large number of goods. However, as one is often interested in only a small subset of the consumer's consumption possi- bility set or even a single demand equation, this degrees-of-freedom problem may be further mitigated by the notion of separability. Separability provides a justification for imposing a restriction on the number of distinct commodities recognized. In essence, the aim is to include in the estimated demand functions price and quantity data and treat it as if it were price and quantity data of a single commodity 3Derivation of these restrictions and their implications for demand analysis are concisely discussed in George and King (1971) and in many texts on consumer demand theory, viz., Green (1978) and Phlips (1974), in particular. 72936 1 t' .I .25 To. ‘c JIN‘ Best I 3' I; a. ‘0 76 (Green 1978. p. 150-152). Hence, the assumption that the consumer possesses a separable utility function makes estimation of demand func- tions manageable. .In this research the concept of strong separability is defined as: £4 . o (3.4) where i 6 group g, j a group m, k a group s f g or m. This means that the marginal rate of substitution between two goods 1 and j, belonging to different groups, g and m is independent of the consumption of goods in any other group, 3. Inherent in the separability assumption is that consumers set aside or commit sums of money to broad general purposes, e.g., meat, dairy products, fruit and vegetables, clothing. housing, private transporta- tion, recreation, and so on. Then, it is assumed, that consumers decide at the appropriate time on the detailed disposition of these sums (Green, 1978. p. 153). In effect the consumer maximizes his utility in two stages. In the first stage of maximization, total expenditures are allocated between 3 separable groups and then in the second maximization stage, group expenditures are split into individual commodity expendi- tures (George and King, 1971, p. 27). In practice, it is not possible to look at marginal utilities to determine the nature of separability. Often arbitrary groupings are adopted. However, for the present pur- poses the research results of DeJanvry (1966, p. 112) are utilized, viz., that the meat category forms a separable group within the consumer‘s consumption possibility set. “u: in I '7 if! 1- ar. indiu' tiers au inevita': 3mm": dI-mubb .115 l CINE! 3"FAA 'fr-vfl 39"; “‘bus | u in" u‘ed 77 Thus far the discussion of demand theory and restrictions derived from it relate to an individual consumer. That is. the maximization of an individual‘s utility function is subject to a fixed budget constraint giving rise to a system of demand equations upon which various restric- tions are imposed. As a practical matter, as statistical data almost inevitably relate to groups of consumers, it is usual to aggregate demands across individual consumers to derive market demand functions. This in effect assumes that the demands refer to a 'representative con- sumer' such that aggregate demand relations may be obtained directly from the representative consumer demands. However, the conditions for doing so are very stringent and unlikely to be met in practice. For instance, the most restrictive condition is that all consumers‘ Bngel curves are parallel straight lines.u Nevertheless, in defense of this approach Houthakker and Taylor (1970, p. 200) argue that ‘of all the errors likely to be made in demand analysis, the aggregation error is the least troublesome.‘ Further, Grunfeld and Griliches (1960, p. 1) conclude that aggregation may reduce the specification error and hence produce an aggregation gain. There is no departure from this standard assumption of the representative consumer in the present research. A long standing problem in undertaking demand analysis lies in bridging the gap between the theory and empirical work. The gap exists primarily because of restrictive assumptions used in the theory and because of inadequacies in the data base. Some of these difficulties such as identifying separable groups and the aggregation of individual uFor details of the implications of aggregation see Green (1978, P- 139ff) and Phlips (1974, pp. 98-100). em r reitera: is a me rrcr~ wu’u'hb 95 Aha. A "1" V “Mos Iovnvi A: ‘1‘; '4 Vin. A.,_~ ‘Id. ‘T. :1 78 demand relationships have been mentioned above. Others, which are not reiterated here are discussed by George and King (1971, p. 20ff). .There is a need, however, for fUrther coverage of certain aspects of consumer theory which relate closely to the nature and modelling of structural changes in demand. These include the common neo-classical assumption that consumer preferences are given, the theoretical treatment of commo- dities or goods as opposed to characteristics of goods, and extensions of the classical static theory into dynamic models of consumer behavior and habit formation. 3/L1 Consumer Preferences and Goods Characteristics In the above outline of traditional demand theory, underlying con- sumer preferences were assumed to be given fer a particular period in time. Although this assumption greatly simplifies the classical treat- ment of consumer behavior, it is rather unrealistic to expect prefer— ences of individual consumers or households to remain unchanged over time or unaffected by their environment. When preferences are allowed to change, the demand system also changes. This is because the indivi- dual preferences underlie the given utility function and its properties which in turn determine the form of’the demand functions. It is not the intention here to set forth a theory which considers changing preferences or which permits maximization of an intertemporal utility function. This has been attempted elsewhere.5 The objective is Pimply to provide insights into how changes in preferences might effect 5A theory of intertemporal utility’ functions is presented by Phlips (1974, Ch. 10). s m n is (I. n) 1 '. 4‘ s..Clau I Jonah?” Mica“: tinst m’ “In to Bun. iv a. h. 'v. 79 demand and what might cause such changes. Moreover, in empirical ana- lyses and estimation of individual or system demand parameters one should be aware of the implications of the assumption of an unchanging utility function over the observation period. There is in fact no unique time period for which the utility func- tion should be defined but these are restrictions on the possible length of the period. Since the consumer derives utility from variety in his diet and diversification among commodities he consumes, the utility function must not be defined for a period so short that the desire for variety cannot be satisfied or too long a period that tastes (and hence, the shape of the function) may change. The (static) theory would break- down if it were impossible to define a period that is neither too short nor too long in the above view. In essence preferences are given. Were it practicable, it may be very important to incorporate into the analysis the fact that preference ordering may change over time (or from household to household)6. Green (1978. p. 26ff) outlines three possible reasons preferences change:7 namely, the effect of advertising, the choices made by other consumers, and the longer term impact of changes in price. Effect of advertisingi Essentially, a consumer's choices are deter- mined by his preferences and those preferences are influenced by the 6 These changes should not be confused with shifts in text book demand curves constructed under ceteris paribus conditions. 7For a discussion of the current state of modelling changes in consumer preferences see Pollak (1978). Pessemier (1978) and Mar- schak (1978). in’nfi‘" .u.vu H ' 2C3}: n 55‘ r11“ 'yv-b ads-- I files c~. RIMS duh-Tn In the sizes uii‘: re MI- "Ov. 0“. 1.": ”is: Pug‘ 4, i1 .Vo 80 information which is available to him. Lancaster (1966) developed the notion that preferences among bundles of goods are not a fundamental determinant of choice, but are derived from preferences among the bun- dles of characteristics supplied by these goods. This notion may be at the heart of the perceived change in preferences for beef over other meats and foOds. Normally advertising is concerned with persuading con- sumers to switch from one brand to another brand of a particular good.8 In the case of meat demand, the concept may be more complex since recog- nized brands do not exist in the market. That is, meat is homogeneous with respect to brands, although there are many cuts and types of beef meats. Adhertising in meats, where it has taken place, has concentrated on ‘commodities‘ e.g., beef versus chicken versus pork. If consumers' preference orderings are based, not on commodities but on goods charac- teristics, then this advertising effort may have been misdirected. If it is possible to interpret each type of meat, e.g., beef, chicken, pork or cut of meat, as goods which supply broadly the same characteristics but in different proportions (i.e., vitamins, minerals, leaness, calories and taste), then a large part of advertising may be construed as an attempt to inform people of the bundle of individual characteris- tics of a given good. This begs the question regarding the cause of the recent decline in demand for meat, red meat and beef, in increasing cmder of magnitude of that decline. Does this reflect a change in M 8Galbraith (1979) stressed the ability of producers to manipulate consumers through advertising to the point where an individual surrenders her consumer sovereignty to the producer or producing organization. See also Galbraith (1973) and Gintis (1972) for discussions of the broader issue of produced influence of corpora- tions and loss of consumer sovereignty. 4 r r’ .XSUEJU «bu-H 051459 y: 9w. ' 'm' .4) ' '5 _ . I": ‘5‘. “hi. 1"". In ‘ .j: I h...- 15‘: p, 'I'A rm _ 81 preferences or a change in information? That is, it may be quite con- sistent for a consumer, who has a particular preference ordering for particular characteristics of a good, to shift from beef to chicken because of new or changed information that chicken provides the bundle of characteristics which maximize his satisfaction in consumption. If the unit of observation were goods then this shift may be perceived as a change in preferences, namely of the good, beef. However, by looking at characteristics, preferences are unchanged, so then the determinant of choice in this case is information. Take, for example, a consumer with a given preference ordering for the characteristic of 'health attributes' who is at that time consuming beef. Assume a report is released, stating that beef is high in cholesterol and dangerous to ones health, and that chicken, because of its low cholesterol level, should be consumed. In response, that consu- mer, assuming his rationality, will shift consumption to chicken, where his given preference ordering with respect to characteristics can be satisfied. No change in underlying preferences has occurred; simply the product, beef, in the eyes of this consumer, no longer contains the bun- dle cfi‘characteristics which reflects his preference ordering and which maximize his utility. Characteristics, therefore, would seem to be the focal point of advertising, namely, to inform peOple of the characteris- tics or the bundle of characteristics that a consumer will receive on purchasing a particular commodity (or brand).9 9This is a point that until very recently has been overlooked in the promotion and advertising dollars spent by representatives of the U.S. beef cattle industry, in their attempts to turn around or at best halt the decline in per capita beef consumption (Beef In- dustry Council, 1982). a. and t ’. new} My” 4‘ v ‘46.... an-‘r4- Ui". ‘I “P 0. yet: 82 This point is taken further by Lancaster (1966, pp. 132-133). He argues that despite the persistence of advocates of the pure theory, market researchers, advertisers and manufacturers act as though they believe that knowledge of the intrinsic properties of goods is relevant to the reaction and behavior of consumers. He notes that the classical theory does not allow fer the introduction of a new good or for the quality of an existing good to change. The theoretical, and indeed the empirical problem arises when such structural changes occur within the last few years of the observation period. In this regard traditional theory would appear to have nothing to say and hence have no predictive value. If it were possible to measure the quantities of characteristics supplied by new or improved goods then a cost of living index could be defined in terms of a representative bundle of characteristics. Such a revision of demand theory may substantially overcome present weaknesses in the classical theory, such as the assumption of fixed preferences. Preferences relating to characteristics change less over time than preferences relating to goods which supply these characteristics in dif- ferent proportions. The apparent difficulty of inventing or promoting a completely new characteristic would support this assumption. The point is that under the neo-classical paradigm the only characteristic of beef is 'beefiness' and the only source of 'beefiness' is beef. Choices by other consumers: Besides advertising, preferences may also change through the choices made by other consumers. When the choices made by others result in an increase in a consumer's purchases it is called the 'bandwagon effect'. When the result is a decrease in a consumer's purchases it is a 'snob effect'. Green (1978, p. 28 and pp. 1H6-1fl8) states that these effects should be incorporated into the tr. ir.‘ 1.5V”; 1‘- .m. ihhrc up, ‘ 83 theory of individual behavior when applied to market behavior. It is unclear which effect would predominate in the market fOr meats. Tb some extent a bandwagon effect has resulted in the trend away from red meats, especially beef, due to alleged health considerations. Estimators of demand fUnctions appear to recognize these factors as a source of preference change, but typically allow for this only in a very general way. Approaches to measure such structural shifts are detailed later. Green (1978, p. 147-1u8) presents a theoretical model of these effects on individual demand. He assumes all individuals have identical price dependent Marshallian demand curves of the form: pi = a + bq1 +‘cQ1 (3.5) where the demand price, pi depends on the quantity purchased by the individual consumer, Q1, and on his expectation of the total amount, Q , purchased by that group of consumers. It is assumed that a > o, and b < 0, giving a downward sloping demand curve in the case of consumer independence. Hence, the demand price may be increased in response to an increase in expected total purchases, where c > o for the bandwagon effect, or decreased, where c < o fOr the snob effect. When present the snob effect, c < o, will make the demand curve slope downwards more steeply, whereas a strong bandwagon effect, c)o may turn the slope to positive. Changes in_price: Price is a factor which may induce a change in preferences. Price changes clearly affect choices through their effect on the set of alternative goods that the consumer can afford within his given budget. The question is whether such a commodity price change stage I {and ‘5de. b .can nap‘km yC. oQUI “a“? 4 b5.. ‘ “In?“ ' "VI rf‘. I'M" «r. 'h;a “ii. I H..- ‘v. ‘5 fig.. -.4: 8H affects a consumer's preferences among alternative bundles. Some would argue that the high retail beef prices of the late 1970's led to a change in preferences for beef in favor of chicken or pork or non-meat food. However, the logic of this argument seems unclear. When beef prices rose to levels unacceptable to consumers, there followed the expected contraction in purchases (a movement up the beef demand curve) causing the demand curve fer chicken to shift to the right. This leads to an increased consumption of chicken at the same price, ceteris paribus. If consumers then deve10ped new habits and tastes from the greater chicken consumption then overall preferences may have changed in favor of chicken. Nonetheless, the shift to chicken fellowing the high beef prices should not suggest a change in underlying preferences for beef at that point. Simply, the consumer could satisfy with chicken his preferred ordering of his characteristic set (in terms of calories, vitamins, minerals and so on), more cheaply and within his budgetary limitations. Preferences only changed after new habits and tastes were fbrmed through greater consumption of chicken, ceteris paribus. This is reflected in his reluctance to increase equivalently his consumption of beef when beef prices subsequently fell. This apparent asymmetry in demand response will be investigated below. 3.1.2 Dynamic Models and Habit Formation As noted above, the static neo-classical theory of consumer demand is based on a number of restrictive assumptions. In particular, the theory implies that the representative consumer solves his maximization problem in each period regardless of any past behavior. There have been many attempts to relax this assumption and specify dynamic models of behaviCI behevic in the fmctic .' .1513. invclv Wile .. zzrst 85 behavior by accounting fOr habit fbrmation and partial adjustment 10 The approaches usually adopted are either to date variables behavior. in the utility function, introduce dynamic aspects into the utility function or to introduce dynamic aspects directly into the demand func- tion. Only the last approach is considered here. Other than specifying a time trend, dynamic demand analysis usually involves lagged variables. Examples of direct inclusion of dynamic pro- perties into demand are the cob-web model, the distributed lag model, a first differences model of changes rather than quantity demanded, and the habit formation or inertia model of consumer demand (Intriligator, 1978, pp. 235-2u2). The three main habit formation models are (1) the inclusion of an exogenously determined time trend to the static model to account for taste changes; (2) the partial adjustment model, an endogenous specification which assumes that adjustment of actual con- sumption to desired consumption is achieved only partially during any given time period because of habits; and (3) the state adjustment model. The first approach is relatively straightforward. The last two are dis- cussed below. Of interest in this research, in addition to an accurate fitting of a model to observed data, is the development of a model that assists in identifying and measuring changes in the structure of those demand relationships. The partial adjustment mechanism as fOrmulated by Nerlove (1958) is an example of a distributed lag relationship. It assumed for prices and incomes, that consumers adjust consumption towards a desired or long-run 10 See, for example, Phlips (197k) and Green, Hassan and Johnson (1978. pp. 93-107). h...“ 'Mun ”Pug 1.. _‘~ 5! t -86.. demand and in each period only partial adjustment to the desired demand level is achieved. Therefore, if (1’ ) (3.6) * represents the desired demand for commodity j in period t, a Nerlovian partial adjustment scheme would result in * qt - qt_1 ‘ 5(qt - qt_1) o< 6< 1 (3,7) where represents the 'coefficient of partial adjustment.’ When €=l the adjustment is total and immediate and gives the original static case. Expressing (3.6) in a linear form with an additive error term and com- bining this with (3.7) gives the estimating equation in observable terms, qt . 6a + dbpt + cScyt + (1-6)qt_1 + Out (3.8) or equivalently, qt " A + Bpt + Cyt + Dqt_1 + Out (3.9) When estimated in double logarithmic form, <5 is the elasticity of adjustment, B and C represent the long-run price and income coeffi- cients, respectively, and b and c represent the short-run effects. Con- ceptually this model reflects the impediment to immediate consumer responses due to imperfect knowledge. As price changes, some consumers will continue to respond to past prices because of habits or institu- tional constraints. Although widely used in empirical analyses, this Partial adjustment model has no solid theoretical underpinning (Phlips ’e'ent 35.33 are i.e., .ll‘ m... the c _ 37 _ 197“, p. 15“). In fact Griliches (1967) showed that essentially the same form of estimating equation, indistinguishable from the static demand model with autocorrelated errors, can be derived from quite dif- ferent constructs. An improvement on the partial adjustment model is provided by Houthakker and Taylor (1970). Their state (stock) adjustment model, in essence, is a habit fermation model11 typically applied to non-durable goods. The basic equation is qt = a + Bpt + Yyt + oat + u (3.10) t where 3t represents a psychological stock of habits and 4 NI NI... Figure 3.3 An Illustration of Joined Segments of Different Polynomials 30More complex forms of these switching regressions are provided by’Goldfeld and Quandt (1973). 31This review will draw extensively on the research of Poirier (1976) among others. disccn alcng ‘. .135. as ra: siste tural 1119? - 114 - Cases of the apprOpriate uses of the standard dummy variable, i.e., discontinuous changes in the dependent variable, have been discussed along with the various shortcomings in their use in econometric estima- tion. Poirier argues, however, that with the exception of inherently discontinuous models of structural change of a qualitative nature such as race, region or education, continuity in economic models is more con- sistent with reality. Where the explanatory variable causing the struc- tural change is quantitative and continuous, then at the very least an interaction dummy should be used. A point to note about spline functions is that while they are con- tinuous at points of structural change they may not be smoothly continu- ous. Segments of the curve representing periods between significant points of structural change are connected by join points. These seg- ments may represent splices of linear, quadratic, cubic, exponential or any other functional form. It is possible to represent these joined segments by a series of slope and intercept dummy variables subject to the constraint that segments join up continuously. However, spline functions satisfy this continuity without the need of such constraints. In the following, the discussion begins with the simplest form, the linear spline, moving to cubic splines, bilinear splines, splines in multiple regressions, and finally leading up to continuous time varying parameter models of structural change. " 1' 1.1193. here and 1 The line the 913.“. - 115 - Linear Splines: The linear spline may be shown for k transformed variables as qt " Bo 4' Blwl + 82112 + + Bka (3.23) where, wi'xt Wj - (xt - x14), for xt>xj_1 - 0 , for Xt5XJ_1 and for j-i join points, j=2, 3,...,k and for time, t=1, 2, 3,...,n. The coefficient. 51, represents the slope of the spline over the first line segment. The remaining coefficients, Bt' represent the change in the slope from interval j-1 to interval j, respectively. Therefore, for example, the slope of the second line segment is 81 +.82. The attractions of this linear version are (a) its linearity in unknown regression parameters, 80, 81...,Bk; (b) the absence of parame- ter constraints: (c) the ease in constructing the transformed variables, "1. W2....,Wk in standard regression programs: and (d) the ease of hypothesis testing of the presence of structural change, using standard student t-tests on each 8. Statistical significance, different to zero, indicates a change in the slope between intervals j-1 and j and hence the notion of structural change at ij_1. Although omitted from the above formulation, it is a straightforward matter to include other explanatory variables within the context of multiple regression. Note, that and (join) points can be specified if it is believed that the data set is numerically bounded, for example, as in the case of income dis- tribution segments which may have a lower bound of zero. (uh? 4" f m”. the rr 1 Ir prcvf 1h€3 .4. nce the seg: - 116 - Cubic Splines: If the original relationship is non-linear and is a cubic polynomi- nal then Poirier (1976, p. 21) shows that a structural change at the join points will yield a cubic spline. In this case the transition at the join point from one polynomial segment to another does not occur abruptly since the curve's first and second derivates are matched. This provides a very flexible approximation to the data. More generally, these are in essence higher order grafted polynomials in a single independent variable joined smoothly at known join points. Feasibly, the first line segment could be a cubic polynomial grafted to a linear segment (Fuller, 1969) as illustrated in Figure 3.3(b). The case of a quadratic form over two line segments, i.e., one join point, and hence one structural change, may be shown as qt ' o + Biwi‘+ B2W2 + B3W3 (3'29) where, w1 ' x: 2 w2 x: w a (x - ‘ )2 for x > i 3 t xl ’ t l ' 0 , for Xt.§ X1 In this case, a t-test on the coefficient, 33 will indicate immediately the gain from adding this structural shift parameter to the model. Variations of these and other degree polynomials may be more appropriate in practice (Cornell, 1976). the c hie, hence 1cm ether expec hall the ante -117- Judge et al. (1980, p. 388) suggest that a major shortcoming with the cubic spline, given its specification in a single independent vari- able, is the implied restriction of the form of structural change. Hence, there is limited scope for identifying the nature of the change. However, Poirier (1976) is clear in indicating that the addition of other explanatory variables is not only a simple matter but to be expected in any formulation of a spline function. Of course, a poten- tially limiting aspect of spline functions discussed so far is that the time of the structural change, i.e., the join points, must be known, ex ante. Bilinear Splines: A bilinear spline is obtained by creating an interaction variable with transformed spline variables. With subscripts omitted for simpli- city of exposition, a relationship between 9, the dependent variable and two independent variables X and Y gives the bilinear form q:a+bX+cY+dXY (3.30) where the main effects are measured by bx and cY and the interaction effect is dXY. Application of this model to a piecewise formulation implies that the XY term operates only over portions or certain regimes of the XY space. Where these regimes are continuous then the resulting E'EEZ'E -118- surface is a bilinear spline. In a notation consistent with linear and cubic splines, the bilinear spline may be represented as qt ' 8o + Blwl + Biwz + ijs (3°31) where, w1 ‘ x Xi-l V t - X1_1, for Xt - 0 , for Xt IA Xi-l i = 2’ 3’ .00, I Y w . Y 3-1 [A t - Yj-l’ for Yt = o , for Yt s.Yj_1 j = 2, 3, ..., J w3 ’ wlwz where, i has i-l join points, Y has j-l join points and time is t = 1, 2, ..., n. An alternative formulation of the above equation is qt ’ aij + bij(xt ’ Xi-l) + cij(Yt ' Yj-l) + dij(xt ‘ xi-1)(Yt ‘ Yj-l) (3.32) where X0 < X1, and Y0 < Y1 and 313 is the value of qt when Xi = xi_1 and ii = YJ_1, bij is the partial of qt with respect to It, 613 is the partial of qt with respect to Yt' and did is the interaction coefficient between Xi and Y . Because of the richness of the kinds of structural change which can be specified, the complexity of the hypothesis testing of these changes is increased though not prohibitive. However, since 119 this spline is linear in its unknown parameters, classical testing procedures can be utilized (Poirier, 1976, pp. 62-6“). Bilinear splines have the advantages over dummy variables which continuity gives and they are less costly in degrees of free- dom. However, higher order forms of this spline are increasingly more expensive in degrees of freedom and meaningful economic interpretation of the derived coefficients is more difficult. Nevertheless, combinations of linear and cubic splines are feasi- ble. Two final points to note are that (a) spline functions may include other explanatory variables, splines with other variables in the equation and interactions with splines, and (b) in specify- ing those functions care must be taken to ensure multicollinearity is not introduced since each of these splines have intercepts ' implicitly or explicitly contained in them. Unknown Join Points: As noted above, a limitation of spline functions and indeed any switching regression whether it uses dummy variables or piece- wise regression, is that the join point or exact time the struc- tural change occurred is often unknown. Models where the join POint(s) is unknown and hence needs to be estimated are contained in the literature, for example, Poirier (1978, Ch. 8): Judge et al. (1980), Silvestre (1969). Peirier argues that spline regres- sions have major advantages in practical implementation, especially when other explanatory variables are added, when more than two regimes or segments are involved, and when the disturbance term is 120 not well behaved. Nevertheless, these models are computationally much more complex and may suffer from the requirement of large sam- ples or large shifts in order to adequately test the model for sta- bility of slope coefficients and of the estimated join points. However, it is generally uncommon that some a priori informa- tion is not available concerning the location of the structural change. If the point in time when the change(s) took place is only approximately known then one approach is to choose several Join points and/or construct several functions and then select the one or combination of Join points and functions which give the smallest residual sum of squares.32 The risk with this latter approach in finding join points is that as the function becomes complex, local minima instead of the global minimum may be located.33 Also, unless the appropriate optimization techniques3n are employed, iterative algorithms can be expensive to run when the number of iterations is large. —___ 32A rather different approach involving the criterion of minimiz- ing residual sum of squares was utilized by Silvestre (1969). Although this method contained some novel aspects, it appears com- putationally prohibitive for all but the simplest mmdels of struc- tural change. Also, more efficient algorithms have been developed which largely supercede this approach. 33Hathematical algorithms which search for the global minimum tend to be iterative rather then analytical in nature. See Hold‘s four rules—of-thumb to find join points in cubic splines, quoted by Poirier (1978. pp. 151—152). 3H3” for exanple Kuester and Mize U973). 121 3.“.2 Models of Quality Change In the previous section a consumer theory which viewed a pro- duct as a collection of characteristics was discussed. This theory, due to Lancaster (1966), has been applied in various food consumption and demand studies. For example, Ladd and Suvannunt (1976) tested the hypotheses that (a) for each product consumed, the price paid by the consumer equalled the sum of the marginal monetary values of the product's characteristics and (b) consumer demand functions for goods are affected by characteristics of the goods. Products derive an inherent heterogeneity because of the array of various characteristics contained in them. Different qualities and quantities of characteristics provide this hetero- geneity and hence models of consumer goods characteristics are use- ful in studies of product differentiation, quality, and grades and standards. As noted earlier, traditional consumer demand theory has not viewed utility maximization in terms of product characteristics, nor does it have anything to say about the effect of quality changes and the introduction of new products in demand analyses. Hohlgenant (1982), however, is an example of one attempt to address this problem in the context of the demand for meats. He hypothesized that unexplained structural shifts in demand fbr meats can be attributed to quality changes in the composition of meats consumed. Rather than presume that negative shifts in the demand fbr red meat have been due to changing tastes and attitudes to health, he hypothesized that this shift is due to substitution of 122 new processed poultry meats for processed red meats. The basis used fbr identifying structural shifts due to quality changes is that final consumption zJ, of each good, J, is fermulated as a function Of quantity, Q3, and quality, 33' of the market good pur- chased. For example, quantity is in terms of pounds of'meat con- sumed, while quality may be the number of grains of protein per pound of meat or some other quality characteristic such that zj-qj sj’ jglf 2’ "'!n (3033) Since it is not possible to characterize quality by a single attribute, s3 is viewed as an index of quality, depending upon a whole set of attributes, s .., amdj) (3.34) j = a3 (alj, 823’ . for all j = 1,2,...,n and where an is the amount of the ith characteristic yielded by one unit of the Jth market good. These attributes might be protein, enerSY. carbohydrates, iron or any other nutritional, physical or if measurable, psychological attri- bute. Demand for each purchased good is derived as qj a qj (p19 p29 °“9 pn! xj;319 829 "'3 Sn) (3.35) fbr J = 1,2...,n where the s 's are given for each product, and pJ i is the price of the jth good. These are constant-quality demand functions. An important restriction of this model specification is that quality changes are interpreted as movements along a stable demand curve through changes in the shadow prices, 'WJ = pj/SJ fior 123 the Jth good. Hence, an increase in sJ decreases NJ causing zj to increase. Wohlgenant showed after netting out price effects between 1970 and 1980 that about one-half of the unexplained increase in poultry demand, and one-third of the unexplained decrease in demand for red meats (beef and pork) was due to quality changes. A significant shortcoming in this model relates to the invariant nature of the elasticity estimates. These results will be further referred to in later discussion of model estimation. 3.4.3 Time Varying Parameter Models The notion of demand estimation that accommodates time varying coefficients or elasticities is not new in the literature. Apart from the early development of the intercept shift and slope dummy variables this broader concept of time varying parameters was dis- cussed in the context of 'time-elasticity of demand' (Smith 1937; Prest 19n9, p. “7). Waugh (196”, p. 8) observed an excessive con- cern with 'the' elasticity of demand for a commodity and felt that in reality demand elasticities commonly vary across markets, across uses, across grades and especially across time and from one part of the curve to another. Already noted is the absurdity of assuming a constant income elasticity of demand. It may be misleading to com- pute a single coefficient that purports to show the response of food consumption to income changes either across sections of time or through time. Foote (1958, p. 83) outlined some of the diffi- culties of interpretation and comparison of estimates of demand 12a elasticities from time series data. In particular he pointed out the considerable variation in elasticity estimates obtained from regression equations run over different time periods and from data other than the average of the period; contrasting years of depres- sion and prosperity. As pointed out by Foote, it seems clear that since an analysis of the cost of public programs depends largely upon the demand elasticity of the commodity, the accuracy and vali- dity of estimated regression coefficients in terms of the time over which that coefficient is relevant bear upon the overall quality and effectiveness of government programs. In the previous section, techniques were discussed which per- mitted the regression coefficients to change a small number of times and therefbre allowed fbr a similar number of changes in structure over that time series. In this section models are reviewed which permit the coefficients to vary more continuously over time. As evidenced by the plethora of approaches and techniques to identify and measure changes in the underlying structure of economic relationships the importance of modelling parameter varia- tion is obvious. A major empirical question posed through this discussion is whether structural changes are inherently abrupt and instantaneous as portrayed by duumy variables, whether they are more gradual but nevertheless definitively identifiable over a small period of time after which a constant parameter prevails as might be modelled with spline functions, or whether in fact, struc- tural changes occur more or less continuously over time. While 125 economists using all three general approaches recognize the possi- bility of parameter instability, a required ingredient of the first two approaches is for most practical purposes, that of knowledge of when the change in structure occurred. Notwithstanding this requirement, a necessary caveat in each case is that while structural changes may be identified from a sta- tistical viewpoint, these techniques provide only limited economic interpretation of the change. Thus, the difficulty of pinpointing the nature of the parameter variation has resulted in the continued use of various constant parameter formulations. The merits of the simpler constant parameter models are well known. However, where there is evidence of instability in the parameters there would appear to be a trade-off between accuracy and complexity. Mundlak and Raussar (1979, pp. "-9) offer a number of justifi- cations for the parameter variation fbrmulation: (a) the 'true' coefficients are considered to be generated by a non-stationary or time-varying process; (b) even when the 'true' coefficients and the underlying structure are stable (since econometric models are sim- plified abstractions of reality), misspecification may cause inac- curate model forecasts, which may be countered by an appropriate parameter variation structure. An important misspecification is the omission of relevant explanatory variables particularly those that relate to structural changes resulting from taste and institu- tional changes or technological developments. Time series of these excluded variables exhibit nonstationary behavior and due to their correlation with included variables, the estimated effects of the 126 latter can‘ be expected to change with time. This time-varying interrelationship can be expected between proxy variables and the desired variable, for example, expectation fermulations, which they proxy. That is, if the relationship between the proxy and its true counterpart is not constant across observations, then the coeffi- cient of the proxy variable will not be constant; (c) an underlying assumption in the construction of aggregate data is that the rela- tive weights of the constituent micro-units do not vary. As the relationship between micro-units is unlikely to remain constant in time series data, coefficients in the aggregate equation will vary; (d) an inappropriate specification of the functional form espe- cially when a linear equation is fitted to an inherently non-linear relationship, can cause coefficients to vary across the sample. This is particularly so for time series displaying pronounced secu- larity; (e) perhaps the most intuitively appealing Justification of a parameter variation structure is that economic relationships can be expected to change over time, at least because of changes in policy which affect the operating environment and hence the behavior of its participants; (f) the evidence that economic rela- tionships are fbrmed by dynamic processes fueled by expectations of futuref structural changes would indicate the presence of parameter instability over time. Certainly forecasts would be improved if account was taken of such parameter variation. On the basis of the above observations it would appear unreal- istic to assume that (a) the economic structure generating the sam- ple observations remains constant, (b) there exists a single param- eter vector relating the dependent and independent variables, (c) 127 there exists a constant set of error process parameters, or (d) there exists a single functional form. However, as Mundlak and Raussar (1979) point out, the central issue is whether or not the explicit specification of a time-varying parameter structure will sufficiently improve accuracy and implementation to out weigh the increased complexities of such a specification. In recent years a considerably large literature has developed along with numerous variants of the continuous time-varying parame- ter model. First, as a more obvious development of the classical linear model, the general model that combines time series and cross-sectional data is reviewed. Then, as a transitional link from previously discussed models to models to be discussed, Poirier's (1976) switching regression version of a time-varying parameter model is presented. Finally, for review purposes, a gen- eral version of the latter model and an important subset of more specific formulations will be presented.35 3.n.n Cross-Sectional Time Series Model In the context of the classical linear model of constant coef- ficients using time series observations, perhaps a logical develop- ment of this model is to estimate a relationship which combines time series and cross-sectional data. Typically, several years of data may be available for a number of households. This estimating approach is not used in this research because of the lack of *4 35This review draws, inter alia, extensively upon Judge et al. (\980) and Mundlak and Raussar (1979). 128 consistent cross-sectional and time series data. However, for com- pleteness and since this technique does allow structural changes in demand to be identified by permitting the assumption of fixed coef- ficients to be relaxed, the approach is discussed briefly. In general, the model may be written K = B + Z ‘11: 11c Bkit xkit + 9’1: (3°36) k=2 where i a 1,2,...,N refers to the cross-sectional unit, household or individual, and t = 1,2,...,T refers to a given time period. The value of the dependent variable in qit fbr individual i at time t and the value of the kth non-stochastic explanatory variable is int for individual i at time t. The stochastic term is assumed to have the usual prOperties of mean zero and constant variance. The unknown parameters or response coefficients, Bkit may vary for different individuals and for different time periods. This general model may be classified, as the subscripts suggest, into various models ranging from the most restrictive case where all coeffi- cients are constant, through to other combinations depending on whether the intercept or slope coefficients are constant or vari- able, whether the intercept and/or slope varies over individuals and/or over time. These and further sub-categories are detailed by Judge at al. (1980). In terms of time varying parameter models the general model where all coefficients vary over time and individuals is the most comprehensive. In practice, however, more restrictive assumptions are usually made. 129 To illustrate the type of structural infbrmation that can be obtained from such a model some of the results of a study by Doug- las (1967) of the secular and cyclical changes in consumer demand for household laundry appliances is briefly reviewed. Time varying parameter models which rely wholly on time series data are presented in the following sections. It will then become clearer Just how closely related are the time varying parameter and the pooling models. Douglas examined changes in income elasticity over time. Elasticity coefficients were obtained for each household over the period 1932-61 (excluding 19u2.u5). She showed that the newer the product the greater the elasticity, which rose when the new product was introduced, reached a peak then declined secularly as the pro- duct experienced increased market maturity, and became negative fer a declining product. Another important result with respect to the present study is that the price elasticity of demand changed over time in response to, first, the maturity of the market and, secondly, in response to competition. For instance, price elasti- city was quite high during the years immediately following intro- duction of the electric and gas washing machines but showed a marked secular decline during subsequent years' (p. 73). She con- cluded that price elasticity declined as washing machines experi- enced increased market maturity and that the price elasticity for total washers, automatic washers, and dryers during the post-war years was probably quite low. From this evidence Douglas surmises that income and maybe price elasticity may be an effective basis for differentiating between luxuries and necessities. Two 130 advantages given in having these elasticity measures through time are (a) as an aid 'in identifying where the position of a given product is at a given moment of time on the continuum between the possible extremes of luxury and necessity and particularly the par- ticular products position relative to others' and (b) that 'changes in elasticity from year to year may be useful in tracking a product's movement through various levels of market maturity' (p. 7”). Since consistent time series and cross-sectional data are not available for the consumption of food, least of all meat, in the 0.8., alternative approaches to obtaining intertemporal change in elasticity estimates must be sought. Also, it is recognized that the demand fer a durable consumption good will have substantially different characteristics to the demand for meat. Nonetheless, the Douglas study is aprOpos to the discussion of time-varying parame- ter estimation and does motivate the search for such a technique in explaining such structural changes in meat demand. 3.4.5 Time Varying Switching Regression Model Poirier (1976) has argued with considerable Justification that the switching regression model more appropriately and realistically reflects structural changes in a model through the switch in param- eter values from one regime to another. Certainly there is more structure built into the switching regression model in the way parameter values change than in the fixed coefficient or in the random coefficient regression model. 131 Spline fUnctions which permit parameters to change over some time periods and remain constant over others have certain attrac- tions. The model presented here is an extension of earlier spline functions except that all slope coefficients are allowed to exhibit continuous time varying behavior. Consider the simplified model, excluding other explanatory variables, qt=a+6txt+et, 133192, 00-91. (3.37) which is time varying in the slope coefficient at (but with a con- stant intercept<1). Assume that 8t is a function of time such that at a Y0 + ylt + 72 (t - t1) . (3.38) Combining both equations, assuming the known Join point t1, gives qt - a‘+ yoxt + y1(xtt) + y2[xt(t - t1)] + at, (3.39) or equivalently, qt - a + {Y0 + Ylt + 72 (t - t1)] xt‘+ at (3.40) The following null hypothesis may be tested: (1) H0: Y0 3 Y1 = Y2 = O . (11) no: Y1 = Y2 = 0 '(iii) H0: '72 = 0 (iv) H0: Y1 + Y2 = 0 where null hypothesis (1) tests for any effect of the regressor xt on Qt: (ii) tests whether the slope is constant over time; (iii) ‘— 132 tests whether the slope varies linearly over time; (iv) tests whether the slope is constant fer t.; t1, but varies linearly for t s t,. Various extensions can be made to this simplified linear spline fbrmulation. These are, among others (i) parameter varia- tion according to a higher order spline, e.g., cubic spline; (ii) addition of Join points to accommodate other points of structural change in the timegpaths of the coefficients; and (iii) time varia- tion for more than one slope coefficient, i.e., the addition of other explanatory variables. 3.“.6 Continuous Time VaryingLParameter Models Consider, in its simplest ferm, the single equation, one explanatory variable model36 y 3 xtBt + u (3.41) t t where Vt is a vector of observations on n dependent variables; Xt is a matrix of n observations on k predetermined variables; Bt' which is allowed to vary over time, is a vector of parameters; and t denotes the time period. The error term. ut is a random vector, serially uncorrelated and distributed with zero mean and covariance 2 matrix, E(ut “t1) = 5 u, and E(ut Xt) = E(ut St) = 0. 36This time series model could be generalized to include both cross-section and time series observations. See, for example, Judge et al. (1980, p. 382) 133 Changes in Bt may be caused by structural variables, or by completely random effects as represented by the error term et in the equation 8 = 80 + H(Xt) + zta + e t (3.42) t where H(Xt) denotes the general function of variables appearing in the system which may be associated with habit formation; Zt represents completely exogenous variables and the error term has 1 2 the usual properties of E(et at) : Oe' and E(et 2t) = E(et H(Xt)) = 0. In demand equations it is not uncommon for B t to depend on past consumption thus specifying habit fbrmation, dependent upon past experience. ,Indeed, in the consumption of meat, it is hypothesized that social change and concomitant shifts in consumer tastes and preferences have induced structural change and, hence, variation into the parameter structure of meat demand. Mundlak and Raussar (1979, p. 12) instance forecasts of meat demand in the early 1970's, based on sample data up to 1970, which consistently underestimated actual levels of demand. They suggest that social changes in the traditional role of women resulting in greater numbers of working women and the associated demand fbr convenience foods, augmented actual consumption levels. In contrast, the recent observed shift away from red meats, especially the fattier table cuts of beef, due to health considerations, would reflect social change but in this case an overestimate with respect to actual consumption levels would be the result. 37 + 134 It is these sorts of changes and influences which can be represented in a parameter variation equation. Following Mundlak and Raussar (1979) equations (3.41) and (3."2) may be combined to produce y 3 Xt Bo + H U(old point), the new point is accepted as an 144 improved one. If a point repeats in giving the lowest function value in terms of U(.). then it is moved half the distance to the centroid of the remaining points. In Figure 3.5 point 2 is an inferior point with respect to 1 and 3. However, X, the new point is even worse. Therefore, 2 is moved half the distance to the cen- troid, C, to the new point 4. The procedure of 'complex' is to eventually straddle the optimal point, 2. Figure 3.4 A 'Complex' of Three Points 145 Figure 3.5 Movement from Inferior Boints in 'Complex' Optimization Procedure (e) Rule 5: The new point is checked 'against the constraints and is adjusted as in rule 2, to satisfy all constraints. (f) Rule 6: Convergence is assumed when the values of U(.) at each point of the 'complex' are within Buunitsu3 for a consecutive iterations (K , a , B , and y are exogenously set in the program). An iteration is the set of steps *— u 3The convergent parameter, 8 is in units of U('). Selecting an appropriate 8 is key to this complex technique. 146 necessary to achieve a new feasible point in the 'complex'. The larger is Y, the more rigorous is the solution process.uu The 'COMPLEX' algorithm requires starting values or best guesses for the xN variables, plus valuesfor K, a, B, Y and 5 (explicit constraint violation correction). The algorithm will then provide values for U, the X's and the number of iterations taken to reach that objective function value, U. Criterion for Identification of 8 Values It is not always possible to identify every unknown parameter in the model. Nevertheless, identifiability implies obtaining estimates within the neighborhood of the real values. Hence, the following criteria are established: (a) each unknown parameter B1, 82,..., 8N must influence U, and (h) each member of the 8 set must have an independent effect on U. The effect of a change in one 8 cannot be offset by a change in another one, i.e., 3U ~ .32 _.. A AU ‘ A B "' aBN 8N (3.53) 3U — AB +eeo+ 331 1 as, 2 *— u 1‘A flow diagram illustrating this algorithm is contained in Knes- ter and Mize (1973. Do 371)- 147 Criterion (b) requires that AU=O, which implies implies A81= 482 = ... = A8" = O in some neighborhood about the optimum. In essence this is a condition of linear independence. There are many optimization techniques which feasibly could be applied to the present problem. The classical gradient method and the improved gradient method are two such search algorithms. How- ever, the former is not selective in that it starts on one slope and stays on it. This risks obtaining a local optima and not a global optimum. Also, if the iterative steps are too large the solution fails to converge and may over shoot the optimum point. The improved gradient method utilizes the COGGINS algorithm, the optimizing subroutine (Box, Davies and Swann, 1969). This method overcomes the local optima problem by iterating along a ridge. However, the problem, with this method is that it is a single vari- able maximization, i.e., of U(X) and is not subject to constraints. It does have application as a building block within multivariable search methods. For both of these methods a minimum or maximum is reached when the gradient vanishes. By contrast, COMPLEX is (a) more efficient than gradient methods in terms of achieving a faster convergence; (b) tends to find the global optimum; and (c) has the added flexibility of introducing constraints. Within the class of multivariable seanch methods the minimiz- ing algorithm of POWELL may be considered (Powell, 1965). However, COMPLEX is to be preferred until iterations get within the neigh- borhood of the optimum (after which convergence is slow with COM- PLEX) or when a high degree of accuracy is not required. POWELL, 148 which makes use of a quadratic convergence, does not perform well when the interim solution is outside the neighborhood of the optimum. Another shortcoming of POWELL is that this minimization of a multivariable, non-linear function is unconstrained. 3.4.8 Legendre Polynomials In optimal control problems the vector of controllable vari- ables. Xe, can be represented in discrete or continuous form. Diagrammatically, the representation of Xe, optimized over the time path 0 5 t g T, is given in Figure 3.6 below. _ i optimal xcoptimal . C. X opt. _. P O 1' O T (a) Continuous (b) Discrete Figure 3.6 Representation of Continuous and Discrete Controllable Variables in an Optimal Control Problem 149 The problem in this research (where the controllable X; are in fact the 8'3. i.e., Sc), may be stated: find 5 such that go will optimize U(T) over the time path 0 g t g T. Legendre polynomials provide a means of representing any con- tinuous function f(x) or f(t) over the interval (-1, 1). This set- of polynomials, {Pn(x)}, which are orthogonal functions, have found application in representing continuous system controlable inputs in optimization studies. Hence, a function f(x) would be represented 8831‘s f(x) ' a0P0(x) + a1P1(x) + ... + anPn(x) (3.54) for x: -l < x < 1. These polynomial functions Ph(x) are defined as: Po(X) ‘3 1 (3.55) P10!) = 3‘ P2(x) = a (3x2 - 1) 1 . Pn+1(x) --E:I [(2 n+1) x Pn(x) - nPn_1(x)] The nth polynomial is obtained by recursion and is based upon Po(x) and P1(x), *— 45 See Manetsch and Park (1981) for a more extensive background and development of this material. 150 The legendre polynomials are orthogonal over the interval (-1, 1) for x. That is, for i¥j (3.56) ll 0 f1 Pi(x) P (x) dx 1 1 1 _{ Pi(x) Pj (x) dx = {1 [Pi(x)]2 dx = 2—fiil , for i=j However, of particular interest here is the values which the a's will take in the function, f(x) in equation (3.54). An important property which allows the 5 vector to be obtained is called the finality of coefficients property. Essentially, this means that the a's are independent. This is a special property of’ orthogonal functions which is shown below. Assume a function f(t) such that: N (3 57) f(t) = n20 an¢n(t) - where t 13 time 34d ¢n is some basis function. A representation of f(t) is desired that improves as more terms are added as N'*“’in (3.57). It is also desired that as terms are added to this approx- imation of f(t), previously determined coefficients, a1, a2,...,aN do not need recomputation. This finality of the coefficients, how- ever, can only be obtained if the [if I’1 P b b. slope, Aq-p -> Af 2 Dz 1' p, Q q! q: P a. slope, Aq-p ‘9 ‘f c ...... d d. slope, Aq-p -> M P 1 7"\\\ D D] 2 Q ql P e. Aslope, Aq—p -> f p; f. Aslope, Aq-p -> At 177' A central focus in this study is on how market responsiveness to changes in quantities available and in incomes has influenced retail prices of individual meats in the U.S. over the past 30 years or so. The importance of understanding the nature of these changes in market response, as measured by elasticities or flexibilities, was discussed in Chapter 3. A number of researchers have observed that flexibilities and elasticities have not remained constant during this time. More particu- larly, they have shown that estimated demand curves (linear in arith- metic values) have generally become flatter over time.18 This may imply an increasing demand elasticity. However, this graphical evidence is not in accord with empirical evidence that, in the case of elasticities, demand has become less elastic (more inelastic) and for flexibilities, demand has become more flexible. Breimyer (1961), Tomek and Cochrane (1962) and Tomek (1965) have been among those to make this observation. Tomek recognized the need to explain these changes in empirical measures of consumer responsiveness. To do this requires that the three sources of change in responsiveness be examined. Most commonly, econom- ists explain changes in demand flexibility as being due to demand curve shifts through time in response to changes in incomes, available quanti- ties of substitute commodities, or tastes and preferences. However, this interpretation ignores the slope of the demand relation, i.e., the relationship between quantity and price, which may also change. A bound on the nature of some of these changes may be obtained by applying the Slutsky relation and the homogeneity condition of consumer 18This phenomenon is illustrated later in this chapter in the graphical analysis of retail meat demand. 178 demand. As incomes increase, the related increases in food consumption generally become smaller. In consequence, income elasticities (in this case), for food will decline.19 From the Slutsky equation, when transformed into elasticities, a decrease in the income elasticity implies, via the fall in the income effect, that price elasticities will also become smaller.20 That is, as incomes rise relative to the price of a commodity, the income effect of a price change will be smaller. Using the homogeneity condition, a fall in income elasticity, assuming no change in substitutes or complements, will result in the own-price elasticity becoming smaller in absolute magnitude (less elas- tic or more inelastic). As discussed earlier in this chapter, the homo- geneity condition also holds for demand flexibilities. Hence, in an opposite direction, as the income flexibility rises, the absolute magni- tude of direct price flexibility will become larger, i.e.,more flexible. Demand flexibilities for meats are likely to rise over time. First, as disposable incomes increase and income flexibilities rise, direct flexi- bilities become larger in absolute value; and second, as an economy develops, consumer preferences change in a way that makes the demand fbr certain foOds more price flexible. Offsetting these effects are the 1 9Empirical evidence supporting this observation is contained in Fox (1958, pp. 125-129, 1&1). 20 This follows from the Slutsky equation Sq Bq 3Q ___.___ (1.--, 391 31:1“l Byp where utility, u and prices, 5 are constant. By multiplying each term by p/q and the last term by y/y the gross price elasticity equals the net price elasticity less the income elasticity times expenditure on that good as a proportion of income. 179 possible changes in prices, number and closeness of substitutes. Changes in consumer responsiveness to the availability of substitutes suggest that the substitution effects, i.e., cross-product flexibili- ties, may also be subject to change over time. Theoretical reasoning suggests that cross-product flexibilities would be higher fer close substitutes. An increase in the quantity available for a close substitute will shift the own-product demand to the left. Own-product demand, D(q1,q2,.,.,qn), 13 a function of cross- product quantities. This shift and, hence, the effect on price of the own-product, will be greater for a close substitute. Another way of viewing this substitution effect is in terms of effective demand. From a position of equilibrium, increases in quantities available of a strong substitute will divert demand away from the own-quantities causing a greater excess supply situation and a greater concomitant own-price adjustment than if the substitute were a weak one.21 Advertising, promotion and the forms of market and consumer educa- tion of substitute commodities will make them stronger substitutes and will tend to make the own-product more quantity flexible and increase the impact of substitutes on demand fOr that.good in question. 0n the other hand, to the extent that attempts to product differentiate result successfully in less substitutability, demand flexibilities will tend to be smaller. 21Note the parallel in the interpretation of cross-price elastici- ties and cross-product flexibilities. In both cases the closer the substitute, the stronger the cross effect and hence the larger the cross-elasticity and the cross-flexibility. 180 On balance, a reasonable hypothesis is that direct flexibilities of demand for individual foods are becoming greater through time. Cer- tainly, food aggregates are becoming more quantity flexible, simply because incomes are rising and aggregation implies fewer substitutes. For specific products like meats, subJect to production cycles, it may be more realistic to hypothesise, not continuous increase in flexibili- ties, but cycling change in demand flexibilities. One theoretical note needs to be made regarding the interpretation to be given to an income flexibility. Houck (1966) has argued that, on theoretical grounds, the income flexibility of demand must be unitary. This result is derived from the homogeneity condition for elasticities when all products are considered. If the estimated income flexibility differs widely from unity, he argues, that meaningful interpretation of it becomes difficult. Houck further states that 'when zero-degree homo- geneity is assumed among prices, income and quantities demanded, the flexibility coefficient on income is not a behavioral measure in the same sense as the elasticity (pp. 228-229). He acknowledges, however, that not only may the flexibility differ from unity in practice, but the underlying conditions of homogeneity may not apply in any given empiri- cal research problem. The interpretation of‘ the income flexibility estimates in the current research is accepted at an intuitive level. A change in the 'aggregate' purchasing power of consumers can be expected to increase the demand for a commodity (except for inferior goods) and hence put upward pressure on market prices. This results in a positive relation- ship between income and market prices. Alternatively, a fall in 181 consumer purchasing power will influence their willingness and ability to pay for a particular good or to buy as much of it. For a given level of product availability over a specific time period, assuming a normal or superior good, consumer demand will not be as strong and so the price will tend to fall adjusting supply with demand. Also note that the stronger the demand for a good in response to a rise in income, the higher the income flexibility (and income elasticity). u.2 Statistical Model and Estimation Procedure The discussion in Chapter 3 and in this chapter provided the main theoretical foundations for consumer behavior and how this theory relates to empirical analyses at the retail market level. This review was desirable as the theory is a useful guide and provides a Justifica- tion for the demand fUnctions chosen. The theory is a guard against 'some of the absurdities and inconsistencies which may arise from prag- matic models if the latter are used without considerable care and exper- tise' (Brown and Deaton 1972, p. 1152). This gives the estimated demand functions theoretical plausibility. Nevertheless, as also pointed out earlier, a substantial gap exists between the theory of consumer behavior, on the one hand, and empirical analysis on the other. It is in recognition of this gap that economists have chosen certain specifications of demand fUnctions over others. Houthakker and Taylor (1970, p. 1) refer to practical considerations in the design and development of their study. They report that 'common sense' and a ‘long process of experimentation and elimination' were necessary before a final equation was selected. of 182 Not all economists subscribe fully to this perspective. In defense of theoretical plausibility, Phlips (197“, p. 112) was encouraged to write Houthakker and Taylor's choice among specification is based on goodness of fit. ...while much has been gained in terms of descriptive power, much has been lost in theoretical plausi- bility. By introducing more realistic changes of the income elasticities, Houthakker and Taylor lose contact with the theory of utility maximization. Indeed, there is no longer any reference to a specific utility function. Much more, these specifications are not compatible with utility maximiza- tion, as they do not satisfy the adding-up criterion exactly. The approach is entirely pragmatic (p. 112). 0n the other hand, Brown and Deaton (1972, pp. 1151-1152), in their review of applied models of consumer behavior gave support, albeit qual- ified, to the pragmatic approach.22 They saw a clear role in many prac- tical situations for the procedure of estimating a set of single equa- tion models, one for each commodity: Single equation models, even if less satisfactory from a theoretical point of view, may still be able to out-perform complete models in terms of past experience and ability to project the future (pp. 1157-58). In empirical analysis it is often necessary, when attempting to measure consumer demand from a time series of annual consumption and Prices of a commodity, to go beyond the specification of the theoreti- cally derived demand function. Host econometric analyses of demand in h 2 2Deaton and Brown meant 'pragmatic in the sense that it includes those variables in which we are directly interested, ignoring or summarizing others.‘ (p.1151) ar 183 agriculture do not use explicit utility function formulations, but use arbitrary or ad hoc reduced forms.23 This approach is more commonly adopted ‘because of increased ease of estimation and the ability to incorporate greater complexity in the dynamic fbrmulation.‘ (Pope, Green and Eales, 1980, p. 778). Moreover, there are variables which directly influence consumer demand in addition to relative prices, quantities and income, as suggested by the classical theory. This is the view taken in this research. The statistical model to be estimated may be written in general terms, the demand for the ith meat specified as (p1 2b (1) -——- = a + --—- CPI t 1 3:1 13 POPC t + c ) + di 2 ...JL___. 1 (POPCoCPI 1: + “1; (5‘4) kt where 1:1,...,u (table beef, hamburger beef, pork, broilers), where pit is the real price of meat 1 at retail in period t, CPI is the Consumer Price Index (1967:100), POPC is U.S. civilian population. th is the domestic consumption of meat j, at retail weights, per person, yt 13 real personal disposable income per person in the U.S.. 2t represents all other variables considered to influence retail prices, and uit is a random error term. Coefficients ai (intercept), bij’ c1 and di are to be estimated.2u The signs expected on bij (i=j), the own—quantity coefficient, are negative, implying a negatively sloped demand curve. The signs expected 2 31One recent exception is the paper by Green, Hassan and Johnson (1978). 2” Later specifications of (5.") will allow these coefficients to be functions of time. 184 on the bij (iij), coefficients on substitute products, are negative also. For complements the sign would be positive. An increase in the quantity available for consumption of a substitute good should cause its price to decrease. This decrease in the price of the substitute causes a decrease in demand for the particular product in question, i.e., the demand curve shifts to the left. Assuming some upward slope on the sup- ply curve, this shift brings about a fall in the price of the product. Therefore, a negative relationship exists between the quantity of a sub- stitute and the price of the product in question. The coefficient, 01, on disposable income may be positive or nega- tive depending upon whether the product is an inferior or normal good. For the flour meat products considered in this analysis a positive sign is expected. u.2.1 Functional Form of Equations Behavioral theory of demand does not provide much guide to the choice of algebraic forms of the equations to be estimated. The compu- tational burden is simplified considerably by using functions linear in the parameters. Most common among these functions are the linear and the log-linear functional forms. The quest for an integrator of demand theory and empirical work has favored the use of the double-logarithmic demand form. This fUnctional form provides a means to implement the desirable homogeneity restriction plus it provides the relevant flexi- bilities (or elasticities) directly from its coefficients. Hassan and Johnson (1976, p. 22) appear less accommodating of the double-logarithmic form in demand theory because of inconsistencies with 185 classical demand theory. Yet, despite its ad hoc framework, these authors use this functional form and concede its superior fit in many cases, ease of estimation and the ready interpretation of the estimated coefficients when compared to utility maximization derived demand curves . Although the linear and log-linear models are the most widely used among linear economic models, there is evidence to suggest that neither of these formulations is to be preferred. Research indicates that more flexible functional forms obtained from Box-Cox transformations, using 25 (Pope, Green and Eales 1980; Chang maximum likelihood techniques 1977). The approach in this research is to test both linear and log linear functional forms and make a choice based on Theil's test of resi- duals (Theil 1971, pp. 5AA). These results are described in a later sec- tion. u.2.2 Method of Estimation Each model of retail demand was estimated using ordinary least squares. To check the appropriateness of this choice some alternative estimation methods were examined. For instance, with price equations estimated separately there is a possibility of contemporaneous correla- 26 tion among the dependent variables. This may result from prices of 2 5Tolley, Wang and Fletcher (1969) found mostly small effects on price and income elasticities and flexibilities of food demand from the use of logarithmic juxtaposed linear forms. 6Contemporaneous correlation among dependent variables and among disturbance terms implies that these variables and terms in dif- ferent equations are correlated, respectively, at a given point in time but are not correlated over time (Judge, et.al., 1980, pp. 2 5-251). 186 one meat being affected by price levels or changes in prices of other meats in the retail market. If this occurs, disturbance terms in dif- ferent equations will be contemporaneously correlated and OLS is not an efficient method of estimation. Estimation will, however, lead to con- sistent and unbiased parameter estimates. In such a case of correlated disturbance terms, Zellner's27 'seemingly unrelated regression' (SUR) will yield asymptotically more efficient estimates than those obtained by OLS provided the set of exogenous variables is not identical in each equation.28 The equation set is estimated as if it were a simultaneous equation system: in the first stage an estimate is made of the error terms' variances and covariances, based on the residuals derived from an equation by equation application of OLS. In the second stage the esti- mates are used to contrast the Aitken generalized least squares estima- tor of the regression coefficients in all equations. To examine the possibility of contemporaneous correlation of the residuals, SUR esti- mates are obtained for the set of meat demand equations. The more usual simultaneity between prices and qualities, using two-stage least squares (2815) is also examined. Finally, the hypothesis that total adjustment to a long-run equili- brium might not occur within one year is tested by introducing a lagged dependent variable in each of the beef equations. Lagged adjustment of prices to a change in quantity available may be due to imperfect knowledge, habit persistence or institutional delays. The coefficient 2 7for details see Zellner (1962; 1963). 2 81h meat demand studies see, for example, Tryfos and Tryphono- poulos (1973), Main, et al. (1976) and Hassan and Johnson (1976). 187 of adjustment, 1, can be obtained by subtracting from one, the coeffi- cient for the lagged dependent variable. This specification permits estimation of a short-run and long-run direct flexibility. u.2.3 Estimation Period and Data Limitations The model was estimated using annual observations over the period 1950-1982, providing 33 observations. Shorter data periods are fre- quently preferred in many studies, primarily because of an explicit forecasting objective or in order to circumvent the need to model such changes in structure as might occur over a longer estimation period. Since an important objective in this research is to investigate struc- tural changes in demand, a long data series was used. Market adjustments are continually taking place and hence, there exists no natural time period (in terms of observation steps) with which to capture structural change. The periodicity in the economic relation- ships concerned had much to do with choosing annual observations. Monthly and even quarterly observations would tend to reflect brief and passing phenomenon as well as seasonal influences on consumption and prices of meats. These rather short-term phenomenon were not of interest here. Even so, the choice of annual data is to some degree also arbitrary where economic adjustments are continuous. Breiymer (1961, p. 60) provides some insights to the problems aris- ing from the choice of data points. He argues that the actual produc- ttion span or cycle from which the empirical data arise is central to this choice. He felt that 188 the results of a statistical analysis can be interpreted in terms of the length of run only by examining closely the par- ticular time series incorporated in the analysis. That is, it is necessary to look into the reasons which explain why the series of annual data depart from their mean as they do. If long-run trends are the dominant factor, the results relate principally to long—run trends. If production cycles are instrumental, results relate thereto. If only erratic annual fluctuations are to be found, results apply to them. It follows that if any continuous trend exists, the longer the time span included in the statistical study the more does the long-run trend influence the result. The shorter the nunber of years that are included, the more do cyclically-evolving or even briefer phenomena. (p. 60) One major difference between Breimyer's analysis and the present analysis relates to his assumption of constancy of estimated parameters. In his study, fixed coefficients estimated over a long series will give coefficients which reflect average response over the period.29 In the following analysis, progressive relaxation of this assumption of fixed coefficients should provide additional insights into the character of these structural changes in meat demand. Actual data used in computations and all sources of these data are provided in Appendix A1 and Appendix C. Derivations of the prices and consumption series, in particular, the fed and nonfed data series, are Given. Some limitations inherent in data used are discussed in the fol- lowing section. ...—— 2 9Breimyer did relax this assumption somewhat breaking the estima- tion period into three periods. 189 1.3 Structural Model and Candidate Variables The estimated structural model may be written, in terms of selected candidate variables, as the following set of retail meat demand equa- tions. RPBFR = f(DCFBFRC, DCNFBFRC, DCPKRC, DCBRC, DICR, Z) (u.5) RPHMBR : f(DCFBRC, DCNFBFRC, DCPKRC, DCBRC, DICR, Z) (“.6) RPPKR f(DCFBFRC, DCNFBFRC, DCPKRC, DCBRC, DICR, Z) (“.7) RPBRR f(DCFBFRC, DCNFBFRC, DCPKRC, DCBRC, DICR, Z) (4.8) A description of variable codes and other variables used in subse- quent analyses is presented in Table ".2. It should be well recognized that beef is not a homogeneous pro- duct, particularly at the retail level. Clearly, at the retail level a 'commodity' such as beef is a collection of distinct cuts and grades, most of which are closely competitive with respect to a given end use. The decision to disaggregate beef into two categories, namely, table beef (roasts, steaks) and hamburger beef (ground beef, processed beef products) is based on (a) an attempt to recognize the heterogeneity in beef especially at the retail level, (b) a desire to relate the nature of that heterogeneity back to the broad categories in beef production, namely, fed beef and non-fed beef respectively, and (c) pragmatic con- siderations with respect to data availability and analytical feasibil- ity. 190 Table 4.2 Description of Candidate Variable Codes variable Code Name RPBFR RPHMBR RPPKR RPBRR DCFBFRC DCNFBFRC DCPKRC DCORMRC* DCBRC DCPMC DICR CPI POPC Z 5 to 24 25 to 64 Units of 'Measure Description 3 Retail price, choice beef, real, retail weights. $ Retail price, hamburger, real, retail weights. $ Retail price, pork, real, retail weights. $ Retail price, broilers, real, retail weights. lbs. Domestic consumption, fed beef, retail, per person. lbs. Domestic consumption, non-fed. beef, retail, per person. lbs. Domestic consumption, pork, retail, per person. lbs. Domestic consumption, other red meat, retail, per person. lbs. Domestic consumption, broilers, retail, per person. lbs. Domestic consumption, poultry meat, retail, per person. lbs. Disposable income per person, real. index Consumer Price Index (1967 - 100). millions Population, Civilian, U.S. - Collective term for all other candidate variables. 1 Proportion of population aged 5 to 24 years. 2' Proportion of population aged 25 to 64 years. *DCORMRC is a variable combining pork, veal and lamb. 191 This classificatory scheme is largely production derived3o since retail consumption (disappearance) data are based on whether or not the animal came out of a feedlot. Feedlot produced beef, steers and heifers mainly, are assumed to supply all choice beef cuts. However, as pointed out in Chapter 2, fed steers and heifers also contribute though to a much smaller degree, to ground beef supplies. This approach is an improvement on studies which classify all steers and heifers into the choice beef category. Hamburger beef is assumed to be provided by non- fed beef, cull cows and bulls and grass fed steers and heifers plus imports. Again, there is a proportion of this beef that finds its way into table beef supplies, though the proportion is considered small. An additional advantage of this classification of beef type is the close relationship between U.S. non-fed beef and Australian imported beef. Both are considered to be manufacturing quality going primarily into ground beef and other processed beef products.31 Since no official data series on table (fed) beef and hamburger (non-fed) beef consumption exists, various approaches to measuring and estimating these data have developed. The simplest approach is to clas- sify all steers and heifers as table beef, although this would result in some overestimation of fed beef. Crom (1970) developed and used a data 30Some researchers have preferred an end-use approach, attempting to measure the quantities of high grade and low grade beef enter- ing the market. See Bain (1977, pp. 16-18) fer a discussion of the two approaches and reference to authors who have used them in classifying beef. 31This dichotomy of beef products is not new. Langemeir and Thompson (1967). Crom (1970), Hunt (1972), Freebairn and Raussar (1975), Macauley (1975), Bain (1977), and more recently, Reeves (1979) used this approach. 192 series for fed and non-fed beef production/consumption which has been used by Freebairn and Raussar, Macauley, Bain, and Reeves.32 An alternative data series of fed and non-fed beef production/consumption has been developed for this research.33 This series, while similar to Crom's takes advantage of new data series available and overcomes problems of discontinued data series. In essence, fed cattle are fed steers and heifers only. Non-fed cattle are cull cows, bulls and grass fed steers and heifers. Adjustments are made to accommodate inventory, and imports to provide estimates of domestic disappearance in retail weights. Limitations of available data prevent any major improvement on this classification. Some 25 percent of 3 fed steer or heifer carcass becomes ground beef. 0n the other hand, a portion of a non-fed (forage fattened) steer or heifer carcass and certain cuts from cow carcasses are sold as choice or table feef. Also, the cut-out of a given carcass into table beef or ground beef depends on relative prices and hence that cut-out will vary over time. In this study fed steer and heifer car- casses (fattened in feedlots) only are classified as choice or table beef. The remainder is classified as processing or ground beef. During years of low farm prices for fed beef, an increased proportion of steers and heifers are marketed directly from forage-based operations and, ‘___¥ 32$ee Bain (1977, p, 16) for an outline of this method of estima- tion. 33Details of the method are presented in Appendix Table A2. These estimates are based around a method developed by Ferris of Michi- gan State University. 193 therefore, do not enter feedlots. These cattle are classified as non— fed beef although meat from these animals will be marketed as table beef. Therefore, some inflation of the data on non-fed beef (or 'ham- burger' beef) occurs during these low farm price years. Perhaps a more accurate dichotomy is fed beef and 'other' beef. However, the only retail price series for lower quality beef is the hamburger/ground beef price series. To the extent that 'other' beef is of manufacturing qual- ity this price series should be representative. The underlying assump- tions and associated limitations should be borne in mind when interpret- ing results. Retail prices of choice grade beef and retail price of hamburger beef, as defined by the USDA, are the normalized endogenous variables in the two beef demand equations. At the retail level these two price series, while being readily available3u would appear to be representa- tive of the prices of the two beef categories mentioned. To the extent that institutional prices for ground beef, i.e., at fast food restau- rants, deviate from this retail series some bias may be present. The retail price of pork and the u-region average retail price of young chickens (broilers) are the normalized endogenous variables for the pork and broiler demand equations, respectively. Separate data series, each for domestic consumption of table beef, hamburger beef, pork and broilers, in retail weight equivalents, per person, were specified in each demand equation. ‘ u The hamburger price series has recently been discontinued. Up- dates of this series would need to make use of the ground beef price data series of the BLS. 194 Personal disposable income, provides the budgetary constraint on demand. This variable is included to account also for the increased purchasing power of the U.S. population over the sample period.There are two reasons disposable income is exogenously determined in this model. The first is simply that beef, pork, and broilers comprise a very small part of total per person disposable income. All red meat and poultry meat amounts to little more than “-5 percent of total expenditures per person (USDA 1981, p. 21). A second reason follows from the classical theory of consumer demand. Since total expenditures made for each com- modity group is determined in the initial budgetary allocation process of the consumer, the group expenditure for individual demand functions is predetermined (George and King 1971, p. 27). Hence, the group expen- diture is a predetermined variable for the demand function of an indivi- dual product within the group. Population is usually incorporated into demand studies by express- ing consumption and income data on a per person basis. This study does not deviate from this common practice. However, it may be of use to restate the implications of this practice, along with its difficulties, especially given the potential role changes in the population may have had on consumer demand for meats in the U.S. Expressing these variables on a per person basis is consistent with the underlying theory of consu- mer choice which refers primarily to individuals. More importantly, for market level demand studies, per person relationships are likely to be more meaningful and stable than relationships between aggregates. A potential difficulty of this use is the assumption that all persons be given equal weight irrespective of age and sex. 195 In principle, a different scale of weights or 'adult-equivalent scale' (AES) should be used at least for each commodity. However, available evidence suggests that equal-weight scales do not produce much distortion (Houthakker and Taylor, 1970, p. 29). This is because the distribution of the population by age and sex is fairly stable. Claffey (1982) estimated an AES for use in meat demand analysis. Since children consume less than adults the AES population is smaller than the official civilian population, causing AES per person consumption to be larger than the 'official' per person consumption of meat. However, following examination, there was very little difference at the national level in the two series. Although, it was acknowledged that the AES approach may have value in regional studies of meat. In the present study, there are, nevertheless, a priori grounds for including specific demographic variables to account for a particular demand relation phenomenon due to the impacts of certain segments of the population. For instance, in Chapter 2 it was shown that younger seg- ments of the population consume more ground beef although older segments consume more roasts and steaks. Like population, changes in the general price level, represented by the CPI, are included implicitly in each equation. That is, prices and income per person are deflated by the CPI to express these data in real 196 terms, hence adjusting for the effect of inflation.35 The assumption underlying deflation of price and income is that consumers are not sub- ject to money illusion. In terms of the theory of consumer demand this implies that the demand functions are homogeneous of degree zero in prices and income per person. However, deflation is also a pragmatic compromise since it is impossible to accommodate the theory precisely. The usual approach is to include in the demand relation all those sub- stitutes and complements as prices and quantities and use a price index such as the CPI for all other goods (expressed in terms of prices).36 This price index may be used as a deflator or as a separate independent variable. Here, CPI is used implicitly as a deflator as these equations appeared superior to expressing it as a separate variable. The variable, Z, in equations (9.5-fl.8) represents all other shifter variables, explaining changes in retail prices of table and ham- burger beef, pork and broilers. Demographic variables, i.e., of specific age segments of the population, have already been mentioned. Theoretically, changes in tastes and preferences result in a different utility function and, therefore, a whole new set of demand relations. 35Tolley, Wang and Fletcher (1969, p. 709) found in price depen- dent equations of aggregate food demand that the total bias in price elasticity estimates due to incorrect deflation was greater than in the consumption dependent case, partly as a result of the effect of correlation of the deflator with the residual. Total bias in income elasticity estimates is small in both cases. Total food is an important component of the CPI and therefore some correlation is expected. Such correlation between deflated prices for individual meats is likely to be very small. 36See Zellner. Gallo and Levey (1980, pp. 20-21) for a discussion of alternative measures of inflation. 197 Intuitively, these changes in demand relationships may cause more signi- ficant shifts in demand than those caused by changes in incomes or quan- tity availability. Tastes may encompass a multitude of effects such as fashion, attitudes to health and preference for leisure. Habits develop after tastes and preferences have been formed. Historically, tastes for meats have not changed rapidly over time. As discussed in Chapter 2 there has been in recent years, a decline in beef demand which some observers feel have been due to taste changes. The treatment of ana- lyses of these changes in demand are discussed later. 4.4 Results of the Base Model of Demand 1.1.1 Estimation of a Base Model The estimated retail demand equations for the four meats plus an equation for ‘all beef‘ were estimated as linear functions using OLS and are shown in Table 9.3.37 For the beef equations, the statistical esti- mates were improved by combining pork, veal and lamb consumption into a variable of ‘other red meat‘ consumption.38 Similarly, in the table beef 37Tests for choice of functional form were inconclusive so the linear form was chosen. Contemporaneous correlation of the resi- duals was tested by estimating Zellner‘s SUR (Appendix Table A1). Notably, unexpected positive and insignificant signs were obtained on pork in the table and hamburger beef equations. 381n the case of the all beef equation, consumption of veal and lamb was esthmated as a separate variable from pork. In this case no improvement was gained by combining these variables. However, in the table and hamburger beef equations this separate specifica— tion was not stable in later analyses. .198 .n .H. .8 won «.Q can so uaunuquaean . .ceos cacao». no « used-s .u. succesum N u . euauaocoou aces-ounce use. . . ..uausoaueooo snags ...oauacu-a ea aa>a u nonuqz1eaau=o 1.: o . a saws-a. .nosau assumes a a. o>«asuoeous« souucuouuououss we once u us“ ca ueeaoauuooo u mw . ace-uouce oaaauass co .— «sooosuu no assuaov uou seasona- couu u o momma aumvcauo I Ixo u 0‘ «sun-«us> usovcoaso on» as can. ecu an msvuoun seasons 0 u 111111111111111111111111 111111111111111111111111umwn11 o Ada .uau on.~ «a. sac. “Mn.ss As~.as .aa.o Aus.so “MHHNC Amw.om_ Aaanaav no a aoo.‘1 mo~.1 omn.n1 1 ~Q0> Q5 nouns: .essq use: 1 o¢a>zaua cannon 111111111111111111111111 eon-H e a AMQoGV o a a a o .. .... as a.” use an... an a... on.. so. . Anu.ov Aaa.os . . . . as es. a... “ma... “mu; “who “an”: issue as .oes n~._ a”. . Aas.no A_a.so . . . . “to sea. as"..- www.mv “Mm.ww www.mw www.mw~ Assamese coon sauces-an .03.“ unofl e o AQNon o «A use one. Ann.no Ann... Ans.av Ass.so Anc.so . has a- na~.1 ~_h._- «sm.ss. amm.on_ assuage coon oas a e a so J 3 e um. Alxc can anonuoun nuansom anon use: on: Home use: some canon woeuo nonsense: smash 111 cameo unease oaxaooo ouanazoo newsman coats season cosaauoo . also: H ”nous: .coawom use couucssacoo vacuole: uncommon mounouum> mucumcmumxm odnwuuwb amoumoaoa NOOuIOnG— venudh 0:» no seen: on. " >o noon-«sum m use: you accuumsvu coals: Human: usualuuum n.e smash 199 equation, broilers and turkey consumption were combined in a ‘poultry meat‘ consumption variable.39 Elsewhere, specific product variables were maintained. Fish consumption was included initially but later excluded because of its unexpected positive sign. In recent retail demand stu- dies using quantity dependent specifications, considerably greater amal- gamation of variables was found necessary, preventing the individual product effects to be estimated.“0 These particular equations form the base model and do not represent the final specifications. Other specifications are tested, in later analyses of structural change, by allowing the regression coefficients to change over time. Nevertheless, several important observations may be made from these equations. First, all the estimated coefficients are significantly above the 99 percent confidence level, except the sUbsti- tute products, pork and other red meat. Other red meat in the beef equations is significant at the 90 percent level, but pork is insignifi- cant in the broiler equation. Multicollinearity among explanatory vari- ables might explain the low significance on an expected substitute like pork. However, the correlation matrix for first-order correlation among explanatory variables, did not indicate this possibility. It is possi- ble‘that the nature of multicollinearity is higher than first order. A 539This procedure saved degrees of freedom and did not greatly af- ‘.fect the other estimates in the equation. Generally, autocorrela- tion was lessened, the standard errors of the coefficients were reduced slightly, and explained variation in the equation was in- creased. u . 0See, for example, Reeves (1979). Freebairn and Raussar (1975) and Houck (1974), where, in fed and non-fed beef equations, cross-price effects were not estimated or at least were estimated as ‘other meat.‘ 200 low level of significance of the estimated coefficient on the pork con— sunption variable has been found in other meat demand studies. These variables were retained in the equations even when insignificant because they had the expected negative sign of a substitute product. It should not be overlooked that pork may have an inherently low cross effect on other product prices. Second. the 82 values on equations, except for broilers, suggests some variation in the dependent variables remains unexplained. However, prices tend to vary considerably less at the retail level (because of certain retail pricing practices, e.g., price smoothing and other lapses from perfect competition) than they do at the wholesale or farm level.In This will tend to result in a relatively lower R2 at retail. Deflated retail prices also tend to fluctuate less than consumption per person and therefore R2 values derived from quantity dependent equations tend to be higher. Moreover, considering the change which the retail market has undergone over the past 33 Years since 1950, it is perhaps too much to expect that the same combination of traditional variables would explain the at times complex and dynamic set of forces influencing “1For empirical evidence, see in George and King (1971, p. 62) the elasticities of price transmission from farm to retail. For a one percent change in the farm prices, each of beef, pork, lamb and chicken, retail prices change by between .5 and .8 percent, in the same direction. 201 retail prices over this period. When some of these changes are expli- citly accommodated into the model specification, a greater degree of explanation can be expected.”2 Third, the Durbin4Watson statistic, an index measuring the existence of first-order serial correlation in the disturbance term, suggests either the absence or inconclusive evidence of such correla- tion. Estimated mean price and income flexibilities, derived from equa- tions in Table 4.3, are presented in Table 4.4. A common practice is to draw comparisons and to establish the extent of agreement between ones own estimates and those of other economists. However, there is a dirth of comparable studies in which flexibilities have been estimated and so no direct comparisons are possible. Qualitative comparisons only are nae. Fox (1953, p. 43. 52) and Breimyer (1961, pp. 61-78) are among some of the earlier studies in which flexibilities were estimated from annual data. Compared to each other and to this study they use dif- ferent functional forms, data transformations and model specifications. Fox used first differences of logarithms whilst Breimyer used loga- rithms, undeflated retail prices, deflated income and included the CPI as an explicit variable. The estimates of Fox cover a period before uzThe coefficient of determination is a useful index of the good- ness of fit of a regression equation to data. Predictive preci- sion_%f the equation, however, may be of more practical value than the . In this regard, the closeness of say, a 95 percent confi- denceaband about the regression is more important than the size of the R , per se (Bartlett, 1974). 202 .m.# ”H369 GH G0>Hw mwufin—HUQQ flOfimwQHNOH BORN mfiflg wufimv UN fiQuQHfiUHMU GHQ mflfiUuuHfin—Hunmdh "GUOZ mo.~ 55.1 “H.1 . ~m.1 No.1 moon HH< qu>zquo Daemon n¢.N ~¢.~1 mo. 1 no.1 om.1 numaaoum we.“ mo.1 mo.~1 we.1 on.1 xuom mm.~ mm.1 mm.1 no.1 ~5.1 moon umwusnamm no.” no.1 -.1 mm.1 No.1 «mom canoe MUHG canon Uzmun Gammon Umzmoon Oahmhzun ommmmon madam assume commence uaoosH madame Ham woumHumn domuwm Hum flowugfiamfloo uwummfion unmade use: meuom mo mmauwuwnwxmflm seeds" was muausmso Homo: comm mo mommaqucm ¢.¢ mdnmh 203 1941; Breimyer‘s estimates are pre-1961. Both studies deal with aggre- gate beef and in most cases have used substantially different variable definitions. Fuller and Ladd (1961) estimated a quarterly model of beef and pork, at the wholesale level, and also aggregated beef. The few more recent estimations of direct and income flexibilities provide only limited opportunity for comparisons. Langemeir and Thomp- son (1967) estimated fed and non-fed beef (farm and retail) own-quantity and income flexibilities using an annual simultaneous equation model. Their estimates of own-quanity and income flexibilities were higher than those in Table A.“ and those published by Fox and by Breimyer. Also, as‘ expected of data from the pre-1970s, the retail cross-effect of non-fed beef consumption on fed beef prices is smaller and the cross-effect of fed-beef consumption on non-fed prices is larger than those estimates in Table ”.9. Non-fed beef (hamburger beef) has become a closer substitute with fed beef and has achieved a greater degree of favour in many consu- mers‘ minds. Crom (1970), who along with Langemeir and Thompson, was among the first to make a serious attempt to disaggregate beef into fed and non- fed beef, estimated quarterly price equations of wholesale demand. In that study flexibilities from linear OLS estimations were not presented. Colman and Miah (1973) estimated a quarterly model, also of wholesale prices, of major meats in the U.K. The quarterly flexibilities for beef (aggregate) and pork were about equal to or slightly less than those for all beef and for pork in Table 4.4. A priori, one would expect that, largely because of adjustment delays due to habits and legs in market flows, the shorter the period between sample observations the lower the 204 consumer response (smaller flexibility) to a change in quantities avail- able. There is simply less time to respond to a given change. Also, flexibilities estimated at wholesale may be expected to be greater than those estimated at the retail level. This is because of the more com- petitive structure of the market at the farm level, if not the wholesale level, than that which exists at retail. For example, as noted earlier, price leveling and price averaging by retailers tend to .restrict the response of changes in quantities on retail prices. Empirical support of these observations is provided in two papers by Ferris (1974, 1981). Using annual data, the direct flexibility for fed beef was -1.5 at wholesale and -.87 at retail. The relative inflexibility of the esti- mate at the retail level would seem consistent with figures in Table “On. The only other available meat demand study in which flexibilities were estimated is that by Roberts and Heady (1979). In an annual model covering the years 1953-1976, retail price equations were estimated, using OLS (linear arithmetic), for beef, pork, lamb, chicken and turkey. Beef, pork and lamb were estimated in carcass weights in contrast to retail weights in the present study. No disaggregation of beef was attempted. Specification of a lagged dependent variable (retail price) provided short and long-run flexibilities. Results of that study are compared in the following section. Bearing in mind that these are not final results, the relative mag- nitudes among flexibilities in Table “.4 seem consistent with expecta- tions. The direct table beef flexibility may be slightly under estimated or the cross flexibility on poultry meat may be over estimated 205 in the table beef price equation. This apparent biasedness, due most likely to misspecification, is largely corrected in later models. Nevertheless, the relative magnitudes of direct flexibilities appear to be in the expected order. For instance, the greater inflexibility of table beef demand vis-a-vis hamburger beef demand reflects consumer preferences for quality of beef cut. A one percent increase in quantity of table beef supplied per person is absorbed into effective demand more readily than a one percent increase in hamburger beef because of its preferred quality status. At the consumer level, another reason could be the more inelastic demand for hamburger beef at fast food restau- rants. Hence, the increase in table beef supply has a smaller price depressing effect on table beef prices than does the increase in ham- burger supply on hamburger beef price. Similar reasoning can be used for hamburger vis-a—vis pork and pork vis-a-vis broilers. Also expected is the greater cross—quantity effect of table beef on hamburger beef than the contrary effect of hamburger beef qualities on table beef prices. Although, hamburger beef quantities appear to have a greater effect on pork prices while table beef quantities have a more important effect on broiler prices. Recall that the higher the cross- flexibility, the closer is that product a substitute. The cross-effect of broilers on table beef may be overestimated. The likely cause of this over estimation is misspecification error due to omitted variables and the constant coefficient specification. This problem is (addressed in the estimation of later models. 206 Nevertheless, broilers (chicken) are the closest substitute in con- sumers‘ eyes for table beef and pork. Table beef is the closest substi- tute for hamburger beef and chicken. Pork is the weakest substitute in all other meat equations. The results raising the greatest concern are the income flexibili- ties. Products which are luxuries tend to be more highly preferred and tend to be consumed by higher income groups. These products are expected to have the highest income flexibilities. A priori, table beef demand should fit this category more so than hamburger beef. That is, table beef should have a stronger income effect and be more responsive to changes in incomes. Multicollinearity between income and consumption of table beef and poultry meat exists, given the high first order corre- lation between these explanatory variables. However, this causes inef- ficient estimators, not biased ones. More likely, the apparent biased- ness in the income coefficient is due to specification error, i.e., the omission of relevant variables, incorrect mathematical form or incorrect inclusion of the error term. Remember that these are average effects over a 33 year period, flexibilities of which are derived from a fixed coefficient model. Broilers have grown to greater prominence in consumption and more recently, attitudes have appeared to have moved unfavorably for beef. Relationships in more recent years, which are also likely to have changed, will be examined in subsequent analyses. 207 u.u.2 Beef Price Adjustment Model Estimated price adjustment equations of retail demand for table beef, hamburger beef, and all beef are presented in Table ”.5. Short and long-run direct and income flexibilities, derived from this table, are shown in Table 4.6. The price adjustment model is obtained by inclu- sion of a one-period leg on the dependent price variable. For various reasons, total price adjustment following a given price change in a causal variable may not occur within one year. Imperfect knowledge, the persistence of consumer habits and institutional impedi- ments to market processes are primary factors. The coefficient of adjustment indicates the proportion of adjustment that is instantaneous. For beef products the proportion of adjustment in prices estimated to take place in one year is around 70 percent. The short-run flexibility for an independent variable at the variable mean may be estimated by multiplying the coefficient of the independent variable by the ratio of the independent variable mean and the mean of the retail price. The long-run flexibility at the variable mean is derived by dividing the short-run flexibility by the coefficient of adjustment. Therefore, the greater the amount of adjustment in the current period, the closer is the long-run flexibility to the short-run-flexibility. From Table 4.6, the direct flexibility for table beef implies that a one percent increase in domestic available supply of table beef per person will result in a .U7 percent decrease in the short-run and .67 percent in the long-run retail prices for table beef, ceteris paribus. Similarly, a one percent increase in real disposable income per person will, as a result of an increased demand for choice beef, cause a 1.28 percent 1208 oocmumooos mom neo.~ A: mu 03H d> HQUM UfiHU ..Aeuu .o .ucm_ .>m>uc=v Ho>oa assumes .ounmwus> woovcoooe oommma mo ousonou .nuuumausuo mundane uozuo mo soaucccaaxo n on soaucaouu . a us» ca use: .cauowucucxs u e m uou n.¢ manna mom .N occusm o: uo "mmuoz 393 832 A . . . . . . . 3 8 3n 3 A: 3 an 8 Ga 3 as s 2 So. «3. 2a. 3a.? «2... 8mg- «8.; ahead as; :4. cannon A2.: :93 God A . . . . . . am «V Awe av “No as Aoo av Ana «v 8 s 8. So. So. as. 93.? 8n... «3.? 2a.- :1: Essay cos. saucepan: Aam.~v Ahm.cv Ana. . . . . . . S e: so so 3 SN 3 E 3 Cm 3 3 A 2. 95. .25. 8n. 5.? 8a... 3.3.. «.5... 8a.:— 933 mean ~32. . or 53 seems sans. ozaoa ogooo omega: oceans 3 : NM. a\o accumsoo mouse Mucus: moumauon < mucosa adamuus> consom use soauoasmsoo odommsoa .mon panama moansuua> amoumcsfioxm manmwum> ucoocoqoa Now—19mmd magnum may uo>o woumaaumm ”cannon and: Humans mo ocoaumsvm usoaumsnm< moans moumsaunm n.c ounmp 209 me. as.“ Hm.~ we. - am.- -.- os.- as.s- nc.s- moon ~H< cameos Na. n».~ mo.a ac.s- ma.- ma.- mm.- as.s- an.- ~m.- no.1 wouwsoaa: on. mm.~ m~.~ no. - no.1 an.- c~.- as... om.- 50.- 5a.- moon o_nae as: sea as: ass . sax as: as: as: as: com scoassane< ca «gas spasm «cos pecan macs steam use; steam macs spasm ounce Hausa: seoauadmuoo «can oxaoo ogzzooo ‘ oaamuzoo oedema: consume: oloo:~ common use compassmcou ofiumoeoo Nam~1omau weaken oz» Ho>o moamlwumm o.v snack "sous: see-smsne< «can; mes-on aumuou mo nouumnuemxoum eleven use Auwucmsa cszuucoq mes chasm 210 increase in retail table beef prices in the short-run, and a 1.83 per- cent increase in retail prices in the long-run, ceteris paribus. Note that the base model flexibilities in Table A.u fall between their short-run and long-run estimates. Since some 70 percent of adjustment occurs in one year, i.e., the current period, estimated flex- ibilities using annual data seem to be predominantly short-run measures. This is consistent with the reasoning of Tomek (1962) that for many foods, including meats, complete adjustment (in the ‘long-run‘) is nearer to one year than to several or many years. Certainly the differ- ences between the short-run and long-run estimates do not appear large. It is expected that the long-run flexibility will exceed the flexi- bility in the short—run. In the short run the impact of imperfect knowledge and habit persistence on the market adjustment mechanism will prevent complete and instantaneous adjustment of market prices to clear the market. Only after some time has elapsed in subsequent periods will these final adjustments in prices take place. This implies greater longer run flexibility of retail prices to the initial change in the explanatory variable. The short-run and long-run flexibilities presented in Table 4.6 are consistent with this reasoning. Larger long-run flexibilities were also reported by Fuller and Ladd (1961, p. 202) and by Roberts and Heady (1979. PP. 35-36) for beef (aggregate) and pork. Authors of both papers expressed concern that these results of long-run flexibilities which are greater than short-run flexibilities, contradict the usual arguments relating to the short-run and long-run demand curves and the behavior of demand elasticity through 211 time. They gave as an explanation the premeditated over adjustment to prices by ‘enlightened' consumers, knowledgeable of the cyclical pat- terns of beef prices. That is, when beef prices are low relative to prices of substitutes, consumers purchase more beef than if the present price relativities were expected to continue indefinitely. This argument rests on the doubtful assumption that consumers are knowledgeable of beef cycles and that they can consistently over adjust their reactions to market prices at the retail level to accommodate for these cycles. A more defensible explanation is one that is couched in the validity of the original specification of the economic model, namely, that prices adjust to changes in supply availability. The clas- sical utility maximizing theory, from which Marshallian demand curves are derived, says nothing in its formulation about the short-run and the long-run (See Chapter 3). Moreover, the distributed lag model has no theoretical underpinning with respect to utility maximizing theory. The lagged specification is ad-hoc and pragmatic. Its justification is based simply upon a pre-specified structure of an underlying economic model of the market under analysis. As discussed at some lengths, it is not appropriate to judge estimated flexibilities against Marshallian demand curves or demand elasticities to ascertain their correctness or otherwise. The underlying causality inferred by a flexibility does not make meaningful this type of comparison. Given this causality, and the 212 related assumption that adjustments in the retail meat markets are through prices, it then becomes clearer that the long-run flexibility will exceed the short-run flexibility in the manner hypothesized earlier and supported by the results in all three studies.“3 4.4.3 Effect of Age Composition and Expenditure Away From Home on U.S. Meat Consumption In Chapter 2 it was considered that variables, in addition to the traditional price-quantity and income variables, may explain some of the changes in demand for retail meats, particularly over the past, rela- tively volatile, 10 years or so. Among such influences on the retail meat demands, were changes in the age composition of the population and the apparent trend towards increased eating away from home and the asso- ciated growth in fast food outlets and institutional food-service pro- grams, e.g., school lunches. In some respects these two developments may be related. For exam- ple, the consumption of ground beef is greatest among the younger por- tion of the population which have contributed to the growth in meals away from home. variables tested in each of the four equations were (a) the proportion of the population aged between 5 and 2“ years and the proportion aged between 25 and 64 years, and (b) away from home expendi- ture on food per person, in real dollars. These are some apparent shortcomings with these proxy variables. Ad hoc inclusion of a specific u3Grounds for rejection of these arguments and this conclusion would seem, therefore, to be more reasonably based on the assump- tions underlying the economic model and not whether there is correspondence between elasticities and the flexibilities. 213 age variable does not address the total age dynamics of population change. On the other hand, the use of an adult-equivalent scale to weight meat consumption by sex and age has not proven successful. After studying this problem, Houthakker and Taylor (1970, p. 29) advocated use of specific demographic variables for population segments where, in the case of a particular commodity, such an influence could be identified. A priori, a positive sign is expected on the 5 to 24 age variable and a negative sign on the 25 to 64 age variable in the hamburger beef equations. This follows from the evidence presented in Chapter 2 that the young segment are the heaviest consumers of ground beef. Therefore, an increase in the proportion of the population in (the 5 to 24 age group, which in the past has reflected largely the baby boom phenomenon, will result in an increase in demand and upward pressure on hamburger beef prices. Conversely, the 25 to 64 age group are lower consumers of ground beef and have as a group decreased since 1950, the first observa- tion point. Hence, an increase in the proportion of this group reduces demand for hamburger beef and places downward pressure on hamburger beef prices. The effect of these age groups on table beef demand may be less clear because fed beef provide table cuts and a proportion of ground beef. Also, table beef is eaten at restaurants and some fast food outlets, by all ages of the population. Notwithstanding this, the expected sign is negative for the 5 to 24 group and positive for the 25 to 64 group on the grounds that the latter group are the heaviest consu- mers of table beef and much lighter consumers of other beef. 214 The expected signs for pork and broilers is even less clear. Pork consumption appears neither age specific nor eaten predominantly away from the home. Chicken consumption is also eaten by all age groups, although in recent years, as with ground beef, chicken parts are eaten mainly by younger consumers and at fast food outlets. By contrast, whole chicken and roast pork is consumed mainly by older segments of the population and at home. On balance, the signs on age variable coeffip cients for pork are expected to be the same as for table beef and signs for chicken are expected to be the same as for hamburger beef. No data are available on the quantity of ground beef or table beef retailed through fast food restaurants or of the total quantities eaten away from home. The only readily available proxy for this data is aggre- gate away from home expenditure on food per person.uu This data series includes all food expenditures, excluding only alcoholic beverages. Individual meats are a relatively small component of this total expendi- ture and therefore their effects on retail prices is likely to be masked. An alternative specification of this upward trend in away from home food expenditure is the ratio of that expenditure to total food expenditure. Intuitive reasoning would suggest a positive influence of away- from-home food expenditure on demand in the case of convenience-type meats, hamburger and to a lesser extent, chicken. As expenditure on food away from home increases, demand for the products increase, putting See various issues of USDA, National Food Review. 215 upward pressure on hamburger beef and broiler prices. A negative influ- ence is hypothesized on table beef and on pork although the evidence is unclear. As the share of expenditure away from home increases the demand for beef and pork roasts and certain steaks, more commonly eaten at home, will fall leading to a downward pressure on prices. On the other hand, certain other table cuts are a common item on many restau- rant menus. As predicted, results of the above analysis were mixed. For this reason, selected results are presented only for the beef equations, and then, only those for the age variables. Coefficients on the ‘away- from-home‘ variables had the expected positive sign in the hamburger and broiler equations and negative signs in the beef and pork equations though insignificant. Multicolliniarity with real disposable income caused some of these problems. The retail price of hamburger beef may not be appropriate when considering the influences of away-from-home purchases of ground beef foods. Counter-intuitive signs were obtained on the age variables in the pork and broiler equations and coefficients were not highly significant. The age variables in the hamburger beef equation were significant and had the expected signs (Table 4.7). A one percent increase in the proportion of the population aged 5 to 24 years will result in a 1.42 percent increase in retail hamburger beef prices, ceteris paribus. The increased demand by this young, heavy ground beef consuming group places upward pressure on hamburger beef prices. A one percent growth in the population of older consumers will result in a 1.24 percent decrease in retail prices, ceteris paribus. A major factor in the ETOch in the 2113 .Ho>oa ucouuoo n any as sou-suocoocu aw coausaouuooouss no uauu ego .eosuse Add new .Iooeouu we consume new seasons. .coauscwshouov 0H nonunuusuo socus31cwnusa I .3.9 mousse co unusuaccooo - as «.0902 I. «can someones: s~.~- ~s.a so on nm «N on n “coauauanawunm ou< . Awq.mv Acm.av Aqo.nv AoH.v Awn.nv nqn.¢v Aeo.nv . as a on. mac. nae. ”an.s1 as~.~- ~s¢.- mos.s- can.a- Noo.~na amen hostages: a . Aoa.o. Aso.~o Ame.ss Ass.v Ass.oc Aan.no Asn.v . on a an. moo. mac. mnq.~ moo.a1 eno.1 o~m.~1 omo.m1 o-.- noon nonsense: c . Aon.no Aom.o Ams.nv Aas.s Aa~.As Ass.so Ann.ns . on H c». one. moo. now.1 ~o¢.~1 o-.1 on~.H1 o~n.u1 ~qo.~n~ use: «finch n . .oh.ss Ann.ss Ass.ss. Aom.. Aaa.~. Aoo.mv Anm.~o an a nu. moo. coo. «an.n ~a~.u1 ooo.1 hon.~1 coo.H1 ~a~.mo moon can-h .- «on: co so on em on n Queue arson uzxmoa umnuzon Quechua acquacou ouaum «aqua: oncogene .3.9 mm. axe cocoon ucosuom ou< common use coauossasoo uauaoson aouaauus> auouosafioxm o~osuue> ucoocooon N¢¢~1Oaan ”nausea 0:» uo>o mauoiwunm n.¢ enemh "coouum was: nqouou uo susuosuum ou< uo oucosaucu oeu mo moumsfiuom 217 proportion of this younger segment has been the U.S. baby boom. This might suggest that as these younger consumers grow older and move into (older age cohorts, characterized by low hamburger beef consumption, demand for ground beef will decline, causing a decline in hamburger beef prices. However, this assumes constancy of age based meat consumption patterns. Militating against this is the degree to which the young con- sumers of ground beef carry or maintain their tastes and preferences into older age. Extrapolation of these trends, therefore, must be con- sidered cautiously since habits and tastes and preferences of consumers can be expected to change through time. The coefficient signs on the age variables in the table beef equa- tion, contrary to the initial hypothesis, are the same as in the ham- burger beef price equation. This possible outcome was mooted earlier. Even though older groups are considered heavier consumers of table beef, fed beef provides some 25 percent of ground beef. Also, as the younger Broup expanded over the past 33 Years, so did the demand for beef, caus- ing prices to rise. Note, however, that the impact of this age variable on table beef prices is substantially less than it is on hamburger beef prices. Also, the coefficient on the 25 to 64 age variable, though negative, is not significant.”5 5These equations were tested over a more recent period, 1960- 1982, although the equations were generally inferior. It may be that young people eat more of both types of beef provided their parents have adequate income. Similarly, poorer consumers over 25 years of age may be heavy consumers of hamburger beef. This indi- cates the need for a more detailed analysis of these factors. 218 4.5 Examination of Structural Shifts in Demand 4.5.1 Graphical Analysis Graphical analysis provides the first indication that the underly- ing structure of demand for a particular commodity is changing. By plotting a series of price—quantity points, valuable insight may be gained about these data and the apparent movements of observation points along and shifts in a demand relation. General reasons for some of these changes in demand structure were discussed in previous sections of this chapter. Many of the specific causes of demand structure change in the meat markets were discussed in the previous chapters. However, as with other means of quantifying structural shifts, graphical analysis identifies approximately when certain shifts occurred but not why they occurred. Plots of retail prices and quantities for table beef, hamburger beef, pork and broilers are illustrated in Figures 4.1 through 4.4. A cursory examination of these plots will reveal the changes in these price-quantity relationships which have occurred over time. Fitting freehand drawn linear demand curves through particular sets of data highlights periods when price-quantity movements along the curve as dis- ' tinct from shifts in demand. These demand curves for table beef, broilers and to a lesser extent hamburger beef, appeared steeper in the 70' earlier years of the 19505 but became increasingly flatter through the 19603. This trend was not as apparent for pork demand. In the case of table beef, between 1968 and 1973, demand shifted outward. Retail prices of all four products rose sharply over this 2159 moom.o~nma mo cowuoesmcoo owumolon one condom durum“ no uon 1 H.e musmam Aunwuwa Hamuouv common use common . . nvp.nn «on..« nun.pm nv_.no «so ow «no.cv «co 0. unu.on noo.ou o.o.o« ’00..0 Daumuuc 1111111111 1 11111 NMH11111 11111111 111 11111 1 111111111111111111 one 1 . mm 1 NQ N100 €00 5%.. we. .5 00. mm. mm . cm 1 1 ON. on. o no. 1 as. a? Ho . No. 00. who 1 Nb who CW. 05 e Om a 1 ch. 1 $5. 1 ms on. wm. an . - nch.po nhu.nl no..0¢ «an... «00.00 N.Q.N0 «n0.na «nfi.ha —QC.OO .no..0— p00.00. .50100. wou.00. 000.0wo oo~.«.. OuO.Q-v Bacon (1931) punod sad sauag 22C) moon nonsense: .u0.nv cummmZUO 11111 ms. ascends: stumuv common use season Na. Nn. co condo-sauce cacao-co was mousse season so some 1 ~.a ensues Nu. on. Ms. . .00 '05..“ —00.-Q «00.00 “'0 F” N.w.06 6N0 000.09 ......... aw“-.-::.:. an. Rhea. no 0%. saw no. as. . no. an. some. as. no. as. an. an. an. on. an. 1 «no.5v $00.0. onc.«n cwh.vn CON.hn 5.0.00 Ohfl.flc nflu.v¢ Ohw.hv Ou0.0h «on.uh nap.nh 00W.hh .QN.OO '05.«¢ 5.0.”. 301168 (teas) punod sad saucy soon no sequelsscou oaumoson use moowwm Mucus“ no Onn.00 060.00 uuxnuo 11111111111111 cm. an. 11 04 0‘ mm. an. some 1 n.e «cause Auzmwoa Humuouv msomuoa use common 000.00 500.N0 . . . ovr.v0 00..u0 who on mum.hn (Oh no 000.10 up" «n :3“: .......... . ...... . ................... . .......... . .............................. #0. No. nc. on. 00. an. no. No. #0. Oh. Nho mm C“. a 50. mo. 55. no. uh. oh. an.ma cm. mm. 00. on. on. ms. mu. 1 hv0.nn 'Oh.bn .No.on hm...0 C0fl.fl¢ unp.no 000.00 -V0.06 Ohv.Nh ncn.ch Nn¢.0r 00¢.hh ano.fih «an... 00'.00 Bxcau (tear) punod Jed cacao 222 muonwoun mo sowuolsnsoo swans-on was moouum Hamuom no uon 1 ¢.q owswam auewwo: Husuouv common use mossom 000.0! OFN.NH 000.00 0.0.00 006.0v rpm..v v0..on «no.vn uuooo.1 ...................................... 1 .............. . ................................. 2 . ........... S. 8. 2. 2. 2 . 2.. 2. on. on. as. 2.9? 8:3 3. S. s. so no no. no 8.. an an. on. mm. 505.?" 000.00 000.Nn 000.00 000.00 000.0. nov.0' 00..00 .Nh.nn .10.60 000.00 060.10 ....ph .nv.mh 600.0b 00000 (Icon) punod 13d asuaa 223 period. Eventually, high prices caused demand to shift back to the left.”6 This shift ushered in a period of volatile quantity—price rela- tionships, associated with a general fall in real prices and a decline in consumption per person from the high levels of 1972-73. Declines in consumption stabilized during 1980-82, although real prices continued to fall. It was during the period of the late 19703 and early 1980s that shifts in both supply and demand occurred and it is the latter part of this period that has caused forecasters the greatest problems. Hamburger beef followed a rather different path to that character- izing table beef. Formost of the 1950s, consumption of beef from non- fed animals exceeded consumption of fed beef (Simpson and Farris, 1982, P. 39). The majority of 0.8. cattle were still produced on grass and a substantial portion of table beef came from grass-fed cattle. During the 1960s and early 19703 fed-beef production grew rapidly and non-fed beef filled largely a residual role in consumption. After this period, with the beginning of high table beef prices, consumers turned increas- ingly to its cheaper substitute, hamburger beef. As Figure 4.2 illus- trates, between 1973 and 1977 consumption of hamburger beef grew rapidly, then remained at the high levels of the 19503. However, as the general demand for beef declined and fed-beef prices fell, consumption of hamburger beef returned to the low levels of the early 19705, and, since 1980, consumption per person has remained relatively stable, although real prices have continued to decline. "GDiscussions in Chapter 2 cover further details of these price- quantity movements. 224 In contrast to beef during the 19503 through to the early 19703, pork has not fared as well. For much of this period the demand for pork has tended to shift downward. More recently, in the wake of the malaise in beef consumption, pork consumption has increased. 1 A further contrast to both beef and pork products is the path of domestic consumption of broilers per person which has been one of steady growth and falling real prices. As shown in Figure 4.4, the retail broiler demand has tended to become flatter over time. These general observations from the graphs are supported empiri- cally in Table 4.8 in which are presented the four base model equations with the addition of a time-quantity interaction variable (T'Q) in each equation. T is time and Q is the respective own-quantity. In the sim- plest fashion, this variable allows the slope coefficient on own- quantity to vary as a continuous linear function of time. The signs on the T'Q coefficients are positive for table beef, hamburger beef and broilers, although the coefficient in the table beef equation was not significantly different from zero.“7 The coefficient sign in the pork equation was negative and also insignificant. These signs agree with visual inspection of the plots. In the case of the hamburger beef and broiler price equations, the own-quantity slope is becoming signifi- cantly flatter over the 33 year period.’48 For table beef and pork, “YIn the ‘all beef‘ equation, not shown, the coefficient on T'Q was also positive and insignificant. The substantial change of the 19708 and early 19803 offset the steady growth trend in the previous period. 8In a price equation the smaller the slope coefficient the flatter the curve. 225 demand may be tending toward a flatter and a steeper position, respec- tively, although not significantly. From these observations from Table 4.8, it can be concluded that at least in the case of hamburger beef and broilers, if the quantity-price ratio in each year remains constant, then, given their declining slope coefficients, the direct flexibility is declining. This conclusion is contrary to the arguments presented earlier: that of a tendency towards increasing flexibility over time. For this to occur, domestic supply per person will need to have increased at a faster rate than real retail prices and the slope over this period. To determine the direction of change in direct flexibilities in each commodity over time, evidence on trends in consumption and prices were combined with information on changes in the slope coefficients con- tained in Table 4.8. This information on changes in the components of their respective direct flexibilities is presented in Table 4.9 together with the expected change in flexibilities over time. From this informa- tion it is possible to hypothesize that direct flexibilities for table beef and broilers have increased, those for hamburger beef have decreased and in the case of pork no significant change is expected. Given earlier discussions on interpreting flexibilities in demand relations, these directions of change are of considerable interest. The lower the direct flexibility, the greater the extent to which increases in quantities available per person are absorbed by demand and hence the 2526 .eoqunuucue amassed uo couuouuueoo uou n.c «Hash sum "was: .oouasuocoocu a scan annoouo-H 0N aH I H. “unseemuuooo suaucmercso sequuooaou as» on eaquuovw ow sous) no ucuwouuuoco 0: u .oaosausa couuoououcu huuucwso 63°10luu s cm cah- ANe.e. Ace.eo Aun.e~. Aea.ao Asa.ev Aan.~s Aan.os ma.~ as. see. «no. nNo. a~a.~1 «no.1 nme.1 -n.1 nan.~e Assumes nauseous Ase.es .ene.s Aun.~o “on.no Ao~.mc “an.no “Ho.av Am.a an. nee. eeo. ooo.1 men.~1 emo.a1 s~n.1 nan.1 nae.oe Aauaeee seem . . may “Ne.n. Ane.us Ane.ms Ace.v Ans.oV Ase es new no heat: 00% Hmofl NQ. KOO. NNO. Mao. Owfioni "an... QQNeNI ORG-l fimoNMd HOOD Homhafififlv— Ase.aV “ca.o Ano.ec Aen.~s An~.ao Ae~.eo Ace.ns ees sn.a se. see. one. goo. Han.a1 "on.1 nee.~1 eon.a1 one.ena Aeneas. coon «seep «one each ounce arson cannon oaxeooo oaseazoo oeeeaco ensue fiancee easeseea .III acne-coo o.3.n nm. axe osoosn common use couuossocoo ovumoloa eo—neuue> mucuschowu egowuus> assesses: monaowus> couuuuuoucm auuucms010suh "panama one: «wagon no utdaUuuuooo odoam so weak mo museum 0.0 munch 227 .moaoofi>m o>onm one Boom omwamo£u0damo .mumo moauo meuou ammo ou osfiH compo amused m wswuuam an oocfimunoo .mumo huwusmsUImao ou mafia osouu umosafl m wswuufim an mocwmunoe .ucmoamqswwm no: I .m.clm m.q manna Boom omcfimunom ommmmuosfi mmcmco o: ommmowomo ommmowosfi voosuficme musaomnmv huwawoaxoam ow mmmmso monommxm 2.5 3.5 one m>aummms m>wumwoo m>wumwmc o>fiumwmc ommoauo Hamumu Hamm A.m.sv A.m.sv snowmen use m>wuamoo m>aumwms m>aumwmc m>fiuamoo maamfiwm>m mofiuwuamso one 3.5 o>Huwmoe m>Hummmo m>wuamoa m>auamoo musmuofimmmoo macaw who on no «com mm o m H« m M m umwusnamm w m an H "aw mwsmno mo soauomufia uosooum umoz mowuaawnawmam acumen mo mucosoaaoo cw newsman Hmsuo< mo sowuqmomaooma scum oosamuno moHuaHHAHXmHm uomufin ow mwsmno mo coauomoan omuomoxm m.o anmH 228 smaller the adjustment of price to clear the market. This could mean that over time, given supplies made available to consumers, that quanti- ties of hamburger have been more readily absorbed by demand than .have table beef or poultry meats. Relatively speaking, the preference for hamburger beef has been growing vis-a-vis table beef and poultry meats. 0n the other hand the position of pork demand has remained unchanged. A simple way of testing this hypothesis is to estimate flexibili- ties, using the base model in Table 4.3, over the first half and compare them with those of the second half of the data period. Flexibilities estimated over these two periods are presented in Table 4.10. As hypothesized, the direct-quantity flexibilities of table beef and broilers increased over the two periods, the flexibility for hamburger beef declined, and that for pork remained approximately the same. An alternative test of this hypothesis is to derive flexibilities directly from equations in Table 4.8, by plugging in incremental values for time, T. For example, in the hamburger beef price equation direct flexibility can be written as f = {-2.244 + .031T)* (q/p), for T = 1,2,...,33. These calculations, contained Table 4.11, can only be taken as indicative of the direction of change and of the relative magnitude of the change. Table beef and in particular, broiler direct flexibili- ties show a distinctive rising trend over time. Flexibilities for ham- burger beef show a downward trend although subject to {considerable annual change. Pork, by contrast, has revealed considerable variability over this period, switching frequently between flexibility and inflexi- bility of demand. 229 «N.N on. ca.1 on.1 N~.1 NN.1 o~.1 No.1 no.N1 ow.1 coon NN< N N N N mousse vowuom mousse mousse o¢A>xaun cannon em.n no.u oe.N1 an.1 .1 No.1 nn.1 on.1 mn.~1 on.1 ouoNaoon oo.N on.~ no.1 o¢.1 no.N1 No.N1 «N.1 0N.1 mn.1 NN.1 xuom SN 34 8.1 3.1 .1 2.1 8.1 3.1 3.1 «21 teen nonsense: no.~ cm.~ no.1 nn.1 NN.1 NN.1 nn.1 No.1 on.1 no.1 mean canoe N a N N N N N a N N N N N N coupon coauoo pounce mousse mousse mousse ooauom weapon mousse vo«uom acumen possum mousse mousse cease Nasusm «one venom oxen: Oaxaca ozzzouo cannula: Quechua coucfiuoa assoc” commas non soNuoISacoo owueuaon «woqlhwon 92¢ oood1amod "accuxum cab nose vouosuumu mouuuuunwxouh season one auuuomsa Namuom ommuo>< 33 user 230 It is not possible to understand the cause of every change in direction in these flexibilities, even if these estimates were identical to actual movements in direct flexibilities. Even when patterns of change have been recognized, interpretation or the relating of these to actual phenomena have to be made with caution. With this caveat, it may be instructive to attempt to relate, in a general way, observations of economic change in the meat markets to the estimated flexibilities contained in Table 4.11. The purpose is to establish the existence of any structural relationship: to simplify the discussions the more recent period of the late 19603 to date is exam- ined. For a considerable period from the 19503 to the early 19703 the market for beef, especially table beef has been characterized by strong demand, expanding supplies and growing consumption per person. Real prices were kept low over this period by low inflation rates, relatively cheap inputs and scale economies in feed-lot production. With no signi- ficant change in the quantity price relationship (slope) over this time (assumed from Table 4.8) and growth in quantities available per person exceeding growth in real prices, the own-quantity or direct flexibility should increase, which indeed it did up to 1972 (Table 4.11). The peak in beef supply did not occur until several years later although one would expect the retail market and consumer response to lead the changes in the direction of beef production. The flexibilities in Table 4.11 for the years after 1972 first decline and then fluctuate to 1982. The logic of these changes is less clear, except to recognize that both demand and supply appear to have shifted over this period (see Figure 231 Table 9111 Direct Flexibilities for Retail Meat Demand: Estimation Used T*Q Interaction Variable Direct Flexibilities Year Table Beef Hamburger Beef Pork Broilers 1950 -.269 -.822 -.916 -.308 1951 -.259 -.573 -.960 -.377 1952 -.291 -.687 -1.023 -.627 1953 -.414 -l.310 -.650 -.461 1934 -.373 -l.625 -.769 -.570 1953 -.398 -l.63l -l.011 -.548 1956 -.427 -l.727 -l.083 -.794 1957 -.404 ~ -1.596 -.886 -.925 1958 -.373 -l.136 -.843 -1.107 1939 -o“3 -0936 “10086 -1025. 1960 -.b76 -1.022 -l.073 -1.285 1961 -.524 -1.018 -.896 -l.588 1962 -.517 -.988 -l.032 -1.500 1963 -.600 -l.000 -l.l40 -1.616 1964 -.667 -l.076 «1.170 -1.691 1965 -.642 -1.050 -.911 -l.765 1966 -.703 -.978 -.832 -1.840 1967 -.776 -.927 ~1.043 -2.053 1968 -.825 -.892 -1.118 -2.054 1969 -.820 -.778 -l.063 -2.150 1970 -.889 -.732 -l.104 -2.473 1971 -.868 -.718 -1.408 -2.509 1972 -.869 -.622 -1.135 -2.657 1973 -.727 -.A73 -.837 -1.889 1974 -.722 -.719 -1.017 -2.210 1975 -.614 01.156 -.733 -2.089 1976 -.818 -1.129 -.830 -2.472 1977 -.880 -1.110 -.986 -2.709 1978 -.824 -.706 -.934 -2.779 1979 -.689 -.4A8 -l.187 -3.273 1980 -.696 -.523 -1.499 -3.415 1981 -.750 -.602 -l.446 ~3.758 1982 -.824 -.597 -1.204 -4.133 232 u.1). In the last few years, quantities available have remained at relatively stable levels and prices have fallen. If the relationship between quantities and prices remained constant as assumed, then as shown in Table ”.11, the table beef flexibilities will have risen. This means that demand shifted leftward in a parallel fashion. As fed—beef production expanded in response to a growing table beef demand, hamburger beef supply availability and, hence, consumption fell from the mid-19605 to a low in per person consumption in 1973 (Figure u.2). With this decline in hamburger beef supply, quantities available to consumers moved more in line with effective demand, and consequently, retail price adjustments became increasingly smaller. In turn, direct flexibilities of demand for hamburger beef became increasingly smaller, in absolute terms. As indicated in Table “.11 this decline continued until 1973. In effect, the demand curve for hamburger beef was shifting leftward and becoming flatter (Figure u.2). However, by 1973 table beef prices were at high levels and consu- mers turned steadily to ground beef and other meat substitutes. For a period, demand for hamburger beef rose (demand curve shifted out) and hamburger beef direct flexibilities rose to higher levels over 197u.77. Over time table beef prices began to fall, hamburger beef again faced increased competition and demand declined. The opportunity cost of ham- burger beef supplies from cows and heifers rose as the demand for breed- ing stock increased. A greater portion of fed beef carcasses was cut out as table beef. As these reduced quantities of hamburger beef could be more readily absorbed in demand, than as reflected in Table ”.11, direct flexibilities of hamburger beef also fell. 233 u.5.2 Slope and Intercept Changes in Demand From the foregoing discussion, considerable change in the structure of U.S. demand fbr meats, especially table and hamburger beef and broilers, appeared to have occurred over the past several decades. Thus far, through the use of time-quantity interaction terms, namely T’Q. changes in slope have been measured as an average linear change over the total 33 year period. However, as may be seen in Figures u.1-fl.fl, over certain periods within this total period the slopes appear relatively stable, while over other subperiods substantial change in slope has occurred. In an attempt to measure these changes, slope and intercept dummy variables were tested separately and within the same equation. The choice of periods over which these variables were to operate was somewhat arbitrary, although based upon the plots in Figures u.1-u.u. While some experiment was necessary, examination of groupings of price- quantity data does reveal periods when these prices and quantities appeared to move along the same broad demand curve as opposed to periods when a shift in demand took place. For each of the four meats being examined, the following subperiods were chosen. 234 Table Beef: 1950-1957 Hamburger Beef: 1950-1958 1958-1965 1959-1966 1966-1668 1967-1972 1969-1982 1973-1976 1977-1982 Pork: 1950-1954 Broilers: 1950-196fl 1955-1959 1965-1982 1960-1965 1966-1979 1980-1982 It is recognized that for table beef the period 1969-82 could not adequately be modelled as a homogeneous period. The latter half of this period was particularly volatile in terms of demand and supply shifts. It is partly for this reason that it is separated from the other periods. Also, in order to keep the number of subperiods within manage- able limits and within the restrictions on degrees of freedom in estima- tion, this grouping was chosen. A further difficulty arises from the inclusion of both slope and intercept changes. When both variables were included in the same equation, except the broiler equation, multicol- linearity prevented efficient estimation of the coefficients. There- fore, the intercept dummy variables were excluded in most cases. The bases for this choice were first, that the primary interest is on how the structural relationship between prices and quantities (slope) changes over time, and second, that the interpretation associated with the value of the intercept is somewhat doubtful since no data points have been observed at or near zero quantities. The remaining discussion focuses primarily on changes in slope coefficients of specific demand relationships. 235 Results of the addition of slope interaction variables are summar- ized in Table “.12. The slope of the demand for table beef has become significantly flatter in each subperiod.“9 This change has been associ- ated with a rightward shift of the curve and substantial increases in supply availability. The slope of the demand curve fer hamburger beef has also become flatter in each subperiod, though significantly only in the fourth period, 1973-76. This period covers the years when demand for hamburger beef expanded as consumers substituted out of high priced table beef. Consistent with earlier results, the slope of pork demand has become steeper over time. In this model, these changes were statisti- cally significant at or above the 90 percent level for all but the fourth subperiod, 1966-79. Only one shift in the demand for broilers was measured, namely, that during the mid-19603. This was the time of an important watershed fbr the broiler industry in terms of diminishing gains in labor productivity, output per farm and feed efficiency (Reimund, Martin and Moore, 1981). The first broiler equation in Table “.12 indicates that the demand curve became flatter over the latter period. Because there was only one shift it was statistically possible to specify a slope and intercept variable as shown in the second broiler equation.; This equation suggests that in this later period the demand curve rotated to give a lower intercept and a flatter curve. u 9Changes in slope are to be measured with respect to slope in the first period, for example, as represented by the coefficient on DCFBFRC. 236 .auuaquuoeeeu .eeaaeuue> sequueuuucu use haisu :uon euusqucu sequence saga .n.¢ ouaek 00m ”mayo: 5 .euaeuue> Anise ecu no reason ecu ou mucus» weaasc any use .cOuuesce some cu au«ucesulnso eauuueeeou esu eucoeeumeu o ewes) .eunaeuue> caduceueucw huuucesvneao one e0~asaus> c>o ache Amn.n. Ae~.ov Asn.nv Aoo.-v Age.~v Ae~.mv Ame.nv Aas.ov "TN 2.. 08. as. new... aflaé? 2: SN. 93.: «3.. :93 32.. undo; a-.oc Aoo.~v .on.n~v Au~.V An~.ev Anh.~v “an.nv 8; 2... «8. «8. n2. o3.~ 23.- ”2.- «8.- 03.2 32.. 2:85 Ame.av .sn.av Aan.. Am~.~V Aoa.sv A~«.nv Ae~.ov th.~v Aon.~v Aon.~. so." 8. 35. a8. a3... 95.. 22.- 08.- 2a.: 25.? 1e... «3.. "$4.: 03:. .28 on. Aan.~v ANn.nv Amn.nv n-.v asm.v Aoa.~v Amo.v Amo.aqv Aun.~v Anh.ev 34 on. So. 3:. n2. 2.... no". «no. .12.- 2o... 3a.? mi... «:6: 32.. «3.. 533.5: Acm.ov Aoo.cv Aoo.nv Aoc.ev Ano.nv Aen.. A~n.cuv Amu.ov nnm.sv 2.“ S. n8. n8. ...3. s3. 3“. St? «8.- 8a.? 2a.? 89%: our; t3 «2.: 53 32. 32. 35 ~96 can”! 2E8 ous—moo 05:58 82sz canal—8 our... :33. 3:53 umeumcou .3.n m axe eaoucn scenes was scuueasecoo cayenne: e eeuaeuue> mucueseflnwu wanmwue> ucovcuaea ease: you unseen dweueu cu newsman unuuuuucn was enema -.¢ o~Amh 237 It should be noted that the income flexibilities derived from the equation in Table ”.12 and calculated at mean values were 1.65, .99. 1.13 and 1.91 for table beef, hamburger beef, pork and broilers (second equation), respectively. The relative magnitude of these values agrees with a priori expectations. 4.5.3 Measurement of Irreversibility of Demand There are researchers who have theorized that these shifts in demand structure are more systematic and follow distinct cycles and phases. The existence of consumption habits as well as cyclical pat- terns of prices and consumption has led to some investigations into beef demand and of the hypothesis that coefficients in demand models are dif- ferent for rising and falling phases of consumption, prices and incomes. This is the hypothesis of irreversibility in demand, discussed in Chapter 3.50 In a quantity dependent demand relation the usual hypothesis is that the quantity demanded by consumers, following a change in prices, is more elastic in an increasing consumption (decreasing prices) phase and more inelastic in a decreasing consumption (increasing prices) phase. This is consistent with the hypothesis that because of habits in consumption patterns consumers adjust consumption by less during the increasing price phase than during the decreasing price phase. Also hypothesized within this model is that income elasticities are higher I See Chapter 3 for a review of literature in which this hy- pothesis has been tested. 238 (lower) in periods of decreasing (increasing) prices because of a stronger response to income changes in these periods. There have been several analyses of demand testing these hypotheses in terms of quantity dependent demand equations, dealing with measures of elasticity. However, at the time of writing no studies have been conducted on the demand for meats when price is specified as the depen- dent variable and quantities are assumed predetermined. In this model specification, it is hypothesized that own-quantity or direct flexibil- ity will tend to be more flexible during decreasing consumption phases and less flexible (inflexible) during increasing consumption phases. From an initial equilibrium, for a given level of demand, market prices adjust less when available supplies are increasing. Existing consumer habits are reinforced by more product at lower prices and since demand and supply are more closely in balance price adjustments are smaller.51 When available supplies are decreasing relative to a given level of demand, prices by necessity, are more flexible in clearing the market. Another, more market oriented, explanation fOr this apparent demand irreversibility relates to the nature of retail pricing. That is, at the retail level, the average response of retailers over one year given predetermined quantities available is to increase prices when supplies are low but not to decrease prices by a commensurate amount when sup- plies are high. Therefore, there is some degree of downward rigidity but uDward flexibility in the process of pricing at the retail level. 1 5 A given level of demand is quite realistic since demand usually moves less and more slowly over time, compared to supply. This would seem the case especially within the short-run period of a year. 239 The a priori behavior of income flexibilities seems less clear. Two arguments may be made, each giving rise to different conclusions. The first is based on the assumption of habits in consumption patterns and institutional impediments to adjustment. As incomes fall, the demand for a particular good falls. Because of the persistence of con- sumption patterns associated with a given level of income, a fall in income is not associated with as large a reduction in demand as the rise in demand resulting from a rise in income. When incomes rise, existing consumption patterns are reinforced and hence prices tend to be more flexible in response to upward movements in income. The alternative argument is based less on individual consumer behavior and more on market behavior. During periods when incomes are low or declining, retailers pay greater attention to competing for the consumers' dollars. The effect of an increasing budgetary constraint on consumer spending is, from the retailers' standpoint, the more frequent offer of specials and greater price discounts, and from the consumer's standpoint, a greater sensitivity in terms of their willingness to pay a certain price for a particular good. Hence, in response to downturns in consumers' incomes a greater direct flexibility can be expected. Con- versely, when incomes are rising consumer demand increases, but as retailers would find, consumers are less sensitive to price increases, since their overall budgetary constraint is less apparent. No theory is available to choose between these alternative hypothesis, hence it becomes an empirical question to be investigated. Irreversibility of demand is usually tested in terms of increasing and decreasing phases of consumption even in quantity-dependent demand 240 models when changes in prices seem the more logical reference point. Logically, the irreversibility should be with respect to the variable to which the consumer is responding, i.e., the particular explanatory vari- able, and not the dependent variable which is the outcome of the causes of irreversibility.52 Given this causality a number of possible explana- tory variables may be used to test the irreversibility hypothesis in beef demand.53 These variables include consumption, production,e and cycles in cattle numbers. In each model tested domestic consumption per person, retail, remained the actual explanatory variable. When changes in aggregate production and in cattle inventory were chosen as the underlying cause of demand irreversibility, changes in these variables were only reflected in domestic consumption in choosing cut off periods and were not specified directly. Therefore, measures of change in flex— ibility remain in terms of retail prices and quantities. Goodwin et.al. (1968) used consumption changes to test irreversi- bility. Uvacek (1968) argued that the irreversible nature of demand curves was linked to cycles in cattle inventories. During supply reduc- ing periods demand curves are more inelastic (flexible). He felt that the high positive correlation between cattle cycles and slaughter levels made possible the link with consumption shifts. Various demand models were tested for all beef, table beef and ham- burger beef. The procedure was to specify an interaction dummy variable A 52The study by Goodwin et.al. (1968) appears to suffer from this flaw in logic. 53This analysis is restricted to beef demand since in an annual model. demand irreversibility is more likely to be identified. 241 on own-quantity, over the period of years, for each increase and each decrease in the chosen causal variable. In most instances this resulted in a large number of interactions (slope) variables. It would not be statistically possible or indeed economically meaningful to specify each and every decrease and increase observed. Therefore, broad groupings of periods were chosen as an increasing or a decreasing period. Subperiods were overlapped at the beginning and ending years in order to achieve some continuity in change of direction and maximize available degrees of freedom. The first set of estimates using direct changes in domestic con- sumption per person were disappointing and at best inconclusive.5u Largely because of the many changes in direction in consumption and the problem of large variances associated with multicollinearity, many of the estimated coefficients were not significantly different from zero. Results of applying changes in aggregate beef production to the identification of quantity interaction variables also proved unsatisfac- tory except for all beef.55 This model estimated in logarithms56 is 5a In the first model tested, one overall slope interaction vari- able was specified that took the value of zero when consumption was increasing and one times consumption in that year when con- sumption was decreasing over the period 1950-82. The sign on the estimated coefficient of this interaction dummy variable was posi- tive, indicating a direct flexibility which is relatively high during periods of consumption decline, hence giving support to the stated hypothesis of irreversibility in demand. 5 5Results may be improved by choosing fed beef and non-fed produc- tion as the underlying variables of change in the table beef and hamburger beef equations. 6The logarithmic model was chosen for convenience in estimation, all other factors being equal. 242 shown in Table “.13. Coefficients may be interpreted as flexibilities. Specification of consumption flexibility was based on eight changes in direction of production observed between 1950 and 1982. Bork consump- tion was excluded from this model since its inclusion resulted in increased variances in other coefficients. Although the differences are quite small, the estimated direct flexibilities agree, except for the last period, with the relative mag- nitude of flexibilities expected under the hypothesis of irreversibility of demand. When supplies of beef are decreasing following the liquida- tion phase of the cattle cycle lower quantities of beef are made avail- able to the consumer. During this time there is an 'accumulated' demand, and prices, being more responsive to supply changes, are more flexible during this period. The opposite occurs when cattle inven- tories are expanding and supplies are increasing. The previously 'pent-up' demand is now satisfied. With supplies moving more into line with demand, price adjustments become increasingly smaller: hence direct flexibilities tend to become lower during this period. Two postscripts may be made to these results. First, these estimated flexibilities measure predominantly short—term phenomena. This follows from annual time series models. Second, the overall hypothesis of irreversibility is expected to be strongest when demand is strong. If an increase in supplies occurs simultaneously with a weaken- ing of demand, implying an adjustment of supply and demand, then the direct flexibilities may not be low as hypothesized. In fact they may be relatively high if the market continues for sometime in disequili- brium. This may partly explain the relatively higher and 'unexpected' .21i3 .n.¢ adamh mom “mono: sass oo~.~- so” . man>o ~m-ehaa as"; mos.~- emu; . ~mn>a senate“ sea oo~.~- so“ 1 oun>a outneas ems; ~»A.a- an“: . nmm>a n~1~has sea coa.~- sea + som>a «k-mnoa an": maa.~- so“: 1 nua>n mm-onoa 2: a: .7 so" + 253 3-32 gnu: us~.~- an"; 1 sins: sn.onos essaoaswosu saunas eusaaaauouu enemasaxoam unease we evsuucuex uuouua no oosuasue: “Mcmuwuuo couwmumoom Mowmeuu oouueuem voueauuem ooueauuem o>wueno¢ mouooouu « «a so u w Amc.mV Aco.cv RAN.HV Amu.nv Au~.nv As¢.Nv Ano.nv nod.~v ama.v Ana.AV $0.93 an; Qo~.~n nso.~ undue moon -< mm.~ ow. 50o. moo. ocu.~ nao.a one. ~uo. ago. due. noo.1 ace. nmn.t uo~c>a sown oun>a nmn>n oun>a mmn>n emu>a amm>n ~mn>o grave cannon Amazuuucwou may . . acoussoo 3 9 mm a\m olooeu eoeuom hem cowuoesecoo ouueeeon no macaquewoa AeeguuueuOA swv seuaeuue> Aboueceuewm euneuue> uconcuooa cameosmoum ween cummouwu< cu someone no women moon uou woolen mucus: cw >u«~«pumuo>otuu nu.¢ o—aeh 244 flexibility during 1979-82 when demand for beef weakened and as a result demand irreversibility appears less in evidence. Also specified in the all beef equation in Table “.13 is a slope dummy on income, DVDICR. This variable is zero for periods of increas- ing income and one times the income level in that year during periods of decreasing income. A single interaction variable to cover all periods of decreasing income was chosen in the interests of preserving degrees of freedom, in lieu of separate dummies for each group of years when income was observed to change. The estimated coefficient on DVDICR is positive and significant suggesting that the flexibility on income is higher during periods when incomes are decreasing. This result agrees with the second proposed income hypothesis that, because of the effect of a tighter budgetary constraint, consumers (and retailers in pricing) are more responsive to decreases in income. As alluded to above and as suggested by Uvacek (1968), changes in the cattle inventory or more generally, cattle cycles may provide the underlying momentum behind irreversibility in demand. This possibility was investigated by identifying upturns and downturns in U.S. beef cat— tle cycles between 1950 and 1982. Seven such periods of change; increase followed by decrease and so on (see bottom of Table “.1“), were investigated. These changes in cattle numbers were used to provide the breakpoints in consumption (slope) interaction variables. The results of this analysis are presented in Table “.1“. Even though a strong positive correlation exists between beef production and cattle cycles, the cattle cycle provides slightly different subperiods to those speci- fied for aggregate beef production. Nevertheless, a very similar .21L5 .n.¢ enamfi 00m "eouoz swag some can; ease Noa.~- nea.~u ~sm.- om~.- so“ + ko>o ~w-a~ms ens: gum; and: an“: os~.~- NNH.~- oom.s msn.- so”; 1 case skunnaa sad sad no” sea hea.~- uHH.Hu sak.- ~s~.u 30H + no>a nk-hoos smug nus: nus: swan soa.au nea.~- osm.- oo~.u ems; s so>n As-nsos sod sou sou soa oq~.un nmu.nu ~o~.t men.n 30H + mo>a nosmnmn and; and; guns swan oo~.~u ama.ao "no.1 ~cn.- gums s ~o>a am-nnon sou sea add sod anH.HI nnn.nt an».1 «05.: so” « Ho>n anucnod coon moon coon soon coon soon coon moon euasanaxoam auooau>en -< ~H< weapons-z swash “as nn< wannabes: eunoh uoouqa no owcugo cauuou ca abandons: uo owcenu »u«n«a«woam uouuaa «o o>uueflo¢ couscous: ensuwswe: «aqueuox veuelwusu haugufiukeab uuewan neuesauem neuoooxm uo sneer . Ano.~v Awn.sv As~.~o “on.~v Aen.Hv Ana.~o A-.~v Ano.flv “so.uv An~.oo Amn.v we a an. ace. moo. can.“ nuo.- oao. «so. hoo.- mac. noo.- mom.u amn.an "an. use» an: . . As~.no Auo.v Ano.nv Aon.~. Aon.uv Asm.ns Ads.v AA~.HV no«.v asn.ov Ana.c «a a as sec. and." o~o.- Hmo. «so. coo.n «so. noo.- nnN.- oqo.u nan.~- can.” «can she . . adh.~v Asa.~. AoH.~v Aon.us Asn.. An~.~v Aoo.~. Am~.v Anu.o Ano.~v Ano.sv Asm.v on ~ as use. saw.“ ~s¢.- ”No. «we. moo. «no. omo.w Hoa.- gnu. auo.u oHo.H- o~m.ss so»; toutasauz . . . “no.nv Ann.~V Ahe.~. Au~.~v Amm.v Aa~.~v Aao.ao Ass.v “on.. “no.~o Asu.n. Aeuuv oo ~ an ace nag." nuo.- use. mac. noo.- «No. use.. nmo.s noc.n «as.- no“.. mom . coon «Hana «3&5 68 33 33 38 33 32. 32 0x8: ogooa demotion 3.1.38 cannon SE :33. acnuecou oeueauoa .36 «m I)... 0.003.. cosmon— uom ooze—assoc 3:038 «anodes? usovceoen e~neuue> zuoueceuowm eonumo oauueo .m.= ma mowcenu co ocean «can now nausea «annex cw zunnanamto>ouun c~.c canes 246 pattern emerges in changes in estimated direct flexibilities; higher in periods of cycle or cattle inventory upturn and lower in cycle down- turns. Again the exception is the last year which for the reasons stated above may be coincident with a weakening demand for meat over that period. The fact that this pattern also emerges for the beef sub- categories, table and hamburger beef may be surprising. More precise estimates may be obtained by using changes in numbers of fed-beef animals and non-fed animals. The problem is how to allocate in any one year the investment component of the herd, namely, breeding cows and bulls and calves which to some extent may be turned out as fed animals or nonfed animals. Finally, note again, the test of irreversibility with respect to income changes, and that the coefficient flexibility on the interaction term of DVDICR suggests increased income flexibility in periods of decreasing incomes. One avenue for further analysis concerns the hypothesis that income and consumption interact systematically during different phases of demand change. This hypothesis was referred to in Chapter 2. “.6 Time-Varyinngwitching Regression Model Thus far all attempts to identify changes in demand structure and to measure those changes through estimated flexibilities have focused on models of discontinuous and disjointed changes in structure. Models estimated first, were of broad changes in slope as illustrated in plots of prices and quantities and, second, were specified in a manner to test the hypothesis of irreversibility in demand relationships. In reality, 247 shifts in demand from one structure to another are likely to be more continuous. Such continuity seems more in accord with the inherent and psychic nature of people and more consistent with the role played by expectations in smoothing abrupt changes in market behavior. One tends to underestimate the continuity of mass economic behavior. In noting this strong continuity in our economic behavior, Fox (1956, p. “18) observed that the structure of demand had undergone a gradual evolution rather than a dramatic upheaval. Discussions now return to the broad changes in demand illustrated by the plots of Figures “.1 to “.“. Results are presented of demand models capturing this continuity of regression coefficients across structural shifts. “.6.1 Linear and Cubic Splines As discussed in Chapter 3, spline functions offer a means of model- ling continuous, though not necessarily smoothly continuous, changes in demand structure. Only cubic splines allow distinct line segments to be joined smoothly at known join points (the assumed points of change). These spline functions have continuity in their second and first deriva- tives. It is this character which gives cubic splines greater reality in economic applications in which a discontinuous shift may be unex- pected, as in shifts in elasticities, flexibilities and in various mar- ginal concepts. The general model applied is a time-varying switching regression model (Poirier, 1976). This model permits slope parameters to change over some periods of time and remain constant over others. And, through 248 the use of splines, the periods of change are modelled in a continuous fashion.57 This model is to be distinguished from continuous time- varying parameter models,- estimated in the next section, which permit the slope parameters to vary in every period. The estimating models are represented below. .Linear_§pline: pit Y01 + Yliqit + YZiqit t + YBiqit(t-tk)Dk (4.9) 3 + 65113::qu1: + Aiyt + et Cubic spline: 2 3 pit ’ Y01 + Yliqitt + YBiqitt + Y41““: t (4-10) - 3 I 3 + YSi qit (t-tk) Dk + 6ji ‘1']. qjt + Ath + ej where the retail price of the ith meat, pit for i=1,...,“ , is a func- tion of the ith meat quantity, and other variables of the composite qit spline in qit; namely time t for t=1,...,33 , ‘Ek over k join points (points of change in demand structure) and dummy variables, Bk = 1 if t ,3 tk’ and 0, otherwise. yOther variables are substitute meat quantities, qjt’ disposable income per person, yt, equation error term, et and regression coefficients, Y, 5, and A, to be. estimated. All variables remain as previouSly defined. In contrast to the specification of most spline fUnctions in time series analyses, the splines in (“.9) and h 57 Models in which change is measured in an abrupt, discontinuous fashion were discussed earlier and include slope-interaction dummy variable models. 249 (“.10) are defined in terms of time, t and with respect to a particular explanatory variable. This is to allow explicitly for a time-varying character in the estimated regression coefficients. Using special notation, the previously estimated base model for each price equation may be written as P = B(X) + e1. Hence, the base model with the composite spline variable, S(X), may be represented as P = 8(X) + 3(X) + e2, where e1, and e2 are error terms. The statistical significance of the contribution made by the addition of a composite spline variable may be tested by following the procedure outlined in Table “.15. The coefficient of partial determination associated with each spline is calculated and its significance tested by comparing resi- dual sums of squares in the spline models with the respective base models as in an analysis of variance table. The F-value is used as the test statistic. Both linear and cubic splines are tested for each meat product and the results of these tests are summarized in Table “.16-“.19. Contained in each table are the respective partial R2 values and the associated F-values. Except for the broiler price model, only equations in arith- metic values (linear functional form) are presented, as these were supe- rior to equations in natural logarithms in terms of signs, statistical significance, and fit to the data. Spline join points are located at points of structural shift in demand identified earlier from plots of 250 TABLE 4.15 Procedure fbr Testing Contribution of Composite Spline Variables Relative to the Base Models of Meat Demand Sum of Degrees of Mean Squares Freedomg Square Residual from B(X) 2e: 27 Residual from B(X)+S(X) Xe: 23 £e§l23 2 Contribution of S(X) Kai-2e: 4 (lei-Xez)/4 2 2 2 2 Partial R (Eel-Ze2)/Ze1 F-value (£ei-Xe§)/4 lei/23 8Numbers of degrees of freedom are illustrative only. 251 .uaocun nu luau «scomuocsm .n .n.v smash oom ak.m an.n~ n.0eoz u:«_am uaaao a:- banana chorus. swan-sou mo «nob .a.n nv.o. we. acumen uaaau c~.. mn._ as. one—am twocas n_o>o_ ase.. o=_.>-m ..oassto o:..>-m a: Hosanna ocsaam so «any nos~e>uu one a nemuues .comuecmeuouon ~emuwem mo aeomouuuoou "moan-sou comma moon o—meh ou comueofiuwoonm mono-ewes unmaue>-olfih mo comuspmwueou "mouoz nnm.~. Anoc.v a~n.~. Ace..~s a-.. Awo.c n_o.u fi_o.. Ame.s Auo.. Aoo.. Ann... so.” am. ”no. a_c. has. om~. - one..- aces. hoc°.- case. oece.- vac. men.m- amm.om as..oo~ ocs.am guano 58 9:8 2.368 2...;sz 5.29-5”. 5.2.1:. 5.2...45 web «to to catalog 2.391% o—nemus> ecu—am ouqnooeou uo «snob Amn.nc Ann... Aka... Ash.k. Ao~.~. Ans..s nom.o Ann.. Ase. . “an.~s .m._ as. use. one. mo~..- mun. - -...- ous.- was. .me.- «Na. osc.n- 4cm.om_ o=._am saunas ...: N. z\w ¢o_a oxaoa oazacoa ozunuzoc .aaqe-hso ~a-e-eso qm~49-e.a e.o octagon «amassed you»; onaewaa> scannw oumuoeeou «o match moon oqneh no~aemue> Auouaeenawm onnemwo> «fiscal—8 mocmgew omasu one have“; we sensuouuucou new ncoquucsm ecu—mm "ecomuoscm oo.ua moon o—peh o—.v o—aeh .2132 .me.n we o:_.>-a ..umsstu - _o>o_ ”no. a. ocauaoacusm .a .ueocqm mm show ascomuocsm .e mo.m ov.v a_ovox onenmm o‘nsu as. been“; a .8238 3:33 we «no... . .A.» on... n». usu.am usage . no." an.n. oh. och—am sauces n~o>o~ 'nc.~ o:—e>-m neumumwu o:—e>-m. ~¢ «saunas ocmnam mo comb nos~e>1m one ~u "nausea .comuucushouoa "nuance mo nueumowuuoou "newnesom new»; homespun: on cowuauam«uomm nouoeeueo ucqxue>-oemh mo conusnmuusou Ann.~. Asa..s Avn.s an... Aqa.. Anm.. ”mo.. aaa.c Asa.s Amm.e noo.. “no.o. Am_.ov mc.~ no. mno. k.c. nan. ems. sec. noo.- mace. mace. occc.- use. one..- mom.ok ”no..- kno.o__ seesaw guano ....a Ne {w 6:. 2:8 oESS ....Tméo 5.3.4:. 5.25:0 5.5.4:. ....o «to to octagon 9:28 2328 nounsmwe> ocflnnm cummoneou mo munch a~o.No Ac»..e aeo.. Ana... Aw~.n. Am~.~. Acm.~s ”no.~. noo.s n_o.¢v Awk.ac no.~ no. ..e. ._e. a... see. as..- m.~.- ova. ~oo. mne.- so..- one..- nah.~n. aca.am gauges .z.n ~a s\w uo_a canoe oazeooa .an.p-hso .aa.e.e.o «aaxe-h.o .:A.+-h.a e.o oxanmzoa pension ocuswcoo costs . . moon weakness: .o_pe«we> new—Am summoneou mo «snob mongemwe> Ahouocaanxm onnewuaa unowcoaoa snowmen ouasu use weenie mo couusawuunou one ecowuocSL enigma "ecouueseu sauna use: mousse-e: n—.v o~aah 253 .mq.n mm o=_a>-g "comuwuu -- _o>o. ’ma. a. u=.u_~¢=uum .n .uuucflq a“ luau ~ncamuuczu .a mo.m Av~.v nocfinam uaasu ecu unocma . gooxuon Acqunscm mo amok "s.n <5.” mm. ocm~nm ounau ma.n .0. an. ocmunm have“; n—o>o~ o—c.v o=_a>-m nauauauu o:—u>-a Na «canyon o=m~nm no onxh uo=~a>au u:- Nx «nausea .:o«uacanuouon manage; mo agomumuuoou "co“uaaem you»; xuoa on noduaumuuuonm nouolauam «chuu>-pa«h an comussuuucou aaa.~v An~.. Anq.u Acv.. Amc.- Ac... ns_.v Am_.v ~o~.. A«~.. ah¢.~. Ana.~v 50.... on.” em. nee. a.o. Nac.- moo. .ooe. noeo.- mace. ~ooo. ”'e.- oov.~ coo.a.- oom.- n...- s._.n‘. oam.am omaau .!a~¢:$.uEQuEBdfifita.P?fito~92fita.Pcfita .fo «fa #5 8:8 BE§8 Sfibaufitgu unnamaa> vaunnm cummcmmou mo nnhoh Ao~.mv as.._. an.._. Aug.“ avm.v anc.~ a_a.. a-.v noo.~. Ann.—V no_.n. .a.. an. eye. nag. o~a.- .ma.- «cc. .Ho. econ. n.c.- cv~.- «...- unn.- o.n.o.~ gum—am “soc“; .x.a ~u xxw. ¢u_o usage .aa.*-s.o .a.¢*-p~a ~=A~w-h.c .aa.h-pva ».c unamua uxunuzua ogunuuo “cuunzou yam“; “he. unnamua> ocm—mw ouunomlou mo munch mopnamua> auOuacaqaxm o~aumua> acoocumoo nocunam uuaau van yucca; no :o«u=a«uu:ou and acouuucab «sauna uncenucacu count aunt an.¢ o—auh ..o>o~ ,mc. a. “c.9‘u.=ufin “oz .. oo.m «no.» ammo. cw“ mocwnnm uwnsu ecu unocma gooxuon >«m_a=cm we each o~.' no.o~ mo. ammo—v ocfinnm umnau hm.“ mo.m. on. Ammo—u oganam hues“; on.c .n.~ on. Hugo=.a. cam—am ufipau Auo>o_ auc.v o:~a>-m nnuaumuu oauc>.u «a .uuuuaa ocfiuam uo unxh no:—u>-m use w: aawauaa .cbwuacqluouoa "canyon mo augoauwwuoou “cofiuuacm cumum uo~m0hn o» :Oguaumumuonm heaving-a u:~>ue>-ui«b mo sawunamuucou ‘25i4 nsm.nv Ash.~. avn.n. ah~.~. na~.~. AH..- Amm.~v ao¢.~. nmv.~. Aom.v .c.~ mm. ~.a. mm.. ~ooo.- booze. n.=.- m... o.«.~_- ~mn.- mv~.- mhn.- -o.~- o=«.gm gmaau Ame.mv A_m.~g Anc.n. goo.“ flq_.~v R~m... n_..~. nc~.~. ha.” co. ~_°. “om.. moo. n_e. - emu. c.n.- on..- msn.- oso.o- oafi_nm hau=m4 Ashen «accuuucsm uuosma.uoa onnson. nun.nv Aom.~ ave..v Aun.~. Aca._v Ana._v nan.~u n.~.nv Ava... nvm.~v nv.~ mm. mac. vuo. wcoc. coco.- osc. o-.m- ~v~.c~_ 50".- oun.- ~m~.- oas.no ucfipnm ugasu Ashen «acbmuucau s-ocmau .:.a ~a zxm «u.o .a.A.*-h.a .an.w-h.o ab.a “hgc ».o usage uzxaua uauagzua uauamuo «caumaou cum»; k..flo»a o~pnwua> o=.~mw uumnontou mo munch mounauua> xuOuacanmxm o.numun> acuvcoaon nonunnm adage can uauawa uo acauaaauunoo as: cacauocah ocuunm ”snug-:9» oouum uo~«0un au.q Quack 255 prices and quantities of each meat, used in section u.5.2 in estimating slope and intercept changes in demand.58 These individual splines, for each line segment, provide constant slope within line segments and con- tinuity across periods or points of structural shift. That is, within periods it is assumed that prices and quantities adjust along a given demand curve. The models presented in Tables 4.16-fl.19 were estimated by OLS with usual regression techniques. Within this framework of least squares a number of hypotheses may be tested. First, the overall test of the con- tribution of the particular spline in explaining variation in retail prices, relative to the base model may be conducted in the way outlined in Table “.15. This F-test is a test of the hypothesis that‘YZ-VY = o 3 in the linear spline and that yz- '6' '74 "Y5 - O in the cubic spline. Acceptance of the null hypothesis is acceptance of no variation in the regression coefficients of the own-meat quantity over time. That is, the classical model of constant coefficients is appropriate. This null hypothesis was accepted in the linear spline table beef model and in both the linear and cubic spline pork models. Everywhere else the hypothesis of constancy in regression coefficients was rejected at the one percent level. Notably consistent with earlier models of structural change, the relationship between pork prices and quantities, i.e., slope of the pork demand curve, has not changed significantly over time. This is not to say that increases and decreases in demand have 58There were fl,5,5 and 2 periods identified in table beef, ham- burger beef, pork and broilers, respectively. This implies 3, u, u, and 1 join points in each case. 256 not occurred, but that the demand for pork has responded almost entirely to the relative prices of substitutes. Second, using the test of equality between sets of coefficients in two linear regressions,59 an F-test of the statistical significance between the linear and cubic spline models may be made. Results of these tests are presented in Tables n.16-fl.19 and indicate the superior- ity of the cubic spline in terms of fit of the actual data in each model of retail meat demand, of course, with the exception of the pork model. Third, as is most usual in regression estimation, the significance of an individual coefficient can be determined by use of the standard t-statistic. This test may be meaningful in the linear spline model. However, in a cubic spline model the spline is a composite variable, coefficients of components of which are not readily interpreted. In this case the previous F-tests provide the relevant information on the nature of parameter variation. Therefore, the procedure of Suits, Mason and Chan (1978), that of providing regression statistics on}2 for the whole regression, D.W. statistics, and on the significance of the par- tial fiz, has been adopted here.60 Examination of the linear spline equations, however, do provide some insight into the nature of changes in demand for beef and broilers. For example, the log-linear spline model of broiler price indicates that the direct flexibility is not constant, that it has been increasing ngor details see Pindyck and Rubinfeld (1979. p. 205) and Chow 1960). 60 See Huang agd Raunikar (1981) where both individual t—statistics and partial R values are used in hypothesis testing. 257 linearly over time (significant coefficient on Q'T), but that the rate of increase has slowed since the mid-19603. This result is consistent with the observations that the consumption of chicken in becoming saturated and that the impact of gains from technology in the broiler industry, which have kept real prices down, may be slowing. This may mean that the competitive advantage in prices enjoyed by chicken over beef will steadily be diluted. Direct flexibilities estimated from price equations in Tables “.16-“. 19 are shown in Table “.20 for table beef and hamburger beef and in Table “.21 for pork and broilers. The linear splines, using a linear functional form, in the beef price equations give larger (and probably overestimated) estimates of direct flexibilities than the cubic spline, which may have under-estimated flexibilities during the 1970s. Also, the range on the linear spline for table beef is unexpectedly large. Consistent with other estimates, nevertheless, is the tendency for table beef flexibilities to rise, especially over the late 19703 and early 19803. This indicates that the table beef market is subject to larger price variations as supply fluctuates. The tendency is for hamburger beef price flexibilities to fall in the latter half of the period. Not- ably, the relatively low flexibility during the early 1970s when ham- burger beef was more price competitive with table beef agrees with' results of Table “.11 in section “.5.1. Very little difference exists between splines in the pork equation largely because they do not make a significant contribution to the explanation of demand for pork. Therefore, the results in Table “.21 are very similar to the model results shown earlier in Table “.11. In 258 Table 4.20 Direct Flexibilities of Table and Hamburger Beef Estimated From Linear and Cubic Splines Type of Spline: Year Linear Cubic Linear Cubic Table Beef Hamburger Beef 1950 - .376 - .383 - .699 - .459 1951 - .358 - .369 - .504 - .347 1952 - .398 - .410 - .624 - .444 1953 - .559 - .571 -l.229 - .889 1954 - .498 - .502 -1.574 -1.142 1955 - .525 - .517 -1.632 -1.177 1956 - .557 - .534 -1.785 -1.268 1957 - .520 - .484 -l.702 -1.184 1958 - .477 - .430 -1.252 - .849 1959 - .667 - .485 -1.065 - .700 1960 - .830 - .497 -1.168 - .764 1961 -1.036 - .521 -1.168 - .758 1962 -1.144 - .488 -1.139 - .730 1963 -1.471 - .536 -1.158 - .730 1964 -1.791 - .559 -l.251 - .771 1965 -l.878 - .502 -1.227 - .735 1966 -2.225 - .506 -l.l49 - .664 1967 -2.596 - .507 -1.094 - .604 1968 -2.907 - .481 -1.037 - .552 1969 -3.040 - .420 - .889 - .451 1970 -3.509 - .394 - .823 - .395 1971 -3.635 - .327 - .792 - .355 1972 -3.849 - .274 - .674 - .280 1973 -3.396 - .189 - .502 - .192 1974 -3.551 - .154 - .803 - .262 1975 -3.171 - .109 -1.359 - .394 1976 -4.419 - .125 -1.395 - .393 1977 -4.968 - .126 -1.442 - .447 1978 -4.853 - .122 -1.068 - .365 1979 -4.229 - .118 - .774 - .303 1980 -4.455 - .148 -1.029 - .450 1981 -5.024 - .209 -1.330 - .629 1982 -5.665 - .298 -1.474 - .717 Note: Flexibilities were derived from linear (arithmetic) equations in Tables 4.16 and 4.17. 259 Table 4.21 Direct Flexibilities of Pork and Broilers Estimated From.Linear and Cubic Splines Type of Spline: Linear Cubic Cubic Linear (in logs) Cubic (in logs) Pork Broilers 1950 - .927 — .984 - .022 -.360 -.834 1951 - .981 -1.018 - .064 -.373 -.810 1952 -l.056 -1.082 - .109 -.386 -.790 1953 - .885 - .903 - .153 -.399 -.774 1954 — .808 - .826 - .227 -.412 -.762 1955 -1.072 -1.105 - .250 -.425 -.754 1956 -1.158 -1.207 - .401 -.438 -.748 1957 - .956 -1.008 - .505 -.451 -.745 1958 - .916 - .976 — .642 —.464 -.744 1959 -1.188 -l.273 — .764 -.477 -.745 1960 -1.185 -l.269 - .808 -.490 -.747 1961 - .988 —l.059 -1.026 -.503 -.750 1962 -1.136 -1.212 - .989 -.516 -.754 1963 —1.253 -1.321 -1.079 -.529 —.757 1964 -l.284 -1.332 -1.143 -.542 -.760 1965 - .999 -l.016 -l.203 -.555 -.762 1966 - .911 - .908 —l.264 -.563 -.763 1967 -1.138 -1.111 -l.421 -.571 -.763 1968 -1.216 -l.l66 -1.430 -.579 -.761 1969 -1.152 -1.087 -1.505 -.587 -.759 1970 -1.194 -1.112 -l.739 -.595 -.757 1971 -l.517 -1.404 -1.773 -.603 -.754 1972 —1.219 -1.130 —1.884 -.611 -.751 1973 - .896 - .839 —l.344 -.619 -.747 1974 -1.087 -l.037 —1.577 -.627 -.744 1975 - .780 - .768 -1.495 —.635 -.741 1976 — .882 - .906 -1.773 -.643 -.739 1977 -1.044 -l.131 -1.947 -.652 -.737 1978 - .986 -1.137 -2.001 -.660 -.736 1979 -l.250 -l.551 -2.360 -.668 -.736 1980 -1.573 -2.118 -2.467 -.676 -.737 1981 -1.577 -2.218 -2.718 -.684 -.739 1982 -1.361 -1.985 -3.010 —.692 -.743 Note: Flexibilities were derived from linear (arithmetic) equations of pork in Table 4.17 and from linear and logarithmic equations of broilers in Table 4.18. 260 the case of the broiler price equation, the cubic spline gave very simi- lar results to those in Table “.11 because of the strong linear effect in both cases. Direct broiler flexibilities have risen strongly over time. The spline function has some important limitations as pointed out by Suits, Mason and Chan (1978). It is most useful when data are uni- formly distributed throughout the observed range and when the scatter of observations is uniformly dense. The broiler data most characterize these aspects. By contrast the table beef data, particularly since the 1970s, does not lend itself as much to the spline function. This absence of uniformity over this period, i.e., 'the thin patches, size- able gaps or isolated points, reduces the discipline (of the function) and the function is free to twist and squirm through the sparse path of the data to yield spurious curvature' (Suits, et al..p. 139). The other major limitation of the spline function arises during attempts to extra- polate beyond the observed range of data. The spline function is not defined outside the range to which it has been fitted, and therefore, as with regression equations in general, some caution is necessary. Mot- withstanding these limitations, the spline function provides a simple means of approximating complicated functions. “.7 Continuous Time-Varying:§arameter Model It should now be clear how potentially misleading it may be to com- pute a single coefficient that purports to portray the response of retail meat prices to consumption and income changes through time or across sections of time. To varying degrees, the underlying structure 261 of economic relationships in each of the major meat markets appears to have undergone considerable change over the past several decades. Such change has been induced by shifts in consumer tastes and preferences and by cycles and adjustments in production. In previous sections demand coefficients were estimated using statistical models which allowed regression coefficients to change a small number of times, and there- fore, allowed for a similar number of changes in structure over that period. In this section an estimation model is presented which permits the structural coefficients to vary more continuously over time. “.7.1 Legendre Polynomials The model employed in the following analysis is based on a particu- lar time-varying parameter model discussed in Chapter 3. The approach involved the integration of the COMPLEX optimization algorithm and Legendre polynomials to find optimal coefficients as continuous func- tions of time, using the weighted least squares criterion. Legendre polynomials, specified into the economic model, provide the means of introducing various degrees of polynomial, with a time dimension, into each coefficient. As noted earlier, the particular advantage of these polynomials over, say, specifying a polynomial into the demand equation, by direct substitution, is the property of finality of coefficients. The addition of a higher degree term in time, does not change the coefficients previ- ously estimated using the lower degree polynomial. Moreover, the same degree polynomial does not need to be set for each explanatory variable. Optimality over time may be attained for some of the E3 coefficients by 262 specifying a linear, quadratic, cubic or higher degree of polynomial while for others the assumption of a constant coefficient value may be justified. Therefore, the B coefficients in each demand model are in essence polynomial functions of time. The role of COMPLEX is to find the coefficient values of the E3functions which will optimize the objec- tive function, i.e., to minimize the weighted regression sum of squares over the time path 0 $_t $_T. The central strength of the COMPLEX-Legendre polynomial method is in estimation of B coefficients which are inherently non-linear in parameters. The present model of retail demand is linear in parameters. This substantially simplifies the estimation procedure adopted in the following analysis and yet does not compromise the property of finality of coefficients which is preserved throughout the transformation. The use of the COMPLEX optimization algorithm is not necessary when inherently linear parameters are being estimated. The time-varying coefficients may therefore be estimated using least squares techniques, applying the standard tests of significance. From equation (3.66) in the previous chapter, a simplified version of the time series regression model of retail demand for the ith meat may be reconstructed as a + e 4.11 Pit YOi +'Ylitzqit j ( ) Initially, let the parameter variation occur only in qit’ the own- quantity of meat consumption per person. Following the general form of equation (3.67). this parameter variation for the ith meat in time, t may be expressed through its coefficient,Y1.1 as 263 Ylit 8 a01 + 311P0(t) + a21P1(t) + ... + aniPn_l(t) n = aOi + kfl akiPn-l(t) (4.12) where Pn—l(t) are the n-l Legendre polynomials defined in (3.62). Sub- stitution of (4.12) into (4.11) gives n 3 P1: ' Y01'+ 301‘+ z akiPn-l(t)qit + 6ji 2 qjt ksl j=1 +rAi yt + et (4.13) n where kfl akiPn-l(t)qit is an interaction variable in qit’ and qjt and yt are the previously defined cross-product and income variables, respectively. Each estimated coefficient, aki becomes a time-varying coefficient, akiPnel(t) and along with the constant coefficients, 611 and Ai’ they may be estimated and tested using standard regression tech- niques. The intercept term is Y and the error term is e . To illustrate the structural form of this demand model, a cubic polynomial in one explanatory variable may be written as Pit a “’6+ ¢1P0(t)qit + ¢2Pl(t)qit + ¢3P2(t)qit 3 . + ¢4P3(t)qit + 5j1 $51 qjt + Aiyt . . (4.14) although the estimating equation is written more simply, 3y‘ pit = ¢o + ¢lzl + d’217‘2 + ¢323 + ¢424'+ 531 jfiiqu. + liyt +‘et (4.15) The regression coefficients presented below are the ¢, 6 and A values. Theoretically, Legendre polynomials may be specified in any or all of the explanatory variables, giving the time—varying character to each 264 coefficient. Calculation of direct-quantity flexibilities, for example, from the cubic model in equation (“.1“). when estimated in linear arith- metic form, is f '[¢P(t)+¢P(t)+¢P(t)+¢P(t)1-q-i£ (416) it 1 0 2 1 3 2 4 3 Pit ' This provides an estimated direct flexibility for the ith meat for each year, t, of the estimation period. Recall that each polynomial is mapped over 0 g_t g_T, so that they are orthogonal with respect to time. The orthogonality is a necessary and sufficient condition for the property of finality of coefficients. Each polynomial in the following estimations, therefore, takes the form I 1" P0 (1:) Pl(t) - -—-- 1 2113.- 2_l P2(t) 2 (32 1) 2 15 2t 3 2_ g;,_ P3(t) = 6 (32 - 1) - 6 (32 l) 105 21: 4 gag _2_£_ 2 ; Pa(t)'—2?(§§'1) ‘ 23 (32 1) +2 945 2t 5 1981 g;,_ 3 32_ g£._ 1’5“" 140 32" 1) " 140 (32 1) + 10 (32 1) for t - 0, 1, ..., 32. Calculated values for each Legendre polynomial up to the 5th degree over the 33 year period, 1950-82, are shown in Table “.22. All equations were estimated in linear arithmetic form using OLS. The initial objective was to permit, with varying degree polynomials, the coefficients of each explanatory variable to vary through time and hence obtain direct and cross-quantity flexibilities for all structural 1265 Table 4.22 Calculated Values of Legendre Polynomials Up to the 5th Degree Year Degree of Polynomial PTl PTZ PT3 PT“ 915 1950 -1.00000 .500000 -1.00000 -1.50000 3.50000 1951 -.937500 .378906 -.653687 -1.53583 3 11u66 1952 -.875000 .265625 -.362305 -1.16979 2.60176 1953 -.812500 .160156 -.122192 -1.32178 2.03086 1951 -.750000 .625000E-01 .7031255-01 -1.11035 1.4u272 1955 -.687500 -.273&38£-01 .218872 -.852665 880075 1956 -.625000 -.109375 .327148 -.564514 373360 1957 —.562500 -.183594 .398804 -.260319 -.55&702£-01 1958 -.500000 -.250000 .437500 .468750E-01 -.392188 1959 -.h37500 -.30859& .416899 .345394 -.629517 1960 -.375000 -.359375 .430664 .62u939 -.766365 1961 -.312500 -.402344 .392956 .876583 -.807043 1962 -.250000 -.“37500 .335938 1.09277 -.760198 1963 -.187500 -.ueuauu .264771 1.26733 -.639510 1965 -.125000 -.48“375 .182617 1.39545 -.460069 1965 -.625000£-01 -.496094 .931396E-Ol 1.47369 -.2ho302 1966 0. -.500000 0. 1.50000 0. 1967 .6250005-01 -.h96094 -.931396E-01 1.u7369 .240302 1968 .125000 -.“8“375 -.182617 1.39545 .460069 1969 .187500 -.46“8“h -.26“771 1.26733 .639540 1970 .250000 -.h37500 -.335938 1.09277 .760L98 1971 .312500 -.4023au -.392“56 .876583 .807013 1972 .375000 -.359375 -.h3066“ .621939 .766365 1973 .137500 -.30859& -.446899 .345394 .629517 1974 .500000 -.250000 -.437500 .4687506-01 .392188 1975 .562500 —.183sgh -.398804 -.260319 .5517025-01 1976 .625000 -.109375 -.327148 -.564514 -.373360 1977 .687500 -.273“38E-01 -.218872 -.852665 -.880075 1978 .750000 .625000E-01 -.703125E-01 -1.11035 -1.uuz72 1979 .812500 .160156 .122192 -1.32178 -2.03086 1980 .875000 .265625 .362305 ~1.46979 -2.60&76 1981 .937500 .378906 .653687 -1.53583 -3.11166 1982 1.00000 .500000 1.00000 -1.50000 -3.50000 1 2 3 4 5 266 variables. This, however, was not possible. Generally speaking, equa- tions were rejected on any of the following somewhat arbitrary grounds. First, the time-varying specification did not make a significant contri- bution to the explanation of the dependent variable, i.e., retail price. Second, the sign of the coefficients do not reconcile with a priori expectations and economic theory. Some exceptions to these grounds of acceptance are made for the purpose of providing comparisons among indi- vidual demand equations or to present results when the null hypothesis of no significant parameter change is accepted. Results of selected equations for five product categories, all beef, table beef, hamburger beef, pork and broilers are shown in Tables “.23 - “.27. Linear, quadratic, cubic and in some cases fourth-degree polynomials were tested. Direct, cross and income effects were examined for each product. However, because of problems in estimation in all but the broiler equations, polynomial specification was limited to own- quantity and income as these were considered to be of primary interest.61 The continuous time-varying parameter specification was found to make a significant contribution, beyond the classical model of fixed coefficients, in each product model with the notable exception of pork. The assumption of the constant coefficient model has been accepted con- sistently in each model of structural change tested for pork. A related conclusion was made in a study using a random coefficient model of pork production (Dixon and Martin, 1982). They explained that the move 61 A model with all coefficients systematically varying seems un- reasonable in economic applications. o. e t .oaaou osu ow nouns-own ea one «a .H-wlonhaom oouuov anoa>oun ecu cu no .uovOI uncoun accuse-noxaooca ones on coauamsluou us«aws>1ol«u mo conusnuuucoo no cooiowuaeumc no uueu on uneu1~ .uouwe H cash a «cases no huaaqocaouo on human clanoo nu newsman A .aaalcehaon sea «0 cannon can an enema» u owes: .ounowuca Anna-acumwe nanosecoou on» can: o~onuu¢> Asoauoouounuv o>wucounauunsl «so ncwuwucsou newloohuoc cannons; osu cu eunuch «at a 2657 Awn.nc Ann.nc 5-.~c “go. o Amo.n. Ana.mv Aoo.av «can "nausea .~.~ . an. go. go. ano.1 nae. «so. 1 can. 1 «no. moo.a1 ono.~aa cannon "unease . . An«.~c Aco.c .-.~c Aoa.nc Amo.sc Ann.oc Aaa.«c As~.nac Aen.uac am a no cue. do. has. “no. no“. 1 nHa.1 an". 1 can. 1 n-. aoa.~1 «sa.aoa cannon Housman ass nn.~ as. «no. so. Asa.~c a~a.~c Aoa.ac a-.«c Aan.nc Aun.~c Aaa.oac Ano.oc «~o. ans. 1 on”. 1 egg. 1 an“. 1 sea. ana.a1 ana.oad cacao: "cacao . . . Aca.ac Aoo.nc Ano.. a~a.nc Aso.~c Aon.oc Aun.oc an a on no: so. use. sno.~1 ace. 1 pan. 1 «an. o~m.~1 san.oaa cannon "cannon-9o .3.n a I o 0:. «1 x1 oquucmuuu acaaa «can canon oaxmooc cease canes human cases cannon announce cyanogen.» macaw->1oaaa no u and» h “co-wow use cowmmmmwcou uuuooaon new-oohnom us «some: a . «can ~H< uou vac-en aqcuud uo nova: wounding“ amaze->1osah c-ouncuucb anon-concxu n«.¢ OHAQH “hum“ ”ouncuuc> uaovnococ £2613 .n~.¢ ounce 00m n .m~.a .Ha-a «am a Asa.v Amu.~c noo.nc Ana. . Ann.oc Aoo.~c Aco.ac Aa~.nc Aso.ac «can ”wooed; no.“ as. «so. no. coo. «no. ~na.~1 mad. 1 oo.~1 he“. 1 “nu.“ a-.~1 -a.on~ cannaco “ecu-uuaao . “an.ac and. c Ana.~c An~.~ac Ann.~c Ao~.nc Ana.¢c Aoo.~c Aoo.oc Asa.ac as N as. “no. «o. mac. One. 1 man. 1 oo.u1 has. 1 new. 1 has. 1 coo. nm~.~1 nae.oH~ omcnaoo "autumn ace . Aao.~c Ana. . An~.~c Asu.adc Aaa.nc Aon.nc Aao.~c a~m.nc Anh.ac on a on. one. so. «Ho. own. 1 saw. 1 o-.~1 ”an. 1 "as. 1 one. nan.~1 n-.~m~ oeaaaoo "guano . Ann.~c Acm.mc has. c Aen.ac Aoo.nc th.uc Awo.nc Ano.nc "a a on. mac. no. mac. nao.~1 can. 1 oao.~1 an“. 1 com." na~.~1 ~a~.ms~ omaamco "uaoquaaao .:.o «a nxm oucaosow - Hosea noun canoe camcoco cecanzoo ness» cease am~aa mmmaa oaanmoa occa.eoo .uoaoaau-c meant->1oaaa co numuu1w “co-wom not nouucasueoo uuuuoaon Heanochdom «o cannon «can one-a you one-ea uuauon no dove: usual-uem «swan->1oeah couuaouuw> ensues-«awn «N.c nan-H sauna "unaccuqs oaoeeoaoo 2139 .nN.c OHAIH 00m A .n~.e ado-a com a Aco.uv Ach.nv Auo.nv Aa~.v on.nv Ano.cuv Avn.cv Amn.~v omhnmzun «undead an.” on. «no. no. sea. I owe. hH°.~1 ans. 1 mac. ~uo.~ I o~n.~1 Nom.cm cannula: "undead 34.3 31V 2:; 3n; A343 36.3 23.8 Nu.n co. one. no. mac. «no. men. 1 nno. och.fl 1 eno.H1 nan.oou ommnmzon "ownso Awa.nv noe.0v Ann.v Aom.Nv Anu.ouv Aoo.cv n~n.nv ca.H no. «we. ...: one. mon.~1 «on. 1 mac. coo." I on~.H1 coo.nH~ ommnhzun ”aduauecsc Auv.nv Anm.nv Aoo.v Ann.Nv An~.o~v Ano.ev Aaa.nv an.u «0. «cc. no. n~c. Occ.nl «ma. I acc. nun.n 1 cum. 1 www.mna ommnhzun "unocwa I O - a a «m \Q ouwuumwooc and: 6H9 0.38 055.02 En: LEE and: 05—958 Gamma—on ammo-coo Queue-Huck and an.) 1.0le no alflulh "a common you now» Inocoo ouuooSoa Haweoehmom uo cannon moon nonw54li= oneuaewu¢> Anon-coucwu you one-en dunno: «0 Home: usual-wee «ciao->1olaa «hand “onacwul> acovcocoa ZJVC .nm.c Ounuh 00m A .n~.c cap-a com - Ao~.c Ana.Nc ~4.Hc Ana.ac a~o.c Ano.c Aa~.ac Ao~.~c Ae~.~c Ano.~c «can ”nausea ao.~ as. use. ...: coco.1 nno. c~o.1 soo.1 «co. coo.1 nnH.H1 one. 1 man. 1 nso.ena acumen ”assoc Asa.ec Asa... Assnmc Ann.c Aoa.oc Aa~.~c “Ha.ac “Ho.~c some "saunas no.“ on. use. ...a coco. nee. ann.~1 n~H.1 oan.a1 con. 1 «as. 1 aao.oa omuaoo "nausea Ann.nc Ano.Hc Aao.~. “a".c Amo.c Aaa.oc Am~.~c Aa~.~c Ann.nc so.” «a. «so. ...a ammo. «no.1 smo.1 coo.1 oHo.1 ann.~1 one. 1 Man. 1 nc~.oaa unseen "cacao .:.o «m -\w oumunnmoua Henna some ounce games games madam cannon ounnczoo omaamoo uses-coo auuuoaau-m uncut->1oaaa co “nu-eu1h "eunuch not acuucasucou uuuuoaon unwaoc>HOA mo oouuuon cooancuuc> unconccqewm o~.c oases xuom new region undue: no «one: nanoseuch us«»ua>1oawh muamx ”ouacauc> ucuocooon 271 .nN.¢ canon 00m A .nN.c onAnh 00m I Ano.. Ana..c Anq.~c Aou.nc Ana.v “an.~c Aam.c Aa~.nc “an.c Aao.~c ado.~c an." «as. coo. so. nooo.1 one. nan. aam.~1 onA.1 no”. 1 NN~.1 can. 1 nu~.1 ~Nn. 1 aoo.on «canons-s sac ”unease Ana.~c .Ho.nc Aoo.nc Aeu.ac Aa~.Nc Ao~.ac Ao¢.~v Asa.nc «can "saunas on.“ one. «no. so. .1 one. moo. nao.~1 ~ou. 1 Ace. 1 «on. 1 ~n«.na canon "unease a~o.cc “om.cc Ann.o~c As~.~c “so.ac Aan.~c .Nn.ac na.~ «no. "so. no. nae. can. onc.~1 «ha. 1 mac. 1 -n. 1 nan.~o ounce ”caucus O I “> a c «m -\m uuwmuuuoo- Hosea some undue canoe names cannon cease oeanmzoo madam omenaoc Dena-coo .uouoaauaa acaauq>1oaas no A 0 1h "noon-m not segue-socou swuncaoo «cfisonhuom no acumen cnoHAamuop unouancacwu mxnhd noquwuc> ucovsocoa nuonuown new van-ea «wound «0 demo: noun-oucm «sumo->1olah 5N.c 0AA¢h 272 toward capital intensive, confined production operations gives a produc- tion cycle with a smaller amplitude, and therefore, the hog prices response has declined systematically over time. This may explain at least the absence of production induced shifts in retail demand. Typically, as a higher degree polynomial was specified the '§2 increased significantly and autocorellation was usually reduced. The standard F-test was used to test the contribution of each polynomial specified. In some equations it appeared that terms in higher degree polynomials were becoming colinear with other explanatory variables causing sign change or a reduction in previously significant coeffi- cients. Illustrative of this behavior is the transition from the cubic to the fourth degree polynomial in the all beef equation. This suggests that caution is required in the use of this approach in model specifica- tion. The linear polynomial is preferred in the broiler specifications, however the preferred polynomial varied in the beef equations. The property of finality of coefficients, while being imposed upon the Legendre polynomials specified in each equation, does not reveal as strong a degree of constancy in the progressive coefficient values of those polynomials as would be eXpected in some engineering studies wherein errors in estimation may be considerably less or may be con- trolled. Nevertheless, the table beef equations, for example, illus- trate the similarity in the coefficient of earlier polynomials after a higher degree has been added to the model specification. Calculation of direct, cross and income flexibilities for each of these sets of product equations followed the formula illustrated for a 273 cubic polynomial in equation (“.16). Derived flexibilities are con- tained in Tables “.28 - “.33. Flexibilities estimated from the all-beef equations (Table “.28) indicate that regardless of the Legendre polynomial specified, direct and income flexibilities are greater than unity in nearly every year. This suggests first that the beef market has been subject to larger price variations as supply fluctuates and second, that increases in real disposable income per person, although increasing demand for beef, have had a more than proportional impact on market prices. Considerable change has occurred between years although it is recognized that all changes may not be significantly different from each other. Nevertheless, the pattern of change is the same among each of the sets of all-beef flexibilities. This similarity of pattern of change in flexibilities among the various degree polynomials was evident also in the other tables of product flexibilities. This reflects the consistency in the coefficient estimation procedures, since the data series of the quantity-price ratios used in determining the flexibili- ties are not obtained from the estimated equations. Also, consistent with previous analyses of the impact of cycle-induced structural changes on retail demand, the pattern of flexibilities of all beef characterizes production cycles over the period. For example, since the late 19603, direct flexibilities for all beef (Table “.28) tended to decline until 1973. Since quantities supplied per person increased only slightly (demand shifted out only slowly in response to increasing prices), the fall in flexibilities was due largely to upward price movements along the demand curve. The record retail prices in 1973 were associated with 274 Table 4.28 Direct and Income Flexibilities for All Beef Batinated Pro- Legendre Polynomial Model‘ Direct Direct Direct Flexibility: Year Flexibility, Plgxibilitz, Flexibility Direct Income 2012523151: quadratic cubic 4th degree linear linear 1950 -1.118 - .909 -1.004 -1.076 1.321 1951 .893 - .746 - .806 - .861 1.209 1952 -1.009 - .862 - .919 - .974 1.255 1953 -1.560 -l.360 -l.436 -1.508 1.614 1954 -l.531 -1.410 -1.482 -1.538 1.600 1955 -1.604 -1.441 -1.514 -1.551 1.672 1956 -1.686 -1.532 -1.612 -1.628 1.754 1957 -1.577 -1.444 -1.526 -1.519 1.668 1958 -l.312 -l.209 -1.285 -1.261 1.450 1959 -l.275 -1.180 -1.263 -1.221 1.450 1960 -l.353 -1.254 -1.352 -1.290 1.491 1961 -1.406 o1.303 -1.414 -1.332 1.528 1962 -1.349 -1.249 -1.364 -1.270 1.499 1963 -1.477 -1.364 -1.498 -1.380 1.581 1964 -1.497 -1.469 -1.622 -1.479 1.701 1965 -1.524 -l.395 -1.545 -1.396 1.698 1966 -1.572 -1.431 -1.589 -l.425 1.727 1967 -1.627 -1.472 -l.636 -1.455 1.781 1968 -1.645 -1.480 -1.643 -1.450 1.783 1969 -1.555 -1.392 -1.542 -1.350 1.672 1970 -1.618 -1.441 -1.591 -l.380 1.699 1971 -1.559 -1.383 -l.521 -1.305 1.663 1972 -1.488 -1.318 -1.442 -1.220 1.574 1973 -1.233 -1.092 -1.189 - .989 1.433 1974 -l.401 -1.244 -1.349 -1.096 1.467 1975 -1.473 -1.315 -1.423 -1.123 1.467 1976 -1.736 -1.562 -1.692 -1.286 1.610 1977 -1.785 -1.625 -1.768 -1.282 1.700 1978 -1.484 -1.372 -1.506 -1.031 1.488 1979 -1.183 -1.115 -1.241 - .794 1.289 1980 -1.250 -1.206 -1.370 - .808 1.316 1981 -1.384 -1.372 -1.599 - .858 1.391 1982 -1.446 -1.480 -l.779 - .859 1.417 ' Flexibilities were derived from Table 4.23. 275 Table 4.29 Direct and Income Flexibilities for Table Beef Estimated Fro- Legendre Folynoaial Model' . Direct Direct Direct Flexibility: Year Flexibility Flexibility Flexibility Direct Income Degree of if, j"""' Folzgggigl: quadratic cubic 4th degree linear linear 1950 - .769 - .639 - .560 - .754 .555 1951 - .709 - .612 - .528 - .695 .518 1952 - .763 - .681 - .583 - .748 .547 1953 -1.035 - .953 - .813 -1.015 .718 1954 - .892 - .845 - .720 - .875 .726 1955 - .908 - .882 - .754 - .891 .773 1956 - .932 - .924 - .794 - .914 .827 1957 - .841 - .850 ~ .736 - .826 .803 1958 - .747 - .767 - .671 - .734 .713 1959 - .841 - .875 - .774 - .826 .727 1960 - .863 - .909 — .813 - .849 .764 1961 - .907 - .963 - .871 - .893 .800 1962 - .855 - .913 - .836 - .842 .802 1963 - .949 -1.018 - .943 - .936 .865 1964 -1.007 -1.083 -1.015 - .995 .952 1965 - .929 - .999 - .946 - .918 .972 1966 - .975 -1.048 -1.001 - .965 1.013 1967 -1.033 -1.108 -1.067 -1.024 1.070 1968 -1.054 -1.129 -1.094 -1.047 1.098 1969 -1.009 -1.080 -1.050 -1.004 1.055 1970 -1.055 -1.128 -1.100 ' -1,052 1.100 1971 - .996 -1.066 -1.041 - .995 1.105 1972 - .967 -1.039 -1.014 - .968 1.074 1973 - .787 - .853 - .830 - .789 1.004 1974 - .763 - .837 - .814 - .767 1.057 1975 - .636 - .710 - .690 - .641 1.088 1976 - .832 - .951 - .924 - .841 1.229 1977 - .884 -l.041 -1.014 - .895 1.338 1978 - .821 -1.002 - .983 - .833 1.207 1979 - .684 - .871 - .863 - .695 1.079 1980 - .693 - .926 - .931 - .706 1.130 1981 - .756 -1.065 -1.091 - .772 1.244 1982 - .829 -1.236 -1.295 - .849 1.302 ‘ Flexibilities were derived from 1.111. 4.24. 276 Table 4.30 Direct and Income Flexibilities for Hamburger Beef Estimated from Legendre Polynomial Modela Direct Direct Direct Flexibility: Y“, LWLML—M. Degree of Polzgomial: linear quadratic cubic linear linear 1950 - .823 - .934 - .566 - .903 1.294 1951 - .574 - .645 - .429 - .628 1.190 1952 - .688 - .766 - .549 - .752 1.282 1953 -1.311 -1.448 -1.105 -1.430 1.864 1954 -l.627 -1.780 -1.429 -1.770 2.047 1955 -1.633 -1.772 -1.482 -1.772 2.185 1956 -1.730 -1.863 -1.609 -1.873 2.322 1957 -1.598 -1.709 -1.512 -1.725 2.183 1958 -1.139 -1.210 -1.090 -1.226 1.743 1959 - .938 - .991 - .903 -1.008 1.734 1960 -1.024 -1.076 - .988 -1.097 1.843 1961 -1.021 -1.068 - .983 -1.090 1.909 1962 - .991 -1.033 - .949 -1.055 1.937 1963 -1.004 -1.043 - .953 -1.066 2.014 1964 -1.080 -1.120 -1.015 -1.l42 2.213 1965 -1.054 -1.092 - .981 -1.111 2.275 1966 - .983 -1.018 - .905 -1.032 2.259 1967 - .932 - .966 - .850 - .975 2.345 1968 - .897 - .932 - .812 - .935 2.424 1969 - .782 - .815 - .705 - .812 2.299 1970 - .737 - .771 - .665 - .762 2.302 1971 - .722 - .760 - .657 - .744 2.359 1972 - .626 - .664 - .578 - .642 2.274 1973 - .476 - .509 - .451 - .486 1.954 1974 - .724 - .781 - .709 - .735 2.058 1975 -1.166 -l.272 -1.193 -1.178 2.449 1976 -1.l39 '1.258 -1.231 -1.144 2.619 1977 -1.120 -1.256 -1.294 -1.119 2.888 1978 - .713 - .812 - .890 - .708 2.424 1979 - .452 - .524 - .615 - .446 1.904 1980 - .529 - .624 - .791 - .518 2.026 1981 - .609 — .734 -1.009 - .592 2.252 1982 - .605 - .746 -1.118 - .584 2.441 ‘ manning. were derived from 111111. 41.25 277 Table 4.31 Direct and Income Flexibilities for Pork Estimated From Legendre Polynomial Modela .Direct Flexibility: Flexibilit : Year Flexibility Direct Income Direct 13%353— Degree of Polynomial: cubic linear linear cubic linear 1950 - .907 - .895 1.063 - .916 .862 1951 - .973 - .939 1.078 - .982 .872 1952 -1.056 -1.003 1.150 -l.067 .928 1953 - .889 - .834 1.122 - .898 .904 1954 - .812 - .755 1.065 - .820 .856 1955 -1.075 - .994 1.322 -1.085 1.060 1956 -l.155 -1.066 1.443 -l.166 1.154 1957 - .946 - .873 1.293 - .954 1.032 1958 - .898 - .831 1.224 - .906 .974 1959 -1.150 -1.070 1.447 -l.160 1.149 1960 -1.132 -1.060 1.494 -1.141 1.184 1961 - .938 - .886 1.463 - .945 1.156 1962 -1.071 -1.022 1.517 -1.079 1.196 1963 -l.172 -1.130 1.648 -1.180 1.296 1964 -1.190 -1.160 1.783 -1.197 1.399 1965 - .917 - .905 1.616 - .922 1.265 1966 - .828 - .827 1.537 - .832 1.201 1967 -1.026 -1.038 1.792 -l.03l 1.396 1968 -1.088 -l.114 1.920 -l.092 1.492 1969 -1.023 -l.059 1.859 -l.027 1.442 1970 -1.054 -1.102 . 1.900 -1.056 1.470 1971 -1.332 -1.406 2.247 -1.334 1.735 1972 -l.067 -1.l34 2.021 -l.067 1.556 1973 - .783 - .837 1.712 - .782 1.315 1974 - .949 -1.019 1.878 - .947 1.439 1975 - .683 - .734 1.636 - .681 1.251 1976 - .776 - .833 1.775 - .772 1.354 1977 - .926 - .990 2.067 - .921 1.571 1978 - .883 - .938 1.992 - .877 1.512 1979 -1.135 -1.194 2.205 -1.126 1.670 1980 -1.454 -l.508 2.503 -1.440 1.892 1981 -l.428 -l.457 2.530 -1.413 1.907 1982 -1.214 -1.214 2.334 -1.200 1.756 a Flexibilities were derived from Table 4.26. 278 Table 4.32 Direct and Income Flexibilities for Broilers Estimated From Legendre Polynomial Modela Direct Flexibility: Year Flexibility Direct Income 1950 - .308 - .305 .915 1951 - .377 - .370 .939 1952 - .427 - .417 .955 1953 - .461 - .446 1.009 1954 - .569 - .547 1.107 1955 - .547 - .522 1.107 1956 - .793 - .749 1.309 1957 - .924 - .866 1.374 1958 -1.106 -1.027 1.396 1959 -l.256 -l.156 1.580 1960 -l.282 -1.l69 1.575 1961 -1.584 -1.431 1.768 1962 -l.497 -l.339 1.719 1963 -1.609 -1.425 1.786 1964 -l.686 -1.478 1.922 1965 -l.759 -l.526 1.954 1966 -l.834 -1.574 1.946 1967 -2.046 -1.736 2.187 1968 -2.046 -l.716 2.209 1969 -2.141 -1.774 2.189 1970 —2.462 —2.015 2.394 1971 ~2.498 -2.018 2.481 1972 -2.644 -2.108 2.569 1973 -1.879 -1.478 1.957 1974 -2.l98 -l.704 2.211 1975 -2.077 -1.587 2.089 1976 -2.458 -l.849 2.302 1977 -2.692 -1.994 2.478 1978 -2.761 -2.012 2.421 1979 -3.251 -2.329 2.586 1980 -3.391 -2.386 2.615 1981 -3.730 -2.577 2.756 1982 -4.121 -2.793 2.944 a First degree polynomial was the preferred form. Flexibilities were derived from Table 4.27. 279 Table 4.33 Direct, Cross and Income Flexibilities for Broilers Estimate From Legendre Polynomial Modela ' Direct Cross Flexibility Hith Respect to: Income Year Flexibility "Table Hamburger Park Flexibility Beef Beef 1950 - .282. - .036 - .091 - .024 .720 1951 - .343 - .042 - .074 - .033 .747 1952 - .386 - .050 - .088 - .042 .769 1953 - .414 - .063 - .129 - .045 .824 1954 - .508 - .067 - .167 - .055 .915 1955 - .486 - .072 - .164 - .066 .926 1956 - .699 - .092 - .203 - .087 1.110 1957 - .808 - .101 - .219 - .093 1.180 1958 - .960' - .115 - .208 - .105 1.215 1959 -1.083 - .161 - .204 - .143 1.395 1960 -1.098 - .175 - .219 - .149 1.409 1961 -l.346 - .220 - .247 - .155 1.605 1962 -1.262 - .224 - .241 - .181 1.582 1963 -1.346 - .266 - .255 - .206 1.668 1964 -l.398 - .306 - .281 - .226 1.821 1965 -1.447 - .310 - .283 - .211 1.880 1966 -l.495 - .344 - .277 - .215 1.901 1967 -1.653 - .426 - .297 - .277 2.170 1968 -1.638 - .470 - .292 - .297 2.227 1969 -1.698 - .508 - .279 - .306 2.241 1970 -1.933 - .609 - .300 - .361 2.492 1971 -1.942 - .645 - .311 - .426 2.626 1972 -2.034 - .724 - .303 - .418 2.766 1973 -1.430 - .519 - .214 - .292 2.144 1974 -l.654 - .582 - .366 - .385 2.465 1975 -l.546 - .477 - .490 - .317 2.371 1976 -l.807 - .651 - .516 - .383 2.661 1977 -1.956 - .727 - .520 - .442 2.919 1978 -l.981 - .774 - .404 - .445 2.908 1979 -2.301 - .812 - .366 - .573 3.166 1980 -2.368 - .833 - .427 - .676 3.266 1981 -2.568 - .909 - .489 - .714 3.513 1982 -2.797 -l.058 - .504 - .721 3.833 ‘ First degree polynomial was the preferred form. Flexibilities were derived from Table 4.27. 280 producers' decisions to withhold from slaughter, cattle for herd expan- sion, and the consequent consumer boycott of beef. Thereafter slaughterings were increased, and retail prices began steadily to fall. This pattern of increased supply availability and falling prices is reflected in increased flexibilities between 1973-77. Direct flexibili- ties fell between 1977-79 primarily. in response to reduced supplies available for consumption following herd liquidation and subsequently rising real prices. Therefore, between 1973 and 1979 the change in all-beef direct flexibilities appear largely due to movements along the prevailing demand curve and not due to dramatic shifts in demand. More recently, however, structural shifts in demand may have played a much greater role. Between 1979 and 1982 all-beef demand moved backwards to the left. Consumption per person over this four-year period has been rela- tively stable, although real prices have fallen, causing the direct flexibility to increase in each year. Over the past several years of a weakening demand for beef, this rising flexibility implies relatively large price adjustments in the market to small changes in quantities supplies. This same conclusion was reached by Chavas (1983) in his analysis of structural change and meat demand. He found in his time- varying parameter model using Kalman filters, that estimated direct price elasticities for aggregate beef per person tended to decline dur- ing the late 19703. In broad terms this agrees with the above result that direct flexibilities would tend to rise over the same period. 281. Hohlgenant (1982) explained much of.this decline in demand, i.e., nega- tive shifts in demand for red meat, as being due to negative relative quality changes which have led to a substitution of poultry meat for red meats. Estimates of income flexibilities for all beef in Table “.28 show a broad tendency to rise through the middle 19603, then fall during the late 19603 to the middle 19703, then to rise again through to 1977, and eventually to be at relatively lower levels in 1982. Up to about 1970, this general pattern correlates positively with the changing direction in real disposable income per person. This means that the income response is strongest when incomes are rising, indicating support for the first income hypothesis discussed in section “.5.3 on irreversibil- ity of demand. However in the latter half of the period 1950-82, the periods of decreasing real disposable income per person were generally associated with rising income flexibilities, and vice versa. This sup- ports the alternative hypothesis discussed earlier. The specific decline in income response shown from 1977 to 1979 was also observed by Chavas (1983. p. 152) over a similar period of the late 19703. From this relative decline in income flexibility he projected a decrease in the long term growth rate of the U.S. beef industry. This down trend in income flexibility between 1977 and 1979 also occurred for table and hamburger beef, ”the components of all-beef demand. Since 1979, however, income flexibilities have progressed steadily upwards for all beef categories. This would suggest the opposite conclusion of Cha- vas, namely, a prospect of an increase in the long-term growth rate of the U.S. beef industry. However, two points of caution are associated 282 with this conclusion. First, recent changes in disposable income may have been correlated with factors not included in these estimates. Second, irreversibility of demand with respect to income levels may mean a smaller response to rising incomes. Considerable fluctuation occurred in direct flexibilities of both table and hamburger beef (Tables “.29 and “.30). In terms of relative magnitudes these values were similar to those obtained in earlier struc- tural models. Table beef direct flexibilities fluctuated in an upward direction throughout most of the 19503 and 19603, and for the early part of the 19703. This reflected the firm upward trend in beef supply avai- lability and hence consumption per person during this period and the declining real prices for table beef. Although the demand curve became flatter (i.e., slope decreased), the rise in the quantity-price ratio more than offset the decreasing slope. Up to the early 19703 consumers had developed distinct patterns and habits in the consumption of table beef. As a result, they became rela- tively insensitive to changes in prices which for the most part were still at relatively low real values. The market, however, was quite price sensitive to changes in beef supplies as illustrated by the ten- dency for direct flexibilities to rise over this period (Table “.29). Eventually, as real table beef prices rose through the early 19703 and demand shifted leftward, consumers became more sensitive to price changes. Since the mid-19703, both supply and demand shifts appear to have caused considerable fluctuation .in estimated flexibilities. In more recent years from 1979 to 1982 direct flexibilities have tended to rise again. Compared to earlier years this increase would suggest, 283 - since real prices have fallen sharply over the period and consumption has remained virtually unchanged, that a substantial structural shift in demand occurred against table beef. Moreover, future increases in supply availability would cause large price variations in the table beef market largely because a flexible demand for table beef would result in propor- tionately large price adjustments to clear the market. By comparison the direct flexibilities of hamburger beef over the last four years from 1979 to 1982 have been lower than previous years (see the preferred linear models). This indicates a relatively inflexible demand in recent years for hamburger beef. At the consumer level, consumers are highly responsive to changes in retail prices. At the retail market level, an increase in supply availability would be closely matched by demand, requiring a less than proportional adjustment in market clearing prices. As noted above, the level of income responsiveness for beef has declined relatively over the past several years. This is not a favor- able trend for the long-term growth in the U.S. beef industry. Most significantly this decrease in income flexibilities (absolutely from 1977 to 1979 and relatively to 1982) occurred in both the table beef and hamburger beef markets. The study by Chavas (1983). in which similar declines in income response (income elasticities for all beef) were reported, covered the period only up to 1979. Since 1979 steady increases in income flexibilities occurred for all three beef product SPOUPings. Perhaps as incomes rise, consumers are tending to spend pro- portionately more on non-meats. In the case of hamburger beef may be overestimated since a priori a stronger demand response is to be expected from table beef over hamburger beef. Certainly the upward 284 trend in table beef income flexibilities is stronger whereas those for hamburger beef tend to fluctuate considerably. As mentioned above, pork demand estimated with a constant coeffi- cient model has been accepted consistently throughout this analysis. This is reflected by the non-significant F-test results shown in Table “.26. This conclusion is also supported by the relative consistency of parameter estimates and by the magnitude and direction of estimated flexibilities presented in Table “.31. The magnitudes of direct flexi- bilities shown in Table “.11 using a time interaction variable, T80, and flexibilities in Table “.21 using linear and cubic spline functions, bear a very close resemblance to those in Table “.31. This is because the additional time-varying specification did not make a significant contribution to the fixed coefficient classical linear model. This is not to say that significant increases and decreases in demand did not occur . It does indicate, however, that significant continuous changes in direct and income flexibilities have not occurred over the 33 Year period. Recall, however, that slope interaction dummy variables testing for abrupt changes in structure did indicate significant demand shifts of slope.62 Chavas also reported an absence of structural change from his continuous time-varying parameter model using Kalman filters. Pork was shown, however, to have significant cross-effects on both beef and poultry. In the present study, pork has shown an increasing income * 6 2The overall contribution of these slope dummy variables was sig- nificant at the .05 level. Intercept dummies were also tested and found to be statistically significant. 285 flexibility over time. Much of this rise in income response has been due to the relatively stable real retail prices for pork and the steadily rising real disposable incomes per person. From an examination of the price-quantity plots in Figure 4.“ it is perhaps not surprising that the linear polynomial model is preferred among the broiler demand models (Table “.27). The polynomial coeffi- cient in these equations indicate first, that the demand curve for broilers is becoming increasingly flatter, as is indeed evident from the plot of Figure 4.4., and second, that the slope on income is becoming smaller (i.e., steeper curve). Also with some exceptions, the quantity-price ratio has risen steadily since the 19503 in response to falling real prices and rising levels of consumption per person. The real price/real income ratio has also risen strongly over this period. As a result the direct and income flexibilities have shown a strong upward movement over time (Table “.32). It is not possible to directly compare relative magnitudes of income flexibilities across meat categories because, for example, of some underestimation in the beef categories. Nevertheless, the comparatively strong upward trend is unmistakable. With respect to the income flexibility, this may indicate continued strong long-term growth prospects for the broiler industry. However, against this conclusion are indications that the market for chicken is reaching saturation. Of particular interest in the broiler model are the cross- flexibilities presented in Table “.33. The cross-product coefficients on the polynomial terms in the broiler equation in Table “.27 are not statistically significant, however, in each case they have a negative 286 sign. This indicates an increasing cross-effect, ceteris paribus. How- ever, the relative size of the cross-quantity flexibilities depend also on the cross-quantity/own-price ratio. Not unexpectedly, therefore, table beef has the strongest cross flexibility of demand. The cross effect of pork on broiler prices is greater than hamburger beef quanti- ties on broiler prices. Chavas' results indicate an ambiguous rise in price elasticities, the opposite effect of a rise in direct flexibili- ties, a decreasing cross-elasticity for beef and a sharply decreasing cross-elasticity for pork. “.8 Concluding_$ummary In this chapter a number of estimating techniques and statistical models were applied to an analysis of structural change in the U.S. retail demand for meats, namely, table beef, hamburger beef, pork and broilers. Each of these approaches is characterized by estimation coef- ficients which were permitted, to a greater or lesser degree, to vary through time. This is a direct contrast to the constant coefficient models of classical demand relationships. The following were the objec- tives of research undertaken in this chapter: (a) to test the hypothesis of no parameter variation in the estimated demand relation- ships, i.e., to accept the classical model of constant coefficients of demand; (b) where parameter variation is identified, to establish if there is any systematic behavior contained in that variation over time; and (c) to relate systematic variation in terms of estimating slope coefficients and flexibilities of demand to structural changes in retail meat demand which have occurred over the historical period. 287 With graphical analyses as a guide, structural changes were identi- fied and quantified using qualitative shift variables, linear and cubic splines fUnctions within a discontinuous time-varying switching model, and Legendre polynomials within a continuous time-varying parameter model. Annual models of market demand for meats were specified as price dependent on the assumption that supplies available fer consumption determine retail prices. This assumption is less satisfactory for broiler demand, however, for overall model consistency and compatibility within the Michigan State University Agricultural Model (MSUAM), all meat demand equations were price dependent market demand models. Flexi- bilities, therefbre, are the appropriate measures of demand responsive- ness . Linear and log-linear fUnctional forms were used in ordinary least squares estimation of retail demand parameters for each of the four retail meats. Alternative estimation procedures were tested and rejected because of inferior fit and unexpected signs on some cross- effects of pork. To investigate structural changes in demand, a long series of annual data was used; 33 observations from 1950 to 1982. Preceding the parameter variation analysis were two related ana- lyses of structural change in 0.8. meat demand. The first, a price adjustment model was estimated and differences in short-run and long-run flexibilities were examined. The short-run direct flexibility was found to be smaller than the long-run flexibility. The impact of imperfect knowledge and institutional delays in the short-run prevent complete and instantaneous adjustment of market prices to clear the market. Only after some time has elaspsed in subsequent periods do final adjustments 288 in prices take place. Arguments are provided as to why this result does not contradict the traditional notion of Marshallian short and long-run demand curves. In the second analysis the influence on meat demand of changes in the age composition of the population and of the trend towards increased eating away from home were investigated. Away-from- home purchases of food showed no significant influence on retail beef prices. The retail price series, for exanple, for hamburger beef may be an inappropriate price when considering the influence of institutional (fast food restaurants) purchases. The growth in the 5 to 2“ age cohort, a result of the 'baby boom' of the 1950s, has been a significant factor in augmenting demand for hamburger beef, whereas a significant negative relationship was identified between older age cohorts and ham- burger beef demand. Graphical analyses, involving plots of retail prices and domestic consumption per person of individual meats, indicate substantial changes in price-quantity relationships over time. Using time-quantity interac- tion variables it is evident that over the past several years direct flexibilities for table beef and broilers have risen; for hamburger beef they have declined while fOr pork they have remained comparatively unchanged. The lower the direct flexibility, the greater the exent to which increases in quantities available per person are absorbed by demand and hence the smaller the adjustment of price to clear the market. Given supplies made available to consumers, this implies that, recently, quantities of hamburger beef have been more readily absorbed by demand than have table beef or poultry meats. Relatively speaking, the preference for hamburger beef has been growing visJa-vis other meats. From this preliminary analysis it is apparent that considerable 289 change in the structure of U.S. demand for meats, especially of beef and broilers, has occurred over the past several decades. An attempt was made to relate in a general way the observations of economic structural change in meat markets to the estimated changes in flexibilities over time. One testable hypothesis explaining systematic changes in the structure of meat demand is the hypothesis of irreversi- bility of demand. The general hypothesis was that direct flexibilities will tend to be more flexible during decreasing consumption phases and less flexible (inflexible) during increasing consumption phases. The differences in the estimated direct flexibilities were quite small, nevertheless, the pattern of change in their values agrees with the hypothesis of irreversibility of demand. Most conclusive results were obtained by using changes in the cattle cycle as the underlying cause of these systematic changes in direct flexibilities of demand. Linear and cubic spline fhnctions were employed within a time- varying switching regression model. This model, an extension of those tested earlier, permitted slope parameters to change over some periods and remain constant over others. These periods of change were modelled in a continuous fashion. With the notable exception of pork, the null hypothesis of no variation in the regression coefficients of the own- quality was rejected at the one percent level. For example, the log- linear spline model of broiler price indicates that the direct flexibil- ity is not constant. It has been increasing linearly over time although the rate of increase has slowed since the middle 19603. This result is consistent with the observations that the market for chicken is becoming saturated and that the impact of gains from technology in the broiler 290 industry, which have kept real prices down, may be slowing. More signi- ficantly, the competitive advantage of price enjoyed by chicken over beef may be steadily becoming diluted. Compared to splines, Legendre polynomials provided continuous time-varying parameter estimation and had the particular advantage over other polynomials of the property of finality of coefficients. The orthogonal nature of these polynomial functions provides independence among the coefficients of each polynomial term added. The same degree polynomial is not required for each explanatory variable while for oth- ers the assumption of a constant value may be justified. There is con- siderable flexibility in its use in providing a time-varying character to regression coefficients and, hence, estimated flexibilities were obtainable for each period. Limitations of this estimation approach include problems associated with usual regression analysis: in particu- lar, specification of the degree of polynomial must be based on economic reasoning. Continuous time-varying parameter specification was found to make a significant contribution, beyond the classical model of fixed regression coefficients. Again the notable and consistent exception was in the pork demand model. The direct and income flexibilities of all-beef demand were generally greater than unity. This indicates first, that the beef market has been subject to larger price variations as supply fluctuates, and second, that increases in real disposable income per Person have had a more than proportional impact on market prices. The level of income responsiveness for beef has declined relatively over the past several years. This decrease in income flexibilities (absolute 291 from 1977 to 1979 and relatively to 1982) occurred in both table beef and hamburger beef markets. Nonetheless, the steady upward progression of income flexibilities since 1979 should not be overlooked. Income flexibilities have continued strongly upward for broilers. A consistent result throughout has been the recently rising direct flexibilities for table beef and broilers and a declining direct flexi- bility for hamburger beef. This latter trend is consistent with the increasing price competitiveness of hamburger beef with table beef, a situation reminiscent of the high beef price period of the early 1970s. Ground beef appears to have become a closer substitute with table beef. Chapter 5 EMPIRICAL EVALUATION OF THE IMPACT OF STRUCTURAL CHANGE ON RETAIL MEAT DEMAND AND U.S. BEEF IMPORTS In the previous chapter various specifications of retail demands for meats, namely, table beef, hamburger beef, pork and broilers, were estimated and results were presented. In the latter section of that chapter, continuous time-varying parameter models were estimated over 1950-1982. These were considered to be a substantial improvement over fixed parameter models estimated over the same period. The notable exception was pork demand. In this chapter, these preferred demand equations are integrated into the Michigan State University Agricultural Model (MSUAM). The MSUAM is an econometric simulation model of the U.S. livestock and grains sectors and world wheat and feed grains regions. Equations linking the retail meat demand component with the MSUAM are derived or estimated. The benefits of including the retail meat demand component into a model like the MSUAM derive from the systematic insights into and assessments of meat demand within a U.S. national and international context. This assessment is greatly enriched by the comprehensiveness of the MSUAM. The analysis proceeds by simulations of the model under varying policy and economic scenarios of national and world agricul- tural markets. Scenarios selected reflect specific structural changes potentially impacting either directly the U.S. demand for meats via the 292 293 estimated demand relationships or indirectly via related national and international agricultural sector markets. Of particular interest in this research is the market for non-fed beef in the U.S. As noted earlier, a substantial portion of non-fed beef consumed in the U.S. is imported, predominantly from Australia. In each model simulation, and hence scenario presented, the impacts of structural change on the level of non-fed beef imports are examined. Scenarios examined over the 1983-1991 period include: (a) model sensitivity to one year and medium term shocks arising from changes in the levels of retail meat demand or U.S. real dispos- able income per person; (b) the sensitivity of key livestock and meat market variables to a one-year or medium-term increase or decrease in U.S. non-fed beef imports; (c) the impact on non-fed beef supply and hence, on farms and retail meat prices of a dairy PIK (payment in kind) program and the implied slaughter of dairy cows to bring into balance supply and demand in the U.S. dairy market; (d) the effect of sustained economic growth or decline in developing market and communist economies on their feed grain imports and hence on the U.S. fed beef market; and (e) the ramifications of changing structural relationships, own- quantity, cross-quantity, and income effects, in retail meat demand as characterized by parameter variation over time. 294 In a number of these scenarios, where appropriate, the differential impacts of the structural change under conditions of a strong (excess demand) feed grain market and a weak (excess supply) feed grain market are investigated. Before reporting the results of these various policy and economic scenarios, an overview of the MSUAM and its structure is given. This is followed by details of the U.S. retail meat demand component and of its integration into the existing feed-livestock component of the entire model. The linkage equations and their specifications are discussed. Assumptions made regarding projections of exogenous variables, base pro- jections of endogenous variables generated by the augmented MSUAM and validation of key market variables are also presented. 5.1 Overview of the Michigan State University_Agricultural Model The MSUAM is an annual econometric model comprising a detailed U.S. agricultural model and a nine region world grain model (Figure 5.1). This multi—commodity, non-optimizing simulation model follows a systems methodology of being essentially problem solving. It is designed primarily for intermediate and long term analyses. To date the model's strengths have been in addressing questions of trade levels, trade policies, domestic grain policies and long-term trends in U.S. and world grains. A primary objective of this research, particularly in this chapter, is to develop an integrated U.S. retail meat demand com- ponent for analysis of the four major meat categories, table beef, ham- burger beef, pork and broilers. 295 U.S. Feedgrains Livestock and Retail Meat Sub-Models International Grains Sub-Model .ouo: .asaup=o_cm< am: so mucocoaeou Loam: Lo ugogozopm —.m scams; a .covaom .rwx covuuzuoga fiqogu 1. mauve; ~ouoxunam xooumo>—d nausea to: :32. 6 _ 222 A comm covuoauogc 4v , s: \ sis “V .m.= 29:233. i a 22223 s25 A: cc.uap=noa cowuoa osoucu nausea csagu v noose Alli... seas. season xooum 37... :28: n Amusoaxuv xuv—oa nogu mugoae_ um: u=o5:5o>ow .m.= w 296 The agricultural model as originally developed by Hondai (1975) and Trapp (1976) contained three basic components: (a) a domestic supply component of wheat, feedgrains (corn, sorghum, barley and oats), oilseeds, fed beef, non-fed beef, pork, poultry and dairy products: (b) a domestic demand component for each of the above commodities; and (c) an international trade component of U.S. exports of wheat, feed grains and oilseeds. Later, Baker (1978) independently developed an econometric and financial projections model of the U.S. farm sector. This model contains a farm cost sector, flow of funds, and an income accounting and balance sheet framework. This component has been used to provide intermediate and long-term forecasts of the farm sector's finan- cial position for use in strategic planning by producers, input sup- pliers and government policy analysts. Wailes (1983) further developed the overall model through re- estimation and respecification, and integrated the financial model into the MSUAM. His major contribution was to develop Ia U.S. agricultural policy component and to set the whole model into a stochastic simulation framework. The U.S. policy component comprises both supply and demand management policy interactions. Supply management recognizes the role of loan rates, target prices, set-asides, national program acreage, diversion payments and recommended voluntary diversion. Demand manage- ment recognizes the Commodity Credit Corporation (CCC) stocks and farmer-owned reserve stocks rules and their relationship to the supply control variables. This policy component is a departure from other U.S. agricultural sector models. In essence the farm program is endogenized in the model. Indeed this is consistent with the considerable discre- tionary authority given to the U.S. Secretary of Agriculture and 297 suggests that the policy process be modeled to interact endogenously with the supply and demand forecasts of the MSUAM. However, since government policies have historically related to the grain market and not to livestock production, no explicit policy frame- work exists fOr the livestock sector. Nevertheless, government grain policies can be expected to have a significant indirect effect on the livestock sector, an effect which would be captured in this model. The domestic-international linkage was based on export supply available from domestic production, and farm grain prices are derived from export prices. Subsequently, Mitchell (1982) added the nine region world wheat and coarse grains model. This endogenized world wheat and coarse grain prices. The world and U.S. soybean model was added by Christensen (1979)1. Therefore, the U.S. domestic component provides estimates of pro- duction, feed and food usage, prices, stocks and exports of all major grains, soybeans and livestock. Livestock numbers and production are estimated for fed beef, non-fed beef, dairy, pork, broilers, turkey and eggs. These commodities, overall, account for some 80 percent of U.S. farm income. Grain and livestock sectors are linked by feed utilization (grain consuming animal units) equations. More will be detailed on the livestock sector, later in this chapter. 1 The initial project began under the direction of vern Sorenson and John Ferris, although other major contributors include J. Roy Black and John Ross. 298 The world agricultural component estimates production, consumption and trade of wheat, coarse grains (corn, oats, barley, sorghum, rye) and soybeans for four exporting countries (Canada, Australia, Argentina, Brazil) and five importing regions (Soviet Bloc--USSR and Eastern Europe, Developed Markets, Less Developed Markets, China, and Undesignated-—trade type cannot be assigned). International and U.S. models are simultaneously solved to provide net wheat, coarse grain and soybean trade by region and world prices. Hence export quantities and prices are endogenously determined. Specific interregional flows are not identified.2 The model is dynamic in the sense that the solution in period t+1 is determined by the solution in period t. No explicit or extensive treatment of risk, uncertainty, investment, or disinvestment is con- tained in the model specification. Certain specifications attempt to proxy various aspects of these notions of model dynamics. For instance, the hypothesized relationships of the role of price or profit expecta- tions and partial adjustment due to information, institutional or techn- ical lags contribute somewhat to the dynamic specification of the model. No explicit attempt is made to include information on detailed resource use or on factors of production. Moreover, the model is not specified to analyze issues of long-run resource requirements, input 2The international component is the driving force for the overall agricultural model and reflects the importance of world trade to U.S. agriculture. In contrast to the more commonly used spatial equilibrium models which allocate trade according to transporta- tion costs, world demand in the MSUAM is allocated among exporting countries according to a trade hierarchy. This hierarchy is dependent upon grain available for export and upon marketing behavior. 299 usage, technological change, investment or environment. The extent of aggregation of the endogenous variables will always be of concern to the modeler in some analyses but not in others. Interaction with the general economy is not modelled explicitly in the MSUAM although there is one aspect of the model which permits some feedback in this regard. The author developed a subroutine which deter- mines endogenously the consumer price index for food (CPIFOOD).3 Using a market basket approach, farm values and market spreads are econometri- cally estimated for each component of the food basket. Retail prices or retail values are then calculated to obtain a retail price weighted index of food. The programming subroutine is contained in Appendix B. Changes in government policies and other economic shocks can then be analyzed with respect to their impact of the CPI for food, and hence their commensurate inflationary impact on the general economy. Overall, the MSUAM may be viewed as a set of integrated commodity models with three important sets of linkages: (a) feed grain--livestock, (b) crop supply and demand and policy management, and (c) domestic- international grain markets. 5.1.1 Model Estimation Procedure The MSUAM, including the U.S. retail meat demand component contains some 215 endogenous variables and 210 exogenous variables. There are approximately 90 endogenous equations estimated by ordinary least squares using data from 1960 to years ranging between 1976 and 1982 3The contribution of John Ferris is acknowledged. 300 depending on the equation, available data and when it was last estimated. The shortcomings of the OLS estimation method are recognized as are its strengths. The MSUAM is in large part recursive, though not completely so. Supply is predetermined while demand responds to current information, including quantities supplied. To the extent that model components are recursive, single equation biasness and inconsistency in parameter estimates will be minimized. OLS has been shown in Monte Carlo studies to maintain the Gauss-Markov property of minimum variance. Variants of the limited and full infbrmation techniques would place a prohibitive burden on degrees of freedom in a systems model as large as the MSUAM. This problem of the undersized sample was also discussed by Labys (1973. p. 138) who noted the necessary trade-off between biasness and efficiency of OLS estimates.u (Thiel, 1971). Moreover, as Wailes (1983) observed, various researchers have found that OLS tends to improve relative to limited information estimates as the model size increases. Finally, a very significant justification for this estima- tion procedure is its relative computational simplicity and inexpensive- [1888. 5.1.2 Solution Algorithm Simulations described in this study were carried out using the GSIM program (Wolf, 1983) developed in the Agricultural Economics Department at Michigan State University. GSIM employs the Gauss-Seidal method of n For a discussion of some of these problems as they relate specif- ically to the application of models like the MSUAM, see Cromoco (1982, pp. 26-30). 301 iteration to obtain a solution to a set of simultaneous equations.5 This is a straightforward numerical method for the solution of systems of equations and is used widely in large, complex economic models. A particular advantage of the simulation approach to agricultural sector modeling is that constituent agricultural commodities may be analyzed simultaneously. Important cross-commodity effects may be con— sidered. For some analyses of the effects of policy changes a simula- tion model of this kind has advantages over the alternative multiplier analysis (Labys, 1973, p. 199). For instance, analysts may consider in a simulation analysis varying rates of change in an exogenous variable or varying levels of several exogenous variables together. This flexi- bility provides a considerable advantage in the evaluation of different policy and economic scenarios in agriculture. 5.2 U.S. Domestic Livestock Model Beyond the detail of the previous description of the MSUAM, there is neither sufficient space nor time to detail the entire model used in simulations to be presented later. Recent presentations of the com- ponent submodels are given in Wailes (1983) and Mitchell (1982). Details of the U.S. retail meat demand component will be covered in much greater detail since this is an addition to the existing model. To do this, however, it seems necessary to provide at least a brief outline of SGSIM is based on the General Analytical Simulation Solution Pro- Eram (GASSP) developed originally at the USDA. 302 the U.S. livestock model into which the retail meat model is being integrated. As stated earlier, the livestock sector includes beef (fed and non-fed), dairy, pork and poultry (chicken, turkey and eggs). The livestock sector is based on a calendar year and crops are based on their respective marketing year. At the farm level interactions between livestock types are limited. Interactions of dairy steers in fed beef production and dairy cow inventory as a determinant of non-fed beef pro- duction are examples. 5.2.1 Beef Supply Figure 5.2 gives a diagrammatric representation of the U.S. beef supply, and dairy supply and demand as estimated in the MSUAM. Submo- dels linking these equations with other livestock and feedgrain equa- tions are shown.6 The specifications of equations of the beef supply model provide aggregate supplies of all steer and heifer slaughter, which is classi- fied as fed-beef production, and of cow and bull slaughter, which is classified as non-fed beef production. With the introduction of the retail meat demand model, estimated marketing margin equations provide the link to fed beef and non-fed beef prices at the farm level. Transmission equations provide the linkage between fed and non-fed beef 6Rectangular boxes indicate stock variables and product flows, while circles represent other variables of influence. 3()3 vealua nee m—aasm »g_.: as. A-neam econ .m.: ea «Luguno~u u~.m ogaa.u n.0vox-n=m »_aaam Lo—_oLa Apanam neon oucacaoaa- -m_c x,_: you so; :o.auauoeg a~_x :O—uunthm go—uuauoua :o_uu:ucgn moon 309 eons uoa-=o= cacao ‘0 “Lan_ coon =o_.u=eo1a .ouox-asm ‘ 82.8 any: ~.ouo¢ =o_uu:uoca even —~u0h nou'sn wpau canoe Louccauwm Lo» mauu.oz meow mucusaumpnuz Lomwoz coon uuvgn poocxom ._ovox-p=m :. a: z.w03 2 Lou sa—m gougo:a_m Locum Lu _o. «seam lacuna>_s team cool as co_uuavoca\ 304 quantities and table and hamburger beef quantities consumed at the retail level, respectively. Both producer decision variables and biological response variables characterize the types of relationships in the beef model. Beef cow numbers are determined by a five-period lag in feeder calf prices, change in consumer price index to reflect the opportunity costs to the enterprise of carrying a cow, lagged beef cow numbers and a time trend. A notable feature of this specification is the asymmetric contraction and expansion phase specified into the five-period lag in feeder calf prices. The steer and heifer slaughter behavioral equation was initially specified to explicitly recognize these biological and decision processes. However, due to the biological inconsistencies of this equa- tion with beef cattle inventories, an alternative model was specified. In order to impose consistency with the inventory of beef carried into the slaughter period, fed beef produced was derived using estimated slaughter numbers and slaughter weights. This alternative model required ,explicit information on calving rates, calf survival rates, replacement rates, veal slaughter, feedlot survival rates, and dressed weights. The trade-off of the additional estimations and information required was greater consistency between projection of the breeding herd and the slaughter volume. An expense of this formulation is the short- run impact of changes in corn and soymeal prices on fed-beef production. For medium-term simulations, however, this cost may not be great. Note also that in the original MSUAM, fed beef production included the slaughter of all steers and heifers and non-fed beef slaughter 305 comprised cow and bull slaughter only. To the extent that steers and heifers are grass fed before slaughter, fed-beef production will be overestimated and non-fed beef will be underestimated. Adjustment for this bias is made in the retail demand model equations which are estimated using the more accurate definition of fed and non-fed produc- tion, discussed in Chapter “. Cow and bull slaughter is estimated as being determined by lagged beef and dairy cow inventories, lagged corn and feeder calf prices, current-period hay price and a three year polynomial lag on the Omaha commercial low price. Estimated results appear to reflect the important influences of the culling decision. Import quotas govern the quantity of manufacturing quality beef allowed into the U.S. Since 1980, the countercyclical meat import for- mula has been used to determine U.S. beef imports (Conable, 1980). The formula is computed as: 5 yr. Moving Aver- age of Per Caput Average_ 3 yr. Moving Average Domestic Cow Beef Annual Annual of Domestic Production , Production Quota - imports . 10 yr. Average of Do- 2 yr. Moving Aver- 1968—77 mestic Production, 1968- age of Per Caput 77 Domestic Cow Beef Production (5.1) The countercyclical import law is designed to permit greater imports of meat when domestic supplies are low and consumer prices are high, put- ting downward pressure on rising beef prices. Conversely, it is intended to reduce meat imports when domestic supplies are large and consumer prices low, hence moderating the downside in beef prices. 306 The average annual level of imports for 1968-77 has been officially determined at 120“.6 million pounds, product weight, or as used in the model, 16“6.“ million pounds, carcass weight.7 The 10-year average of domestic production of quota meats for 1968-77 is 23,18“ million pounds, carcass weight. From this quantity, 220 million pounds for average total carcass weight of live cattle imports, is subtracted for that period. This gives 22,96“ million pounds, carcass weight. In the MSUAM the domestic production of beef cows is aggregated into estimated non—fed beef (cows and bulls) produc- tion, although historical data on beef cow production is available. Therefore, a simple transmission equation was econometrically estimated with beef cow production (QCW) as a function of non-fed been production (nrsro'r).3 These calculations for the import quota give the adjusted base quantity of imports for a particular year. However, the key point of reference for imports estimates is the 'trigger' level, which is 110 percent of the adjusted base quantity. It is only when imports are expected to exceed the 110 percent level that the President must 7The conversion from product weight to carcass weight is estimated from data in USDA (1982a, p. 155). 8The transmission equation is QCW - 239.275 + -8498 NFBFQT (2.40) (36.35) 8/m = .021 i2 = .989 DW = '98 no turning point errors 1964-79 307 restrict total imports to the adjusted base quantity level, but to not less than 1250 million pounds, product weight. In the model the endogenously estimated non-fed beef imports are measured and determined by the beef import quota. That is, non-fed beef imports are assumed to be equal to the beef import quota as determined by the counter-cyclical meat import formula. Actual imports may be greater or less than the quota or the trigger level depending upon pol- itical circumstances in the U.S. and/or economic conditions of beef sup- ply in major beef exporting countries (Table 5.1). No attempt has been made to estimate these factors nor project their possible impacts on the actual level of non-fed beef imported by the U.S. It is therefore assumed that beef export supply from beef exporting countries is completely elastic. This implies that exporters to the U.S. will always try to maximize profits by exporting quantities of beef up to the maximum levels allowed under the countercyclical meat import law. Table 5.1, on the operation of the beef quota, indicates that except for three years (1973. 1978-79), of the past ten, and five of the past 18 years, beef imports have been less than or approximately equal to the trigger level of the quota. In five of the past 18 years actual imports have been less than the base import quota. This suggests that the specification of U.S. beef imports to equal the adjusted base quantity may tend to slightly underestimate, on aver- age, actual beef imports. The results of a recent study (Harris and Dewbre, 1983) suggest that trigger level import restrictions are 3(N3 nausea mo newton .m.= .mopvv pe.u*mmo sou» uo>veov um=oun .mucoeootaa «cvoeumoe zeoucopo> on. m.<¢> "camp :. -—-mm .z.z an noon—go; .momp sore Noe-mm .4.au .Louuooeoz eon» nooeeo or Annpnom .z.ov zap pouspuaoeoacaouo mpo>op sumo—Lu ropes an apex.p .m.<¢> cocoon ..~.z .~_—oeum:< m. coco .mcovuo.eumoe oz mcopauvgumue oz mcovuo.eamoe oz Lama's» o>oao m. .ooccoemam coca ounces. oovcoomom cog» .oomoosr mauooa can» Pomona. «Bone 55 meovuoweumoe o: "ooucuomam coca ounces. meovuu.Lumoe o: .ooocoomam ooze ounces? ooocoomsm cog» 39533.. 55 venous, mouoso .m. m. .m. m.5; mouoao mucosa mm.<¢> mouooo ooocoomam cos» venous? mauoaa .Loomveu xopoo m. soomcvz aware: ecu noncou unouxo .m. ecu—ooN so: use ovpoeumo< gas: m. acopuuweumoe oz mcopuoreumoa oz acoraopeumoa oz o ~—.p em. cm. mm.p mm.— so.p op.p mp.p mo.p aw.— cm.— op.p up.p o..— mo.p mm. Na. mm. o m.mm~— «.mmop n.~—mp ~.mmpw a.vuo~ m.~m~p m.~vup m.moo— m.omep n.¢~m— m.o~mp m.o—mp a.pnmp m.pmcp p.~vmp m.~m—p m._o—p o.-m o.~mop o.-~p ~.w—mp —.vemw c.~enP m.m~wp m._omp m.~onp m.pom— —.w~mp o.mmmp m.m¢mp m.opm— m.-cp m.po¢— c.oovp m.mmmp o.s—mp w.mmmp “.mmcp o.oom— m.mcup p.o¢o~ m.mmmp m.mmcp n.smm— ¢.¢mm— N.omv_ «.mmm— ~.epv— o.mocp «.mump m.mcm— m.m~mp ~.wmmp m.o—N— m.nm—p o.pvpp ms_== szsgm.a3 assoaaov mosses =O.Pp.z mam. Nos, .mm, comm_ ass, one. has, cam_ mam, «as, m~m_ New, _aa. ohm. ass. was. “was camp was, . am< as nurses” _a>u4111 zcmao Someone ugomeu co mucoEEou mueoosn poauu< Loom_gp zuwucaao poauo< omam so cease: easmsfiuex . mmo_-momp "some mo mueoos~ go one; ueooen “no: .m.= so copuogooo . ..m apnea 309 unlikely to be imposed during the 19803 and that a free market will pre- vail largely‘ because of the limited ability of suppliers to meet the quota levels. Results of the sensitivity of this assumption is Appendix Table A3. 5.2.2 Dairy Supply and Demand As specified by Wailes (1983) four behavioral equations make up the dairy model. The decision process in the model focused primarily on the dairy heifer equation (Figure 5.2). This equation is determined by a gross margin variable reflecting the profitability (or opportunity cost) of the dairy enterprise, cull cow (Omaha utility cow) price, and dairy heifer numbers lagged one period. Dairy cow inventory is determined by lagged dairy cow numbers and a five-period polynomial distributed lag on dairy heifer numbers. This lag assumes that the age structure of cows in the herd follows a fixed pattern and changes over time only in terms of changes in the dairy heifer inventory. Milk production per cow is estimated exogenously to the model and is multiplied by dairy cow numbers to generate total milk production. The highly aggregate demand component is completely described by a single milk price equation. This equation is determined primarily by milk disappearance, the ratio of lagged milk price to the milk loan rate, and the deflated milk loan rate. 5.2.3 Pork Supply Figure 5.3 presents a flowchart of the pork supply relationships as estimated. Also shown in this figure are broiler supply, and egg and 310 a gun mm“; LP /\ 1 :o.uuanosa EL. uc—IOLLou u m=.xoouou m o z—z: F—mu :.5 m cameos sun :0 w a ”500.5 . . .5 mecca: no. A—ooom can hoses» can a—oean .o——osn we. zen; .m a so censure—u n m a; .u co.uu=eogn Lo—.oga 353:: co.uo:ectn Lone; cuzo—z hwztah mo.ea . . . Loo :o_uo:v ceou -otu am e: 09.nn poosxom oo_to gap—op» . oo.Lo zuztoh . pu>vz atom Lt vcosoo can Apnoam hogan» Lou =o_uu=u .95 so u x.oo:m to—voeo Apooom moon Eula: 5&3. a 38:53 Luz :o_uo=c -oen even one oo.tn zoos newesw uvuwwzon po>og ELou e co.aos=moou BS 6835 ou—La :Lou cu a=.3ottuu “gem Poaoh usuassz 10m —euo» 2.3.5: Lavoz =a.u_g osucz moves ctou-moz z $238.5 IIIIIIIIIII :o_uu=oogo zoom—oveeu zoos mow Laooaz : :35. to; menace we. Aponom mum 311 turkey supply and demand. The submodels linking these equations with other segments of the model are shown. Production of pork is determined by the size of the pig crop and the weights to which they can be economically fed for slaughter. Separate equations are used to represent fall farrowings and spring far- rowings. Fixed technical coefficients are used to obtain litter sizes per sow. Pork production is then the product of slaughter weights and hogs slaughtered. The spring sow farrowing equation is recursively determined by lagged sow numbers, corn price, soymeal price and the pork price. Fall sow farrowings are determined by a similar equation. Included into these specifications is the asymmetric response of hog prices during the contraction and expansion phases of the hog cycle. Pork production is estimated as a function of lagged pork production, last fall and this spring's farrowing, and the change in the hog-corn price ratio. 5.2.“ Poultry Supply Separate turkey, broiler and egg production equations are estimated in the poultry sector model (Figure 5.3). The model does not capture the complexity or rapid structural change that has occurred in this sec- tor over the last 20 years or so. The primary reason, at least ini- tially, for including poultry in the model was to identify feed consump- tion by this sector. This has been achieved reasonably well. Despite the annual specification of a sector, where supply turns over more than once a year, year to year variation is explained very well. 312 5.2.5 Livestock Feed Consumption Figure 5.“ illustrates the economic relationship among wheat and coarse grains fed to livestock and the relevant linking submodels. The livestock-feed relationships link the livestock sector to the crop sec- tor. Feed consumption is estimated for feed grains (corn, sorghum, bar- ley, oats), soymeal and wheat. Wheat fed to livestock is estimated directly whereas feed grains and soybean fed are based on grain or meal consumed per standardized animal unit. Wheat comprises a relatively small share of livestock feed in the U.S., where the amount of wheat fed depends primarily on availability and the price relationship between wheat and feed grains. The number of grain consuming animal units is derived from numbers of livestock in each category, weighted by feed consumption per animal type. The feed weights were obtained from USDA sources. Grain consump- tion per animal unit is estimated from livestock and feed prices. Total feed grain consumption is then obtained as the product of feed grain consumed per grain consuming animal unit and the number of grain consum- ing units. Soybean meal consumption is obtained in the same way as feed grain consumption. Clearly, some aggregation bias is likely to occur in estimates from this simplified feed consumption model. Feeding rates are not fixed in reality but differ across livestock classes. To date, development of a more disaggregated model has been prevented by an absence of historical data identifying feed consumption by livestock type or feed. Feeding rates and hence grain consumption also differ across time for given animal types. Livestock and feed price relationships do not capture the 313 h Louoom zuopmo>wz l _oaoz-n=m cameos woo: praaoz use: Posse< as? usamcou cwoeo Loo ooeomcou .m.= so co.uoe=m:ou mcpoemooom one pooz: so aeozozopm e.m oeouvm ourea .paosxom ooveo :Lou :ovuusoogo ouo o coruosuoeo Lopvoea o eovuuoooeo zoom covuoaooeo moon ooeueo: covuosooeo soon now meoewoe arson mzoo moon mzoo ALPoo zuoumo>wz mo gossaz mcwoemoooutwl zuoumo>vz so tossm -cou mcwoea -ooom p33 mpoooz-o=m —ocovua=eoa=~ Pmosaom ovate ovumoeoo move: poemc< newssmcoo :woew mo consaz opuoz moves mmoonumogz .>Fz 314 total variation in feed consumption caused by cycles in livestock pro- duction. Much of this variation in feed grain consumption is of a short term nature. Therefore, the above method of estimation should capture satisfactorily long-term variation in feed grain consumption. Because livestock estimates are on a calendar year and the crop sector is on a crop marketing year, synchronization of the interface of the two sectors was necessary. To achieve this, livestock numbers in t+1 were estimated. Since all explanatory variables for livestock sup- ply are predetermined, they can be led forward over a period. 5.3 Model of U.S. Retail Meat Demand In the foregoing sections an overview of the MSUAM was presented. Model components of 0.8. livestock and livestock feed consumption were also outlined since these submodels form an integral link with the retail demand for meat. In this section the retail meat demand model is described in detail. In Figure 5.5 a block diagram of the U.S. retail meat demand model is presented. A block diagram embodies a complete model description and has a one-to-one relationship with the set of equations describing the mathematical model. All relationships among variables are shown either as a summation, 2, or multiplication, n, except where a function is generated. In this case, each functional equation is symbolized and denoted by F1 for i = 1, 2, ...,11. As in previous diagrams in this chapter, flow or stock variables are shown in rectangles, while circles represent other variables and influences. The FORTRAN coded version of this block diagram is contained in Appendix B. s [aponqng 4 loans, poison ; ‘1 315 33... .2 22.8 :32. a... 3 8.2:. .8; a.» 9.3: 5%.. .8 5...... 1 255.. + 5.53 m , m as; S m n. m: m a n. «it. is... a 52.. A.— E a .5 sit a o «unwound i it; a E a .88. the 885 . ..a 5 fl 6:: ~ .1 to: u . 1/4 a nail . ...s . ,. . 59: so . N s. A. I; F. 316 The retail demand model contains 11 behavioral equations each of which have been estimated by ordinary least squares techniques. Four of these are price dependent demand equations, each of table beef, ham- burger beef, pork and broilers. These were estimated in the previous Chapter “. With one exception, the other estimated equations are mark- eting margin or by-product value equations, which provide the transmis- sion between retail and farm level product prices. The exception is an equation which estimates the ratio of slaughter of fed steers and heifers to all steer and heifer slaughter. Its purpose is to provide a more accurate determination of the quantities of fed and non-fed beef production. In addition to these estimated equations, there are ten linkage identities which either transform farm product quantities produced (car- cass weight) to retail product quantities consumed (retail weight) or transform retail prices to farm prices. Variable codes and units of measure of variables used in Figure 5.5 and in specifying the retail meat demand model are presented in Table 5.2. Some of these variables have been defined elsewhere, however, to facilitate the following dis- cussion, these are included here. A shortcoming of the specification of fed and non-fed beef produc- tion in the original MSUAM, was the aggregation of all steer and heifer slaughter as fed beef. Discussion in the previous two chapters highlighted some of the problems arising from this aggregation.9 A por- tion of steers and heifers is raised primarily on grass and not in 9The nature of this aggregation bias was the subject of a recent paper by Martin (1983). 317 Table 5.2 Variable Codes and Units of Measurement Used in Specifying the U.S. Retail Meat Demand Model “33:19 Variable Name Unit of Measure BCALFP Kansas City feeder calf price, medium no. l $/cwt steers, real BFCOWP Beef cow price, Omaha commercial, real $/cwt BPVBFR By-product value beef choice yield, ¢/lb grade 3, retail, real BPVPKR By-product value pork, retail, real ¢/lb CHIKQT Chicken production, broilers on ready- million lbs. to-cook basis CHIKPT Chicken price, farm level, real t/lb CHNGBRR Change in chicken price, farm level, real % CHGBCP Change in beef cow price, Omaha 1 commercial, real CHGFBPT Change in choice Omaha steer price, real % CHGPKP Change in pork price, farm level, real % CORNPT Price of corn, No. 2 yellow corn at ¢/bu. Chicago market, real CPI Consumer price index index DCBRR Domestic consumption broilers, retail weight million lbs. DCFBFR Domestic consumption fed beef (choice), million lbs. retail weight DCPKR Domestic consumption pork, retail weight million lbs. DCBRC Domestic consumption broilers, retail lbs. weight, per person DCFBFRC Domestic consumption fed beef, retail lbs. weight, per person (table beef) DCNFBFRC Domestic consumption non-fed beef, R. W. lbs. per person (hamburger beef) DCORMRC Domestic consumption other red meat, R. W. lbs. per person (pork, veal, lamb) DCPMC Domestic consumption poultr meat, R.W., lbs. per person (chicken, turkey DCPKRC Domestic consumption pork, retail weight, lbs. per person DICR Disposable income per person, real FBEFQT Fed beef production, steers and heifers, million lbs. dressed weight FBEFPT Fed beef price, choice Omaha steers $/cwt. GFVBFR Gross farm value, beef choice yield, ¢/lb. grade 3, retail weight, real GFVHMBR Gross farm value, commercial beef cow, ¢/lb. retail weight, real GFVPKR Gross farm value, pork, retail weight, real ¢/lb. MMBRR Marketing margin, broilers, retail weight. ¢/lb. real 318 Table 5.2 (Continued) Variable Code MMBFR MMHMBR MMPKR NFBFIMT NFBFQT NFVBFR NFVHMBR NFVPKR PORKQT PORKPT RPBFR RPBRR RPHMBR RPPKR RSLFSHSH Marketing margin, fed beef, retail weight real Marketing margin, commercial beef cow, retail weight, real Marketing margin, pork, retail weight, real Variable Name Non-fed beef imports, carcass weight equivalent Non-fed beef production, cows and bull slaughter, dressed weight Net farm value, choice beef, retail weights, real Net farm value, commercial cow, retail weights, real Net farm value, pork, retail weights, real Pork production, liveweight, marketing year, Dec.-Nov. Pork price, 7 markets, barrows and gilts real Retail price, table beef (choice grade), real Retail price, broilers, (young chicken), 4 region average, real Retail price, hamburger beef, real Retail price, pork, real Ratio slaughter, fed steers and heifers to all steers and heifers Unit of Measure ¢/lb. ¢/lb. cllb. thousand tons million lbs. ¢/lb. ¢/lb. ¢/lb. million lbs. $/cwt. ¢/lb. ¢/lb. ¢/1b. ¢/lb. ratio 319 feedlots. Such aggregation, therefore, is likely to result in a sub- stantial underestimation on non-fed beef production. It was also recog- nized that although range-fed cattle typically contribute some cuts of table beef, fed steers and heifers contribute significantly to processed/ground beef production directly, or by supplying fat trimmings for mixing with lean ground beef. Ryan (1980) reported that 23.2 percent of all steer and heifer car- casses are of processing quality.1o Although the general order of mag- nitude of this percentage appears to be in the correct "ballpark", the degree of accuracy of this figure appears spurious. This percentage can be expected to vary with changes in both input and output prices faced by the cattle feeder, and to vary with movements in the cattle cycle, grading changes, and with various shifts in retail meat demand. In Appendix Table A2 is contained the derivation of the data series of fed and non-fed beef supply and of table and hamburger beef consump- tion used in estimation of retail beef demand in Chapter “. In essence, the quantity of all steer and heifer beef (federally inspected and other) is adjusted by the ratio of slaughter of fed steers and heifers to all steers and heifers (RSLFSHSH). [This ratio, RSLFSHSH, is equal to 11 the number of marketings of cattle on feed in 23 states (MRCF23) 10Using this figure, Martin (1983) estimated that over the 1961- 1979 period, fed steers and heifers contributed an annual average of 3“1“.“ million pounds of lower quality beef compared with 1675.2 million pounds a year of high quality cuts produced by non-fed steers and heifers. It is not stated if this is carcass or retail weight. Nevertheless, the potential bias of including all steers and heifers in the fed beef category may be only half as great as if it were excluded. A bias occurs whether non—fed steers and heifers are included or not into fed beef production. However, the bias is greater when it is included. 320 divided by the number of all steers and heifers slaughtered (SLSR + SLHF). Therefore, the quantity of fed steer and heifer beef was reduced by the ratio and non-fed beef production was augmented by the non-fed steer and heifer beef residual. Therefore, to bring estimates of fed and non-fed beef production provided by the original MSUAM into line with the more complete estimates of table and hamburger beef production/consumption in the retail demand equations, RSLFSHSH was econometrically estimated as a behavioral equation (see function F“ in Figure 5.5).12 A priori, the ratio of the slaughter of fed steers and heifers is determined primarily by the prices of inputs to the fed beef production process, i.e., mainly feed costs and feeder cattle, and the relative prices of outputs, namely choice beef and beef cow prices. Structural capacity of feedlots may be captured by lagged production or some other proxy variable for capacity. Two versions of this ratio equation are shown below in equations (5.2) and (5.3). 11 Data on MRCF23 (USDA, 1982a) has been discontinued and replaced by data on marketings of cattle on feed in 13 states. 12Because of the manner in which fed beef production, FBEFQT, is estimated in the MSUAM (i.e., slaughter weights times numbers) and because of FBEFQT's limited involvement in other estimated equa- tions, the adjustment achieved by this steer and heifer ratio equation, in large part, overcomes the biasedness inherent in the original MSUAM specifications. 321 RSLFSHSH . '418 - .075 coanprt_l - -ooa BFCALFPt_1 (5.2) (7.35) (4-34) (6-67) , FBEFPT . - 102 ifiififif' + 960 RSLFSHSHt_l (3-17) (20-08) 31m = .020 i? - .96 . DW = 2-66 ‘ 1960-82 RSLFSHSH - '276 - ~113 CORNPT - .004 BFCALFPt_1-r]:(K)FSLFSHSHC_1 a/m - -025 i2 = ~94 DW = 2-46 1960-82 (5.3) In the estimated model, corn prices, CORNPT, and beef feeder calf prices, BFCALFP, were chosen to measure input costs in the previous period. Current corn prices would capture the short-run impact of changing input costs, however, the long-run .impact is more closely reflected by corn prices lagged one period.13 A positive sign is expected on the RSLFSHSHt_1, as a reflection of the longer term growth trend in fed beef production. The ratio of fed beef price to non-fed beef price, lagged one period was included to reflect output prices. As the price of choice beef rises, the profitability of fed beef production increases, and hence the steer—heifer ratio variable can be expected to rise. Therefore, the expected sign of this variable is positive. How- ever, a negative sign was obtained in each formulation of this equation 13The coefficient on current period corn prices in equation (5.2) is also negative although smaller in magnitude and with a smaller t value. 322 (5.2).1u For this reason, this output-price ratio variable was dropped and equation (5.3) was used in the retail meats model. This equation performs an important role in the retail demand model. Essentially it endogenises the substantial impact of changes in fed beef production, at the margin, on non-fed beef production during periods of expansion and contraction in the beef cow herd. Following the equation fer RSLFSHSH, six identities are estimated. Those identities provide values for the domestic consumption, retail per person, of each explanatory quantity variable used in the price depen- dent retail demand equations. Although it is not shown specifically in Figure 5.5 each of these consumption variables is solved simultaneously in the overall simulation model. Details of the factors used to make conversions15 of product from live to retail weight are contained in the FORTRAN program presented in Appendix B. Hence table beef, hamburger beef, pork and broiler consumption are simulated endogenously and are estimated directly from the simulation model. Domestic consumption of red meat, DCORMRC, and of poultry meat DCPMC, are only partially endogenously simulated in the model. In other red meat, pork is endogenous, but veal and lamb are not, at least at the retail level. Veal and lamb consumption are set at a constant average level of 1973-1982. For poultry meat consumption, other chicken meat is 1“Simultaneity bias between fed beef prices and the steer-heifer slaughter ratio may be partly responsible, i.e., an error in vari- ables problem. 15These conversion factors were obtained either directly from USDA (1979) or have been indirectly estimated from data in USDA (1982a). 323 not determined by the model and likewise it is set at the constant aver- age level of 1973-1982. Turkey consumption is endogenously determined by the model at the farm level, although not at the retail level in the MSUAM. Therefore, turkey production was converted to retail weight and to a per person basis for inclusion in the poultry meat consumption variable, DCPMC. The retail price dependent demand equations are central to this retail component of the MSUAM. These estimated equations were presented, along with test statistics, in Chapter “. For convenience of exposition, however, they are presented again in Table 5.3. The table beef price equation is specified with a second degree Legendre polyno- mial in own-quantities consumed. The hamburger price and broiler price equations each have first degree Legendre polynomials in own-quantities and disposable income per person. The preferred pork equation is a traditional fixed-coefficient model: the null hypothesis of fixed param- eter coefficients could not be rejected. Justification of the specifi- cations and estimation of these equations is contained in Chapter “. The two other types of behavioral equations contained in the retail meat demand component are the marketing margin equations for each retail product and the by-product value equations for beef and pork. Estimated equations for the meat market margins, the difference between the retail price of a particular meat and its farm value, are presented in Table 5.“. These equations allow for the absolute margin via the current farm price variables and allow for the percentage margin via the annual per- centage change in the farm price variable. The current farm price vari- able can be expected to have a positive sign on its coefficient as 324 Iowan common one now ooumaauoo mos coauoouo xuoa onh .mN.¢ manna mo Amy ouosuoow com H.n manna aw ao>am mum mooauwanuov ovoo oanmqwm> A .mmmalcnma aouw voumawumo .m.e edema aw coaumsvo xuoa on» news vouooaoo on has one NmmH .sao>euooeaoe .A~.s one n~.s .su.s mosses ea venues nooo mum mom «maaionau ooauoa one now ooumawumo onus mcowumavo wonqoun was moon wowuanao: .woon canon 059m e~.~ as. one. ‘ Anm.ao aflo.mo Aoo.mv Aem.so Home 80. -68 8o. +828 new. +238 258 .. AsH.~v Aa~.sv Aom.~v Ash.mv omumoa NON..-UMhmmzon HO¢...QMhmmUQ Nan...Nm~.no I mammm oofium HOHHoum am.~ so. nee. Ako.nv Awm.mo 68 So. +938 an; .. Amm.mv Aso.fio Aflo.flo Aas.ev unseen osso.d..oeemezoo Ham...oaeaaoo sma...amo.~a . memes coupe seem an.8 so. see. Amo.Hv Ae~.mv Aao.mv Aam.v Ede So. .183 25. +338 So; . ogooo «.3. I hoe.mV Amm.oav Aen.ev Amn.~v madam was..soemnezoo NH».H..oaeoeoo omn.8..~oo.sm a ensues cause some nonsense: He.H we. «so. Aum.No. Asn.nv Asa.o Ash.ao 88 and + 88 ~88 .. Max:368 2:. 1 93828 8.... H 1 . “so.mV Ao~.~V Awe.mo Aao.mv memes an“...mesoe om~.H;Tommmeoo nem.~..s-.ms~ . memes euuee omen «Heme .3.9 Na. axe n.mmaouumoum voumaaumm manmwum> uaovcoeoa moaumaumum mwmaasm Homo: cameos umoz .m.= may ow meowumavm oowum umoz Hfimuom m.n ounce 3215 Semis . . . 8.3 $93 Sod 8.3 SO: :33: as N no «mo «00.: one. moo. NmH. mom.a uoaaoum . . . $9: 2.3 5.3 8.3 5.2 288 on N i So .50. o2. 23.- an. 02.8 58a: .38 go :«wumx . . Ace.eo Amo.nv Ao~.sv Aeu.so coon 2.5% . 8.2 5.8 Rid and a: 583. 2 N 3. 20. 8o. Sn. 88.- . 8m. Se..- coon e8 .38 mm Em 588 6.88 888 58a: 28:8 8.30.2 .5882 Shanna 8.856 8.528 .588 8888 8388 nowmog oowum sham :« owamzu ooaum Bush ucouuao pomnmsoush moanoaum> mwoumamdmwu oHpmauw> ucooaomon Nwmauooma "mcoaumocm moflwumx ocean afiooozuauom oouoaaomm c.m oanme 326 margins tend to adjust over time with trends in farm level prices. A negative sign is expected on the change in the current farm price, since short-run increases in farm prices tend to decrease the margin, reflect— ing the hypothesis of price leveling or smoothing practiced by retailers. Lagged margin is included to reflect the longer term growth trend in the real cost of providing retailing and wholesaling services and to reflect the upward trend in value added in these services: hence a positive sign is expected. Some studies use real wage rates to cap- ture this variation in margins, however, lagged margins would appear to have greater dynamic content and are therefore preferred. Two conflicting hypotheses may be proposed fer the coefficient sign on throughput, as an explanator of margins. One hypothesis is that mar- gins decrease with increasing throughput as overheads are spread (Reeves, 1979). This is the cost-side explanation. Alternatively, increases in throughput are associated with demand-pull pressures. As demand strengthened, more livestock are slaughtered (following the ini- tial stock retention phase), and margins widen in response to increased profitability in the subsector. Therefore, a positive relationship exists between throughput and margins. Some cost gains are achieved with greater throughput, however, the stronger influence toward widening margins is the increased throughput being drawn into slaughter plants. In the non-fed beef margin equation, throughput was not included. It has a positive sign but was not significantly different to zero. In slaughter plants, both economies of size and the throughput related variation in margins is more likely to be related to their primary func- tion of fed beef slaughter. In many respects non-fed beef slaughter will more likely fill a residual role in slaughter plants. 327 The above hypothesized pattern of signs on regression coefficients is revealed in each of the estimated equations, except for broilers. The broiler subsector, unlike the beef and pork subsectors, is very highly coordinated and concentrated. Moreover, price levelling is not expected to be important in such an industrialized subsector where the same participants often control production, distribution and marketing (Reimund, Martin and Moore, 1981). Therefore, a positive sign on prices and changes in prices of broilers would reflect the ability of producerawholesalers to influence their 'profit margin' positively with own-prices. Throughout in this equation reflects the increasing economies of size experienced in this subsector over time. By-product value equations were estimated for beef and pork to pro- vide the transmission between farm and retail prices. In years of strong product demand by-product value is a more important component, in absolute terms, of the total livestock carcass. By-products are, there- fore, estimated as a fUnction of current and past retail prices and the CPI. The same by-product value equation was used for fed and non-fed beef prices. The estimated equations for beef, BPVBFR, and pork, BPVPKR, are presented below in equations (5.“) and (5.5). severe - -4-931 + -l98 RPBFR - -080 RPBFRt_1-°46S CPI (5.4) (1986) (6‘70) (2-58) (1-70) 8/m - '118 i2 - '68 pw - 1°65 1960—82 BPVPKR + .200 + -o79 RPPKR - .024 RPPKRt_l - -517 CPI (5.5) (~20) (5-36) (1-65) (3-39) 81m - -139 i2 - ~68 on a 1.56 1960-82 328 To complete the integration of the U.S. retail meat demand model with the farm level components of the MSUAM, price transmission identi— ties were required for each of the four retail products. Hence, farm product price is equal to the retail price less the marketing margin plus the by—product value (in the case of beef and pork) divided by a factor for converting retail weight equivalents to farm product equivalents (except for broilers which is already on a ready-to-cook basis). 5.“ Prqjections of Key Expgenous Variables Simulations of economic and policy scenarios regarding the lives- tock subsector in the U.S. are dependent upon the assumption underlying projections of certain key exogenous variables. In this section the future trajectories of these variables in the model are given. Assump- tions made about these key variables are drawn from those made in Michi- gan State University's 1983 Spring Report on the long-term forecast of U.S. and world agriculture (MSU, 1983). Details of these assumptions and the source of the projections are not repeated here. Nevertheless, some aspects of these projections are given. Table 5.5 contains actual and projected values of U.S. population, inflation, interest rates and real disposable income per person. The forecast on inflation reflects the low inflationary growth expected over the next few years. Inflation is expected to increase gradually from 5.5 percent to 7.0 percent for the 1986-1991 period. Steady real disposable income growth is assumed over the medium-term. 329 Anna—v 2m: "muszom om.m $0.5 cm. omme pm.N oo.p— on.¢ v.0mN pmmp om.m 05.5 mm. come ~¢.~ oo.—p m¢.¢ —.mem camp cm.m 00.x om. omov em.~ oo.—P m~.¢ m.me~ amm— om.m om.c om. Oman —N.N om.o— mm.m v.m¢~ mam— om.m co.m mm. omum en.m oo.c— mm.m o.—¢N Ramp om.m om.m om. comm mo.m om.o~ ¢¢.m m.mm~ amm— oo.m om.¢ w¢. comm Po.m om.m o~.m m.om~ mmmp mn.m oo. ma. omen NN.N om.m Np.m N.¢mm vamp ON. ow.m mm. oemm om.m om.o~ oo.m o.~m~ mmmp mo. 1 mp.m Na. Chum m¢.v om.N_ mm.N m.mNN Nmmp po. 1 oe.op mm. psmm m©.m Fm.mp NN.N N.NNN —mmp mm.mu om.mp m—._ NmNm m~.m ¢N.Np n¢.~ o.m- omm~ AN. - mm.__ mz._ oomm mo.m so... AP.N o.MN~ meme meum no.5 op.p mmmm c~.¢ F—.m mm.p m.o- wump N~.~ m¢.m No.p m—mm so.m mm.m mm._ p.mpm unw— No.~ ek.m ma. meum so.m m_.m Pa.p a.mFN okm_ cc 1 ep.m eo.— ompm pm.m a~.o —m.— m.m~N mnmp u a a cameo; N z oo._ mcowp_wz Loo» \m «as esa «mom. secs ozooc .11. zuwn Hem o_eemoameo Mao -aflmmwa _eom .=Omeoa ones muse xooes goes so; oeoocH umocoucH amoeoucu oopce -opzaoe "oucozu mo mouoz owmomoom_o pooz oases nozzmcoo cowp_>wo me~-mnm_ ”oeoo:H opoomoamwo ecu mouoz umocoucz .cowpo—ecfi .coFHapzooq .m.= mo mos—o> oopoonoee oco .oauo< m.m o2: 330 International income growth assumptions are presented in Table 5.6 for the major grain exporting and importing countries or regions. Growth is expected until 1987 when long-term growth rates are assumed to be achieved. The lower near-term growth rate is to reflect a slow recovery from the current recession. Medium population estimates based upon fertility rates, mortality rates, and immigration patterns affect a gradually declining population growth rate throughout the world (Table 5.7). The 1983 base yields and growth rates are reported in Table 5.8. Yield assumptions are obtained from historical trends, USDA projections and from observations by agents in the field. The 1983 base yields are the trend value for 1983. Expected yields for 1983 are substantially higher for the U.S. due to the PIK program. Yields are also expected to be higher in Australia following the recovery from a drought in that country. Except for the U.S., total cropland used for wheat and coarse grain production is exogenously determined in the world component of the MSUAM (Table 5.9). Growth rates in the cropland base are based on his- torical trends and authoritative opinion.16 5.5 Validation of Model Validation of the MSUAM has been conducted elsewhere aha will not be repeated here.17 In this section model validation willlfocus on the 16Further explanation on the nature of these and other general as- sumptions in simulations of the MSUAM, especially in the grains component, see Schmitz, McCalla, Mitchell and Carter (1981, pp. 229-238). 1 7See Wailes (1983), Schmitz, McCalla, Mitchell and Carter (1981, PP. 227-229). and various MSU Agricultural Model QUarterly Reports. 331. Table 5.6 International Income Growth Assumptions: 1975-1991 fZREaTIGrowthrRatesa R9910" 1975-81 1982 1983-84 l985-86 1987-9l % per year Canada 3.60 -1.75 1.55 3.25 3.70 Australia 2.80 3.00 2.82 3.25 3.70 Argentina 2.80 ~4.50 1.80 2.50 4.00 Brazil 7.90 0 .70 4.00 5.00 Soviet Bloc 2.80 3.00 3.00 3.00 3.00 Developed Markets 3.90 1.50 2.76 2.90 3.15 Less Developed Markets 5.30 1.60 3.63 4.20 5.00 . Mainland China 3.08 3.50 3.25 3.00 3.00 Source: MSU (1983, p. 142) “Index of GNP, 1970-100. Table 5.7 P0pulation Estimates for 1982 with Projected Growth Rates for 1983-1991 11...... p.273??? {.5328 United States 229.80 .96 Canada 24.60 1.33 Australia 14.80 1.08 Argentina 27.70 1.08 Brazil 135.00 2.19 Soviet Bloc 382.20 .71 Developed Markets 532.00 .50 Less Developed Marketsa 2340.00 1.93 Source: M50 (1983, p. 143) aIncludes Brazil. 332 Table 5.8 Wheat, Coarse Grains, and Soybean Yield Assumptions: 1983-1991 T983—8ase 8ush§l s/Acre. Commodity Region Yieldsa Growth/Year Wheat United States 38.00 (34.80) .60 Canada 29.25 .45 Australia 19.65 .15 Argentina 25.60 .30 Soviet Bloc 28.72 .32 Developed Markets 54.99 1.19 Less Developed Markets 21.85 .45 Coarse Grainsa United States Corn 118.00 105.35) 2.35 Oats 55.40 54.60 .40 Barley 49.30 49.80 .80 Sorghum 60.30 59.40) .90 Canada 45.18 .88 Australia 21.90 .20 Argentina 49.92 1.32 Soviet Bloc 32.38 .48 Developed Markets 58.04 1.04 Less Developed Markets 19.57 .27 Soybeans United States 32.50 (31.10) .30 Brazil 26.20 .30 Source: MSU (1983, p. 144). aCoarse grain yields are based on a 56 pound bushel. b1983 base yields are trend yields, not expected yields,.except for the U.S. Base yields for the U.S. are expected to be higher due to the PIK program. U.S. trend yields are in brackets. Actual and Projected Cropland Harvested Area 333 Table: 5.9 of Wheat and Coarse Grains: 1975-1991 'Less De- De- veloped veloped Uniteda Aus- Argen- Soviet Mar- Mar- Year States Canada tralia tina Bloc kets kets mpauf -------------- million harvested hectares ---------------- . 1975 261 17.3 12.5 11.2 146.2 47.8 155.5 1976 271 19.0 12.9 13.0 146.2 48.0 162.0 1977 277 18.0 14.3 9.8 148.0 46.8 159.0 1978 267 17.8 14.9 10.8 147.0 47.5 160.0 1979 275 17.0 15.3 9.4 144.0 48.2 162.0 1980 287 19.0 15.8 11.4 145.0 48.2 163.0 1981 296 20.5 16.2 11.1 145.0 48.2 165.0 1982 296 20.6 16.6 11.2 143.0 48.2 166.0 1983 272 21.4 17.0 11.3 144.0 48.2 168.0 1984 265 21.0 17.3 11.4 145.0 48.2 170.0 1985 277 21.1 17.6 11.5 145.0 48.2 172.0 1986 282 21.2 17.9 11.5 146.0 48.2 173.0 1987 288 21.3 18.2 11.6 147.0 48.2 175.0 1988 290 21.4 18.5 11.7 147.0 48.2 177.0 1989 293 21.5 18.9 11.8 148.0 48.2 178.0 1990 295 21.6 19.3 11.9 149.0 48.2 180.0 1991 299 21.7 19.8 12.0 150.0 48.2 182.0 Source: MSU (1983, p. 145). aIncludes soybean and cotton area. bmillion planted acres ___...—u-——Iu Ill—III" —"—' 334 retail meat demand component and on farm livestock product prices which are estimated by the retail model. Table 5.10 compares the predictive accuracy of (a) the individual equations when simulated as single regression models, (b) a dynamic simulation of the model when actual lagged endogenous variables are used, and (c) a dynamic simulation of the model when lagged endogenous variables are produced by the model. These dynamic model results are generated by the full simulation model when all equations are allowed to interact, in contrast to (a) above when equation performance is viewed in isolation. The criteria of perfbrmance used in validation of the model are Thiel's inequality U-coefficient and the root mean squared percentage error. The U-coefficient takes values such that O 5_ U 5_ 1, where a value of zero implies a perfect forecast and a value of one implies a complete lack of relationship between the predicted and actual values.18 Coefficients presented in Table 5.10 for each model indicate that the equations may be used to project changes with a reasonable degree of confidence. The root mean square error percentage is the root mean 1 8The coefficient is calculated from the formula 2 2 2 U = 20>i - A1) 20.1) + 2(Pi) n n n where Pi = projected value, A1 = actual value, and n = number of obser- vations. 335 Table 5.10 Model Validation: Thiel's U and Root Mean Squared Error as a Percengage of the Mean: 1975-1982 ‘Ifidividua1"Equaticns IDynamic Model Act. La qed Endo . Pred. La ed Endog. Thie175 TGEEV Thiel's RM§7 Thiel's RMS7 U Mgan U Mean U Error RSLFSHSH .0125 .00 .0157 .03 .0111 .00 DCFBFRC .0036 .14 .0284 8.33 .0207 .91 DCNFBFRC .0110 .63 .0292 4.33 .0179 .04 DCPKRC .0117 .52 .0230 6.29 .0190 4.69 DCBRC .0131 .00 .0131 .00 .0131 .00 RPBFRC .0248 .00 .0550 1.38 .0536 1.10 MMBFR .0231 .99 .0364 .04 .0231 1.61 BPVBFR .0591 .25 .1634 .11 .0146 .51 RPHMBR .0361 .03 .0648 -4.17 .0636 2.93 MMHMBR .0437 2.04 .0750 1.42 .0757 .05 RPPKR .0194 .04 .0431 9.48 .0402 7.21 MMPKR .0210 1.34 .0189 1.14 .0139 .20 BPVPK .0536 .43 .0847 3.54 .0808 2.33 RPBRR .0204 .06 .0545 1.10 .0529 .72 MMBRR .0191 .30 .0475 5.15 .0431 1.04 CORNPT .0510 .48 .0991 2.20 .0918 1.79 PORKPT .0108 .02 .0708 17.02 .0614 10.30 FBEFPT .0017 .00 .1091 .83 .0955 3.95 BFCOWP .0000 .00 .1258 2.20 .1251 2.67 CHIKPT .0000 .00 .0699 8.06 .0715 6.91 336 square error divided by the mean of the actual values over the period, expressed as a percentage. This measures how well the model equations, on average, are tracking historical values. The results of greatest interest in Table- 5.10 are those of the dynamic model simulations. Except fer the farm price of pork, all variables display a percentage error of less than 10 percent, with most variables having a percentage error much less than five percent. Surprisingly, the model using predicted lagged endogenous variables perfbrmed better far most vari- ables than the model using actual lagged endogenous variables. In later simulations, however, i.e., beyond the historical period, actual lagged endogenous variables are used while they exist. Validation was conducted over the relatively recent historical period, 1975-82 because of the importance of a model's ability to track the most recent years of history. Overall, the ’model perfbrmed well with respect to their criteria, especially given the market volatility of these years and the fact that most of these behavioral equations were estimated using data over the 1950-82 period. The model can be expected to provide a reasonable representation of the behavior of retail meat demand for use in policy analysis. Of course, model testing is not complete with dynamic within-sample simulation and the calculation of certain performance criteria, above. At this point a model is only conditionally accepted. The model must also perform satisfactorily and coherently with respect to impacts of specific changes in key exogenous (and endogenous) variables. That is, 337 variable responses to such impacts must have coherence and correspon- dence with reality. This sensitivity analysis is conducted in the fol- lowing section. 5.6 Base Level Projections of Endogenous Variables: Abtual and Simulated Results In Figure 5.6, simulated values for key endogenous variables in the MSUAM are shown for the period 1975-91, along with actual values from 1975 to 1982. The projections are conditional upon several assumptions regarding the nature of future changes in population, income, crop yields and other key factors. Also, this simulation represents the base model run fer what is considered a most likely scenario of future market conditions. Results from the base model take into account the 1983 Payment-In-Kind (PIK) program for wheat and corn in the 0.8.19 Hence, the base model reflects a relatively weak grain market, especially for corn, with low prices and high yields. The impact of the coarse grains market on the livestock subsector will be discussed below. In general the behavior of the model was satisfactory. Simulated values of the ratio of slaughtered fed steers and heifers to all steers and heifers slaughtered, tracked closely the actual series of data from 1975 to 1982. The behavior of simulated values of fed beef production also approximated actual values reasonably well. Note that beef cow numbers, non-fed beef, pork and broiler production have been exogenized 19An analysis of this program using the MSUAM is contained in Ferris et al., (1983). 0’83 ‘. .IBNPr—I OOICO‘. ‘0 one—rr—a “9869'. ‘6 CID-PF“: 338 “If C“ “II O—HBD .fl.=.'. ‘. ...-‘PP-I’ .086.‘ "O Uta—rr—a IMHO U «mm: W '60 81!!" M mass 10 It 81(188 N “If“. 0.“ T 0.. O.“ 0.. 0." 0.1 Do“ 1 0'. r j —; r I— r i :m 1.77 1”. 5.5 t“! t“ I“? I“. I”! M wow-no “I, mum 10 I W /-'-\. / l II C 1 / I I l . ‘ .\5 I, \ ” \.-~o 2 a 0 I— I T I— 1 l i I915 I”? I.” I.I 1”! 1” 5m 1”. I”! run CHIN! PM?!“ u .. 0"---1 I: . " so J U a I j— 1— l I 1 1 fi 337‘ 1.77 1.7. 1.1. I”, ms I“? I... I”! run Figure 5.6 .0 .. I! . u + ..'... \\ Al .1 3 \~. I a q '1' W I 0‘ .1 '1; a d I W q 'I’ 'I ” 1 coal. 3 J u d a a a. ' I— j I 1‘ r 5 1m um I”. lul was I“. I“? 1”. 3“! an no I" With 84 1 n 1 'IO.... 4'". “‘5. i l .’ I I 1' 'a' o‘ .J' I {I ""' CIWYII 1. I T r j I F I 197‘ I." 1m 1.01 I”: I“ I”? I”. lfll run not PM?!“ 3‘ .. x. J a's‘ ’0 \‘ 0’ 22 J I" \" ~...-...-’I I . I I 'I 20 a l ,9 .. l {I . I f f j—7 fir“ f I 197‘ I." I.7I 5'1 I“: I“ I”? I”. lul VIII 0 . . Base Leve1 Proaections. Actual and Simulated Results «tn ant unarroa «In 301 unarroo Pasca- nn‘ ...rr°° 339 wow-rm Kl? "IE! "flifllflfll 3‘ 30.. 3 a ‘ ’0‘...“ 0‘ L a 0... / “... I“. .',/ .. .1. \q n -" ‘-. I ‘\_ ,. 8 \ j ..I' , "J ..‘ I. ‘ suu/ I C I I I “ 0 0 4T ’j I r“ I r‘ I 4T I I I 1 I I IO?! 1'7? 3'7. ICC! III) I... III? I... I’ll 107' I077 II?‘ 1.01 III! I'll II‘? III. I‘ll run run HIIPIHI UDHIIFIMI ‘0 30 .. - — ncIunL 3' . m~smuhun C a t I ‘ I . ,f' 8 x” ' u’ ‘ If I /.“'"" ' ...-0. '0' . “...,.w' II I 0 ' ‘ - scrum. ‘“”IMIJ"I IO r. ‘T‘ ”T7 l 1 r l O ’1 T' r l ”T l I If" It" If" I." 1‘” III I." II” II" If” II"' It" I‘M II” III II” II” I." ma M can mu ll... our I”?! 2 8 LI. I I I. 0"“ % ..."..’.. U 0'. . ..." U o a ..4 F P 0 U I I 0 s. q | ‘ — “rm --mnun ... NI. 9,: ,‘ ”“‘ "qu " m OJ 0 I T I r r I I I r— l I I I I ‘.7‘ ‘.1, |.,. '..| '..3 ;|.‘ 19.7 I... IOOI 107' 1.71 307. I’ll IIOJ I... 1.07 I... :93‘ I“. 79' Figure 5.6 (continued) I"'“‘ '0‘ ounces Ilc.‘ has .q. an ..C.‘ .R‘ ....M Ina-8‘ .n‘ coccot Illlfluiilfl! 340 :3 “‘13::- ’. u a I I I j I T I I.” If” I”. “II I.” 1" I." I“. H“ "I! IU'II‘II‘KI “I "O. “U. a ’ d / I,” I. 'I C. CI. 0. < I. '”" “I ' I 1 r I T I I IN” I'"' It" I.” I.” I!’ II"' III I.” HUI M‘lIflTCIQU'HCI U. 1'. ...—"'s... 3.. u r r 7 I I I— I If” It" If” I." I.” III I‘n II. I." I“. "II“. m mun I“ ...ot"-..".u¢‘.. '1'. .m.o" ‘0. ‘ ‘“'IMMM“I l l r ”r j— T’ 4T I!" It" It” I.“ I.” III I.” I.” 5“" I”! ....R‘ ’a‘ ...‘.‘ ‘ n‘ ‘ ....I‘ ’0‘ none-u IIIDI3IUUIIKI m m . us . so . u so so It a as . a . wum‘m I. s j“ 1 ”T I T F‘ I ma urn ms no: nos not an ms m: In- mm. usual mot I. I . I . CI- .u as so a so . a “sauna ll j r Ifi j I j I me an mu m: ms nu m1 use no: u- "Immune. to u . "-°".\..-. ‘I' ...\ 0". _/‘ .~."".,.0 I " ‘ —nvu~. wuwm a T I I I I I I ms m1 ms nu ma use an use no: rm mm. nun cal-uni: u fl.. "lflfll '““IHIJ"-l 2' r I r I I I I II" It" I.” IN“ IND III I." I'D "II I“. Figure 5.6 (continued) 341 20 Nevertheless, the projected values over over the historical period. 1983-91 are endogenously determined by the dynamic interaction among all equations in the model. The simulated series of farm product prices tracked reasonably well over the historical period, although there were some problems with underestimation of fed beef price in 1976 and overestimation in the last two years of this period for fed and non-fed beef prices. Most of the overestimation problems have been carried over from a similar overesti- mation in retail beef prices, table beef and hamburger beef. The likely cause of these problems is the estimated relationships for farm product quantities, namely, fed and non-fed beef production, which as evidenced in fed beef production, is also overestimated in the late years of the historical period.21 Otherwise the tracking and direction of simulated values of retail prices are reasonable. As noted earlier, results of non-fed beef imports presented in simulations represent the estimated adjusted base quota level of beef imports under the countercyclical formula. The assumptions underlying the use of this level to represent U.S. beef imports are (a) that the elasticity of the supply of beef exports is completely elastic, and (b) that the level of U.S. beef imports will continue to follow historical levels of actual imports which have fallen between adjusted base quota 20In large models of this type it has sometimes been necessary to exogenize equations during the historical period in order to achieve satisfactory model convergence. Convergence is achieved when the change between iterations is less than .1 percent. 21This would indicate a need fOr some re-estimation and respecifi- cation of these weaker parts of the original model. 342 and the 10 percent higher trigger level (Table 5.1). Hence, model esti- mates of non-fed beef imports are more representative of long-term import levels. An analysis of the sensitivity of this assumption is presented later. The estimated retail prices determine to a large degree the level of farm prices and therefore a similarity in the behavior of these two price series emerges. Simulated retail prices track quite closely actual values with exceptions in some years. As with farm prices, there is an overestimation in the last two years of the historical series. Simulations of retail consumption per person represent a close approxi- mation to actual observations, especially in the case of hamburger beef, pork and broiler consumption. For the projection period to 1991, the results indicate a continued decline in beef cow numbers to 198“ and thereafter an increasing trend in numbers peaking at 50 million in 1989. After this year the down phase of the cycle is evident. Because of the derivation of fed-beef production specified in the MSUAM, fed beef production fOIlows a similar cyclical path. Non-fed beef production is also determined by cow numbers, although it more directly responds to changes in the prices of corn, feeder calves and hay. Growth in fed beef production and the increased supplies of beef available for consumption depress retail beef prices and cause a subse- quent decline in the farm price of fed beef. With this decline in fed beef prices the ratio of the slaughter of fed steers and heifers and all steers and heifers slaughtered declines until 1988 and thereafter increases. Projected beef imports follow a cyclical pattern at levels 343 generally below the historical period. This is because beef cow numbers and overall beef production are rising for much of the projection period. Therefore, according to the countercyclical formula for beef imports and the rising phase in the U.S. beef cattle cycle, the adjusted base quota is relatively low until 1990. If the recent past levels of actual exports of beef to the U.S. are continued, then this suggests that for most of the 1980s, either voluntary restraint agreements (VRA) are likely to be imposed on countries exporting beef to the U.S. or more restrictively, the quota levels will be applied. 0n the other hand, if available supplies of beef from exporting countries are low, as predicted by Harris and Dewbre (1983), then U.S. quotas levels may not be prohibitive. Beef import quotas turn upwards after 1988 as U.S. beef cow production and total beef production turn downwards. Projected retail prices of table beef reveal considerable fluctua- tion over the 1980s, reaching 95 cents (1967 dollars) per pound in 1991. This compares with 84 cents per pound in 1982. Retail prices of pork and broilers are projected to rise above historical levels to the seem- ingly overestimated level for pork of 99 cents per pound and an cents per pound for broilers in 1991. .These rising real levels of retail broiler prices represents a significant turn around in the historical decline in real broiler prices.- This would indicate a declining com- parative advantage of broilers over beef at the retail level. The loss in advantage is illustrated by the levelling of retail broiler consump- tion per person and the steady upward trend in table and hamburger beef consumption. Retail pork consumption, by contrast, is projected to oscillate around the 1982 level. Overall, the interactive response 344 among endogenous variables seems plausible and consistent with a priori expectations. 5.7 Sensitivity Analyses A further test of the coherence (logic) of an economic simulation model may be made by conducting sensitivity analyses on certain key variables. Two analyses of particular interest are (a) the effect of a short livestock feed supply caused by a one—period reduction of 20 per- cent in feed grain and soybean yields, and (b) the impact of a one- period increase in non-fed beef imports22 of 700 million pounds and a permanent increase in imports of 400 million pounds. These simulations, examined below, are compared with the results of the base level simula- tion. 5.7.1 impact of a Short Livestock Feed Supply In terms of the livestock feed markets in 1983, the base level simulation reflects the planted acreage and yield expected to arise under the newly instigated PIK program. Strong participation in the program reduced planted acres and increased average yields, since pro- ducers nominated, under the program, their lowest yielding land. A reduction of total grain supplies is expected to follow. Hence, the program will strengthen grain prices, for example, from .882 dollars to 1.008 dollars per bushel for corn between 1982 and 1983. Soybean prices will indirectly benefit from this strengthening of grain prices. ' How- 22All beef imports are assumed to be of non-fed beef. 34S ever, the PIK program is anticipated to maintain prices and reduce some- what the grain surpluses rather than to cause a tight grain market and high grain prices. Besides farmer participation in the scheme, weather conditions remain an important factor in determining the final level of prices and the condition of the feed market. Livestock producers have benefitted from the surplus condition of the feedgrains and soybean markets. Because of low feed costs, lives- tock producers have been shielded from the full ramifications of low real prices and a soft demand for their livestock products. To some extent the PIX program has reduced this shield. Nonetheless, grain markets still could not be characterized as tight while U.S. stocks of wheat and feedgrains remain at relatively high levels, historically (MSU 1983. pp. 147-148). Certainly a tight market would result if the effects of PIK were compounded by drought conditions in major feedgrain and soybean producing areas in the U.S. or by major U.S. grain trades internationally. To simulate such a tight livestock feed market, U.S. yields of soy- beans, sorghum and corn were reduced to around the levels of 1980, a year of excess demand for livestock feeds.23 Base model yields in 1983 were reduced from 32.5 bushels per acre to 27.0 bushels per acre for soybeans, 60.3 to 47.0 bushels per acre for sorghum and from 118.0 to 94.9 bushels per acre for corn. This represents about a 20 percent reduction in overall yields. The effect of these yield reductions when simulated in the model was to increase corn prices from 1.01 to 1.14 23These three livestock feeds comprise the large bulk of livestock feeds. 346 dollars a bushel (13 percent), and to increase soymeal prices from 3.86 to 6.03 dollars per bushel (56 percent) in 1983. Wheat prices also rose, due to indirect effects, from 1.36 to 1.39 dollars per bushel (2.2 percent). The 1983 drought in the major grain and soybean producting areas of the U.S. was fbrecast by observers at the time of writing to reduce yields by almost double the yield reduction simulated in this analysis. Nevertheless, the nature of the impact of this drought should be clear from the scenario presented here. Table 5.11 and Figure 5.7 present the short and long-run effects of these yield reductions on key farm and retail level variables. Because of the lagged specifications in many of the estimated equations often no immediate impact is felt. In this discussion, the immediate impact refers to 1983, the year the change was made; the short-run impact relates to 1984 and the long:run impact occurs in 1991.2” The short-run impact at the farm level of an overall 20 percent reduction in soybean and feedgrain yields is to increase the price of soybean and feedgrains in both the short and longer terms. This reduces the short-run production of fed beefzs, pork and broilers, each of which relies on livestock feeds. Non-fed beef production, by contrast, being predominantly pasture and range fed, does not depend on grain-based livestock feeds. Hence, the higher feed costs cause a shift of beef production from feedlot produced beef towards range fed beef. 24 This is not a total multiplier although one could easily be cal- culated. 25 Because fed beef production is specified to change primarily in response to beef and dairy cattle numbers short-term responses are reduced. 347 .mCOu coasnua ana.~ one: mmad :a moon uo uguomfi« dusuu< .uuosc mama vuumsnvm ecu no “use“ “mauum ecu mu maze v .uzmqua aawa cu ma sous) .xuoq uauuxu muzwaoa mmmoumu :« can manage“ «can use ecuuusvoun Enema Ho>ua Hmsuum cm ma nwaan .mHu>uH paofim no :oqmmsumwp you uxou some magnumvoom H.s m.~o .I n.~o o.n| n.no w.~n m.oe o.om .ucu.-_.H«z muuoaxu umz .m.= Ho. ma.~ w. m~.~ oo. no.“ ma. do.~ Na. nunmsnxw sauna upon one. -a.~ aoo. man.“ «No.1 cam.d c nnc.a eon.H .mnn .Hwn muuoaan use: .m.= n.1 m.- H. «.mH H.a m.e~ o ~.na ~.m .uau\w muofiuoun a. n.cn o. ~.on n.~ ~.m~ o m.c~ ~.mH .usoxw xuom «.1 H.e~ ~.| m.- o. o.n~ o m.m~ o.cH .uau\w moon pouncoz «.1 n.m~ «.a ~.n~ o. n.en o n.w~ ~.- .uaoxw moon pom “unmade: scaav mouwum Hamm o a.m~ o «.ma m.t c.- o q.- ~.- .mn~ .me mamauoum ~.n w.- ~.I m.n~ m.1 n.- o o.o~ c.ma .mau .an xuom N. w.“ A. w.m N. e.m c o.« <.e .mn~ .Hwn moon voulcoz c. m.- n.~ a.m~ on c.e~ o ~.~H o.mH .mAH .mwn «com com occuuuspoum lush nousmssam woman: “Ho.u new. One. man. o~o.n now. a com. Nam. cuumm tumoum owuma ~.| ~.~c o n.0e a: 0.5m c H.0n n.9n occuaaux newscasz new moon Hm>wq omen ~u>u4 ~o>04 ummn Hm>un ao>uq omen ~o>oa ~o>uq omen Ho>o4 Amsuo< mugs: moanmwnm> uncommom sou..— omcmnu camp .8: 09550 omen scum mucosa umma scum own—28 when Havana» 304 Havana» 30A umpuuur so; umpuuar so; new“ Ho-mmoH amaso>< «was mean Nwaa mama a“ uommBH uoauwmlmno "Aumxumz comm waouumv mmvamww :Hmuwpomm a“ ommmuoma ucouumm ON m mo uomumm HH.m manme 1348 .uzmfio: nuance a .oHo>oN vuoqa uo scumosumqv wen axon moms ~.n o.~n N.a N.~n N.a| N.om o N.me N.mq .ung muuaaoun n. e.Nn H.~n n.5n c.Nt n.5n o N.en ¢.wm .nnd xuom N.H o.om ~.N N.en N.” N.oN o n.NN e.nN .maa moon nowusnaez o m.Nm c.~| o.Ne o.n c.0q o «.me a.~n .mp4 anon oases acomuom yum couuolsmnou c.n. o.nq N. m.~n a.~ o.cn o e.gn w.cN .nuxo unmade»: N. N.aa N.H w.n¢ ~.e a.as o A.an N.oc .ndxo xuom o.~1 N.nm ¢.u n.nn N. N.oN o c.wo o.~n .a~\o moon nomusaam: o.Nc n.am N.HI n.cm m.~ N.N~H o m.eo ¢.nm .naxo moon smash Aouo~fioo essay magnum nuance Nessa omen Huang ~o>oa omen Nosed Nessa omen ~u>ua Ho>oa omen Hose; Hssuu< mugs: moanuqum> oucoauoz scum 09:20 omen loam 09:30 omen loam owes—8 comm loam 09.25 comm smegma» so; "mvnuwr 30A umvdour so; smegma» sea Haaa Nmnmmau uwmuu>< «may mama Naoa Aeaaenueoov Ha.m magma 349 .mmma ea nomdaH eoauom1meo."mvamaw cameo comm ea coauoswmm unmouom am we humane . N.m unawam 000— 000. 000- 000. 000. .00. 000- 000- 0000 0 n b n 0 p - 0.0 . 0.0 0000..- 00~0dx=0 ...o. 0000: 00¢..III. u 0.0 r . a a a 0 8 0.. 0 0 0.. 0 .... u 0.0 0 8 ..0 0 ... a u 0.0 a t 0.0 u 0.0 0000080 0000 .0.: 0000 0000 «00— 0J0- 0u0- 0n0— 0J0. ufl0a 000- 0000 000. .00— 000- 000— 000— 0000 .000 000. 0000 0000 . . 0.0 . . . . . . . 0000: 0000.50.08... .. 0.0 0005: 00 0050—0 ...: 00008 0000' 4000: 0060 II I u . 0 0 0 0 n I 0.0 a 0 U N . . 1............ u v 0.0 0 0 . ... u a I u 0.. 0 0 u 0.. u 0 00 _ 00 8000 0000.58 300 0000 00 C—JJ-Oz. 00. 80089 350 Avuaafiucouv 0000 .... .... a... a... a... .... .p.. a... .p.. P P p P P p P 0— —000.— 00000050 ...: 40000 0000 ll fig .8 .00 r 00 00 00—00 004—000 0.0000 0000 —00— 000— 00— 000— 000— —00— 000— 000— 000— ? P L p p P 0 0 fi 0— 82.: 33422:... 68.. «2. l r 00—00 000;! “Nth. ‘U‘ $6339 OHIO-O ‘00. 0.9386 n.m ouswfim 000» —00— 03— 000— 000— 000— —00— 000— 000— 000— P P P P P P 0 ON 085: 00000—50 ...... 40000 0000 ..II I a. 00—00 0000 ...—0000 0000 —00— 000— 000— 000— 000— —00— 000— 000— 000— b P 0 P P P P 0005: 00004330 ...: 4000—— 0000 II t 00— 00 0000 04000 00 00 00 00 00 00 00 00 00— 0— — 00— 00— 0o— 00— 00300 000 00080 UUIP' 0.00. 00030 351 Associated with this shift is a decline in the proportion of fed steers and heifers slaughtered. Beef cow numbers decline in response to two factors. First, since fed beef production is the driving factor in pro- fitability in the U.S. beef cattle industry, beef cows are slaughtered and total numbers decline. Second, with the shift towards non-fed beef production, which increases in the short and longer runs, more cows are slaughtered. This increased beef cow slaughter brings about a reduc- tion, albeit small, in the short-run level of U.S. beef imports. Since the reduction in feed grains and soybean yields is a one-period impact, this reduction occurs only in the short run. An expected result of reduced soybean and feedgrain yields is a reduction in U.S. net exports of feed grains. The immediate impact is a 4.6 percent reduction in feedgrain exports. In the. next period the reduction rises to 7.9 percent and then returns over the longer term to the base model levels. The effect of these yield changes on retail consumption and prices is also shown in Table 5.11. The 20 percent decline in yields in the first period causes no immediate impact since farm production, and hence retail consumption, is predetermined with respect to retail prices. The short-run impact in the second period, however, is a 1.3 percent fall in retail consumption of table beef per perSon.I Pork and broiler consump- tion also fall, by ".2 percent and 2.4 percent, respectively. By con- trast, hamburger beef consumption rises by5.8 percent in 1984. The impact on table beef consumption is greater in subsequent periods, ris- ing by 2-9 percent over 1985-91. 352 Retail prices rise fbr each product. Table beef prices increase by 1.2 percent in 1984, compared to less than one percent for hamburger beef, 5-7 percent for pork and 5.6 percent for broilers. Interestingly, the increased supply availability of hamburger beef per person might at first be thought to decrease its retail price. However, the reduced supplies of competing livestock products result in a net rise in ham- burger prices. At the farm level, estimates in Table 5.11 indicate that a 20 percent reduction in soybean, sorghum and corn yields results in a 1.7 percent increase in fed beef prices, and a 1.6 percent increase in non-fed beef prices. Pork and broiler prices rise by 10.5 percent and 7.4 percent respectively in the short run. the that, as expected, the price responsiveness of fed beef is greater than non-fed beef, and the responsiveness of farm prices is greater than retail prices. Over the longer term, at both farm and retail levels, production adjusts to these price increases to bring about market adjusting price declines. 5.7.2 Effect of Changes in U.S. Beef Imports Several simulations were conducted to assess the effects of changes in U.S. beef imports. In the first simulation, beef imports were increased by 700 million pounds in one period, thereafter reverting back to the level of the adjusted base quota on imports. The choice of this increase of nearly 50 percent of base level imports and of almost 20 percent of base level non-fed beef production was not purely arbitrary. This scenario is illustrative of the effects of a substantial relaxation of the beef import quota in one year. It also allows for comparison with other studies using similar levels of change. Results of these 353 effects on farm and retail prices and quantities are examined relative to the base model and are presented in Table 5.12. The immediate impact of an increase in imports on the U.S. beef subsector is the lower farm prices for livestock and feedgrains. In the short run, non-fed beef production increases as female stock are slaughtered. The lower prices of fed beef, pork and broilers cause a decrease in production. After several periods non-fed beef production returns to the base level and the differences in prices become less apparent. Over the longer term, the aberrations, caused by the one-shot increase in imports, dampen and farm prices and quantities return more or less to the base model levels. Note the decrease in beef imports to levels below the base quantities following the initial impact of aug- mented imports. A significantly greater impact is evident at the retail level. Overall, however, given the nearly 50 percent increase in the 1983 U.S. adjusted beef import quota the impact is relatively small. The increase represents approximately no percent of actual beef imports in 1982. In Table 5.13, a comparison is made of the effect on farm and retail prices of this 700 million pound increase in one year with a 400 million pound increase either as a once-and-for-all increase or as a permanent increase over the whole projections period, 1983-91. An immediate impact of the 700 million pound increase in beef imports is to decrease table beef prices by 4.6 percent and fed beef prices by 7.3 percent. For the same increase in the level of imports, Freebairn and Raussar (1975, p. 687) obtained smaller decreases of 3.0 percent and 5.5 percent in respective prices. This seems consistent 354 .NmsH nu uuueeaa use; .m.= Hmsuum no usuuwoa 0: was Hu>uH duosu sump vuumshvs NmaH use 00 unouuun HxnhHoumaonu00< HH.m uHess sum HH.m uHeeu sum SH.m oHeeu sum .1: 0'0 6 mchumkum o 0.H0 H. 0.H0 H. m.no H... n.00 0.00 .25 .a .HHZ 3.3me “.02 .05 o m¢.H o 0N.H Ho.t no.H Ho.l Ho.H N0. Hosmsa\w uowum cuou NHo.I NNo.H soc. 000.H «No.1 owm.H co“. nnc.H ween.H .mAH .HHn muuoaaH moon .m.= N.» 0.NN o N.0H N.I 0.cH. 0.: H.MH N.0 .uzuxm muuHHoum o n.om o N.0m N. N.nN ¢.| 0.0N N.mH .usuxm anon H.I H.0N . c n.NN N. c.mN H.HI n.0N o.cH .usu\w 0000 vuwlcoz N.I n.0N H. H.0N H. n.0n H.NI 0.0N N.NN .usoxm upon vow AeueHHoo neaHv muuaum sham o o.nH o ¢.nH on o.NH o c.NH H.NH .mnH .HHm muuH—oum o 0.NN N.H| n.nN H.| m.HN o o.oN c.0H .mAH .HHn xuom H. N.N o 0.0 n. e.m c 0.: «.c .mnH .HHn moon vomlcoz H. n.HN H.| 0.0H as c.0H o N.0H c.0H .mnH .HHm moon v00 uaoHuusvoum Show nounmsmHm wowHu: woo. new. «ca. 000. oHo. man. 0 cow. N00. OHumx vacuum ouumm H. H.0e c m.0¢ N.I 0.00 o H.0m 0.00 mGOHHHw: pmuoaasz sou moon Hu>ua omen Hu>ma Hu>o4 omen Ho>oa Hu>oq comm Ho>oa Ho>oq ommm Ho>oq Hcsuo< muHca mmHanum> umcoammm scum «mango umsm scum omcmso ommm scum omcmco omen Edam uwcmsu «mum "manomeH "muuoaeH “muuoasH ”muuoneH cummmuucH vommouucH vommouocH vommouucH HO0H HmumooH owwuo>< emoH nmaH NmmH mmaH :H uumaeH onumm mac "manomsH wmmm NH.m anmH .m.: :0 wmmmuucw mvasom :OHHHHE 000 m 00 uuwmmm 355 £32.. 0033. a 0: 0.00 0 0.00 0: «.00 0 0.00 0.00 .m00 80000.00 0.: 0.00 H. 0.00 ~.: 0.00 0 0.00 0.00 .000 spam 0.: 0.00 0.: ~.00 0.: 0.00 0.0 0.~N 0.00 .000 amon uwwuanewz 0. 0.00 0. 0.00 0. 0.00 0 0.00 0.00 .000 umo0 000m0 . a cannon yam c00unesmcoo 0.: 0.00 0 0.00 0.: 0.00 0.: 0.00 0.00 .00\9 muo000u0 0.: 0.00 0 0.00 0.: 0.00 0.: 0.00 0.00 .00\0 xuom 0.: 0.00 N. 0.00 0.: 0.00 0.~: 0.00 0.00 .Aon uwmn nowusaemz 0.: 0.00 0. 0.00 0.: 0.000 0.0: 0.00 0.00 .00\0 0000 oHAmh oua0000 00000 nou0ua 00auwm 0o>o0 «mam 0o>o0 00>00 omm0 Hm>u0 00>~0 uawn 00>00 Hu>o0 oman 0m>o0 Haauu< ou0:= ao00a0uu> umcoaaom scum owcuso 00:0 loam ownono 0000 £000 owcuzo oou0 noun o0cmsu mama 0000030 0000030 “nquQEH "au00060 vomaouuc0 vmoumuuc0 conuouuu0 vuomouucH 0000 00:0000 omauu>< 0000 0000 N000 00m==0ucouV ~0.m o0nme 356 Table 5.13 Effect of a One-Period and a Permanent Increase in U.S. Beef Imports One-Period Increase Permanent Increase of in U.S. Beef 400 Million Pounds Imports (mill. lbs.)__ in U.S. Beef Imports Ease 700 400 Long-Term a Level Immediate Maximum Response Response Variables l983 Impact 1983 Response l99l ---Change from Base Levels ------------ Retail Prices (¢/lb.) b Table beef 94.5 -4.3 -2.5 2.7 $1 -l.0 Hamburger beef 68.0 -2.6 -l.5 2.0 l - .9 Pork 7l.l - .7 - .4 - .6 (8 - .6 Broilers 3l.6 - .9 - .5 — .6 (l) - .4 Farm Prices ($/cwt.) Fed beef 28.7 -2.l -l.2 -l.2 (0) - .6 Non-fed beef 23.3 -l.l - .7 - .7 (0) - .4 Pork 24.9 - .4 - .2 - .3 :8; - .3 Broilers l3.l - .6 - .3 - .4 O - .3 Corn Prices (Slbu.) l.0l - .0l 0 - .02 (4) - .Ol u. 5. Beef Imports (Tmi :453 ° 700 400 4:7 (4) 354c bs U. 5. Net Exports feed 60.3 - .l - .l - .5 (5) - .3 Grains (mill. tons) aReal prices are in l967 dollars. bThe number in parenthesis is the delay of the maximum response in years. cMinimum response to 400 million pounds permanent increase. 357 with this study's results since in their study, the 700 million pound increase was only 35 percent of actual imports. However, a significant departure of the results of this study from their results is in the effect on non-fed beef prices. In the present study the estimated decrease in hamburger beef prices is 3.8 percent and for non-fed beef prices the decrease is 4.7 percent. Not only are these responses sub- stantially lower than their estimates of 5.2 percent and 13.1 percent, respectively, but the impact on non—fed prices is less than on fed beef prices. An economic explanation of this result is that beef imports are more closely substitutable with fed beef than they are with non-fed beef. However, neither the available evidence nor the model specifica- tion support this conclusion. In subsequent periods, however, the results from both a one-period impact and a permanent increase show a larger effect on hamburger beef prices than on table beef prices. This suggests a more likely explana- tion for this departure. That is, the long-run specification of fed beef production, which responds primarily to changes in cattle inven- tories, causes an apparent underestimation of fed beef production and hence an overestimation on the short-run price response. This in turn would cause an underestimation in the hamburger beef price response. The immediate effect of this one-shot increase of beef imports on retail and farm prices of pork and broilers is smaller than fcr beef products. There is a high degree of compatibility between the percen- tage changes of these estimates and those obtained by Freebairn and Raussar (1975, p. 687). The effect at the retail level was a decrease in pork prices of less than one percent, and a 2.8 percent decrease in 358 broiler prices. At the farm level, the immediate impact is larger; pork prices fell by 1.6 percent and broiler prices fell by 4.6 percent. This result reflects a greater degree of substitution between hamburger/non- fed beef and broilers than between hamburger/non-fed beef and pork. Also presented in Table 5.13 are the results of a one-period increase of 400 million pounds in U.S. non-fed beef imports. As expected the immediate impact is a little over half that described in the previous analysis. However, a 400 million pound or 28 percent increase in the 1983 quota level is a more realistic level to examine in terms of a permanent increase in beef imports.26 As expected, the major impact is in the first or second periods. Also as expected, a permanent increase in beef imports gives a decrease in farm and retail prices in the long run. Nevertheless, changes in non-fed beef imports have a small long-run effect on retail and farm prices of meat products. As in other studies, this is due to the long-run response of domestic beef supply to price (Arzac and Wilkinson, 1979. p. 305). 5.8 Economic and Policy Simulation Analysis In this section, results of specific economic and policy simula- tions are presented. These simulations illustrate the fallowing direct impacts on retail demands for meats: (a) a sustained increase in demand for individual meats; (b) a sustained increase in real disposable incomes; (c) an analysis of time-varying parameter changes in demand; (d) a proposed reduction of the dairy cow herd to bring in to balance 26See Appendix Table A3 for the results of a 10 percent increase in each year. 359 demand and supply in the dairy market; and (e) changing economic growth in major grain importing regions in the world. 5.8.1 Sustained Increase in Retail Meat Demand A sustained increase of 10 percent in demand for each retail meat was simulated (Table 5.14). This increase represents a continuous shifting out of the demand curve over the projection period and reflects various scenarios of these retail products. In the case of table beef such a continuous increase in demand may be due to one or a combination of potential factors. These include a favorable shift in tastes and preferences towards table cuts of beef; the steady shift of the 'baby boom' group into older age cohorts, traditionally characterized as heavy consumers of table beef (steaks, roasts); or a successful advertising campaign aimed at increasing demand fcr table beef. The sustained growth in demand for hamburger beef may also reflect a steady shift in tastes and preferences toward leaner meat or the effect of meat grade changes and changes in product specifications which more accurately reflect, at reasonable prices, the qualities of leaness and taste desired by consumers. Similar factors may also explain such shifts in pork and broilers. In the case of broilers, however, the individual growth in demand may follow from a continued preference for white meat and the alleged health factors associated with its consump- tion. The initial impact of an increase in the demand fcr each product is an increase in farm and retail prices, since quantities are predeter- mined in the current period. The associated immediate rise in prices of 36“) 33.5 530.33 05 £53.30 5 532.3 3033...... :5 5c .3339. 33:00: 553:0 .80— 3 38?. 3:83.. :2..0:5 2: 0 .33» 5 3:83.. 55:2. 0:» .3 3.3 2.. 5 535:3»: 5 .305: 2:6 .3335 .3330 we 5 $00— :5 0:98.. 5 33.65 0:33.. 00 23 33 0000— 5. 335 35085 :5 05 350.. .a....oa Noa— e. sou..a ..aN. ...- 0N0 3.0- o c. - .NV N. - p. - _. - AN. ..N- N..- ..N. .000 ..m- o.m .nap ....a “sagas: coon .m.= 33 3.20030 _. - 0N. N. - a N. - 0N0 m... N. - 0. Any _. o p..- 000 N..- a .: ..Fa: ascoaxm so: .m.= N. 0N0 o. .. P.N an. p.n o.N N. - 0.0 o. - P. - ..n Am. o.. ..F au.ca ecoo N.0- 0.0 N... N. - m. - Am. 3. - N. - c. - 0.0 a. - N. - N.N AN. N.m 0.. .co....: nausea: sou moo: .H:m a cue. Nun. N. - _ N.N- N.Nn N. - N N..- 9.0- m. - m N.o.. ..n 3.00.02n N _ a. - a. - nus N o.N o.m N. - . a. - s. - N._ N N. - ..N saga N.:. o N.N N. - _. - m a”. a. - o.m m ..o. N.» ..N- m o.NN- 3.3 cuss ao.-=oz _ o N s N _- s - m a _ o. - c. - _ a..- a..- ¢.N m a.NN N.m_ ages so: .23 33:: E... ”UN- 3 Nua- N”: m” - N m..- m..- s. - N N. - a. - N.N- Mm. m.N - a.N a.o_.o.N ~.— w 0. 0. - 0.0 0 0% 0.0 m. a 0 m. n m. .. 0.0 3 N. - 0.0 0.8.. _._- N M.N- .. - N. o o.p N. - N.N a m.:. o.m N.o- Mow o.._- N.N moon coacsae.: _ N p- a - m N m. - N. - p 0..- ...- N. c ..m. ..00 cuss «so.» .a,\. "mou_aa ...soz 3 o: o: 026— 32 3.; 00:29 33:33.. : a 3: 3.. 38.: 3:83 3:330 33.: 3:93 3:33 om - z 3.00 3: a 3:903 33:. 3 :0 325...; 3.803: .:::::.. mmmmw: am_.oama. 0:00. -Ncamm -oaa0 -Ncogm -meos -scomm .1mmmmuwmruwriii .NmmmwpiNmmnINMmmpmmm::::.:::::::m:mmb:Nmmn:w_:~N: 0235505 00230 330.35 0:950 "33: .3» 0:930 :33. 05 5 0:33.. 00 .3 38.55 35330 a ha 333 00.0 03:... 361 competing products following the increase in table beef demand is caused by a series of effects at the farm level. The rise in fed beef prices increases the corn price and feeder calf price which subsequently dam- pens fed beef slaughter and supplies available for consumption. with lower quantities available, retail prices of competing products rise temporarily. The following short-run response is a decline in competing product prices. Beef imports rise in the initial period following an increase in demand for table beef. This rise is associated with an inventory build-up in response to higher fed beef prices. In later years the sustained increase in table beef demand appears to have a negative influence on the level of beef imports. It is the rising beef cow inventory and the associated rise in the supply of cow beef which, according to the countercyclical formula, reduces the level of non-fed beef imports. Also, the increasing quantities of fed steer and heifer beef increases the supply of processing quality beef obtained from the lower quality cuts from these carcasses (Martin 1983. p. 6). Note also that the corn price increases and U.S. net exports of feedgrains decrease when the level of demand increases for high grain consuming livestock products. As expected, the opposite effect occurs when hamburger beef demand increases. 5.8.2 Sustained Increase in Real Disposable Income Per Person Real income growth projections contained in this model reflect the recent turnaround and expected growth in the U.S. economy. However, these base model rates of growth in real disposable income per person were increased by two percent in each year. As expected, this increase 362 has a substantial positive effect on retail and farm prices for meats, the percentage increase more or less reflecting the income responsive- ness among meat categories. The impact was greater in later periods. U.S. corn prices responded in an upward direction while U.S. net exports of feedgrains declined in each period. With rising beef demand beef cow numbers increase as producers hold breeding stock from slaughter in expectation of greater returns in the future. As a result, beef production, especially non-fed beef production decline in the short-run. This reduction in non-fed beef production, and indeed total beef production impacts directly on the countercyclical formula for U.S. beef imports and augments the level of beef imports permitted under the quota. Over the long term, as producers begin to release cattle for slaughter, both fed and non-fed beef production increases. The latter increases both from the non-fed steer portion of the herd and from cull cows and bulls. In turn, cow beef and total beef production increase and hence imports are reduced. 5.8.3 Structural Changes in Demand Parameters In the previous Chapter u, estimates derived from time-varying parameter models indicated considerable variation in the flexibilities of retail demand for meat. Following from this result, three broad developments may be identified fer fUrther investigation. First, over the past several years own-quantity flexibilities of all beef, and in particular, table beef, were shown to rise. This rise indicates a weakening of demand for table beef. That is, relatively large price adjustments in the market price are required to clear the 363 market in response to small changes in quantities supplied. This recent trend has been explained in other studies by the negative quality changes which have led to a substitution of poultry meat for red meats. Second, among red meats, hamburger beef, because of its attributes of leaness and convenience in preparation, has been noted both in earlier discussion and in other studies, to have achieved greater preference among many consumers. Hence, hamburger beef appears to have become a stronger substitute for table beef. Third, income flexibilities for both table beef and hamburger beef, although rising steadily since 1979, have shown a relative decline since 1977. This down trend may suggest a decrease in the long-term growth rate of the U.S. beef industry. These three scenarios, of certainly many possible ones, are presented below. The choice of these simulations was based primarily on the desire to investigate the effect on the U.S. beef subsector of a continuation of recent trends and development in the structure of retail meat demand. (a) Continuous Increase in Own-Quantity_Slope Coefficients The prime objective of increasing the slope of the table beef demand relationship, i.e., making it steeper, was to investigate the effect of increasing the own-quantity flexibility for table beef on the beef subsector (Table 5.15 and Figure 5.8). This represents a continua- tion of recent trends. The coefficient for table beef quantity was increased linearly by one percent such that the original coefficient, , becomes .01 T for time, T = 1, 2,...,9, over the projection period 1983 to 1991. .3654 ..N.N uNaNN Nome .NN.N aNaaN ammo ....NNou NNNN :N «ouNca NaoNa .onaoco:. N¢0¢00 acouco: 0:0: .NNN N. o a N. N. o N.N- N. - c .NNNs «Noose. cue: .N.= meow choc0voom N. a a o a o N. a o .a .NNNs NNaoaxN Nu: .N.= N. - N. - N. - N. o : N.N- N. - o .NNNN ouN.N atom .N.= N.N a o N. o a N..- N. - o neoNNNNa Nassau: zoo NooN a oN- N.N- ..N- N. a a N.N N. - a acaNNoLN N. - N” o N. a o s. - N. - a Naoa N.N- N. o N.N- N. - N. - N.N N. - N. - Noon soc-co: a N N o N. - N. : N.N- N.N- ..N- mean so. .0..va "mOUvLQ Ea..— a 0N- N.N- N.N- N. o o N.N N. - o NaoNNoLN N. - N. o a o o N. - N. - a Nana N.N- N. 0 0.0- 0. - 0. - 0.0 m. - 0 woos guacaneN: N N 0. o N. - o o N.N- N... N. - No»; 32:. .nNNN ”NauNaa NNNNNN N _, 0.... :uNaoNNN> ummmmwmm -o¢0::¢ -o¢ ::¢ oua0 -oz :50 -3: :30 30:00 mo“ :30 -o¢ :20 o 30:00 - :o -oea 0::0 -00000 -3250 -0:o0 -0Logmil. -9500 reason No—Noco u:aeoa yuan somuzneNz v:aeoa 0mo0 «Nae» :1:- ANNaauszoe~ uoNpaa< No0:agu 00:00 “Nassau neaeaa Nam: 00:00: :0 mucoNoNONoo0 one—0 auvu:aaa-:30 :— oonaoco:_ maca:Nu:ou a 0o uoo0Nu 00.0 000:» 365 ochaN Na NNNNNNNNNNN NmuN NNNNN cN mmuagucN uoNNNasNN N.N NNNNNN NNNN . ON on no .N No No Np Np .N NN ON on no No NN on on on Nm Nm om bLLp__FPFh Pb PFb _ PF __ ¢.O NNNNNNNXNJN NNNN NANNN 2N NNNNNNzN ouNcqasz.;-: NNNN NJNNN No NNNNNNNxNNN NNNNNN Juno: mace-I:- _ rmé x a N.N M f N.o N3 .N..... ..N5 '11.. I.— - N.N DHKUUD— LJUXHIDHNJHP? Oh... OUZCZD 366 The simulated results agree with a priori expectations. The con- tinuous time-varying increase in the own-quantity coefficient increase the effect of a given quantity of table beef on the table beef price, decreasing the retail price in each period. The immediate impact of this one percent structural change is a .9 percent decrease in table beef price. In the short run, i.e., the next period, the higher rela- tive prices of substitute products brings further increases in supply availability of those products, causing downward pressure on all other retail and farm level prices. The effect of this structural change is greater at the farm level. Lower fed beef prices result in reduced feeder calf and corn prices. Beef cow numbers are reduced in response to both the declining profitability of cattle production and the increased slaughter of breeding stock. As a result non-fed beef produc- - tion is increased and U.S. beef imports reduced. In the longer run with the continued run-down of the beef cow herd, non-fed beef production decreases. With fed and non-fed beef production falling below base model levels U.S. beef imports would be expected to increase. Somewhat surprisingly, however, imports continue below base levels. A likely explanation, given the components of the countercycli- cal formula in equation (5.1), is that the three-year moving average of production is falling faster than the ratio of the five-year to two-year moving averages for per person domestic cow-beef production is increas- ing. This apparent disfunction of the countercyclical formula was reported in another analysis (Simpson, 1982). In this analysis an oppo- site case was observed, where the trigger level, and hence the adjusted base quota level continued to increase even though production was 367 increasing. This being so, the beef import formula appears to be inef- fectual in achieving its purported countercyclical function. (b) Continuous Decrease in Income Slope Coefficients Using the same method, the coefficient of disposable income per person was allowed to decrease linearly over the projection period. Separate experiments were conducted for table beef and hamburger beef with the decrease being one percent in each case (Table 5.16). Since the result of this structural change is to increase prices for the given changes in income the overall effect is to cause decreasing income flex- ibilities over time. This declining income responsiveness of consumers operating at the market level implies a declining long-term growth in demand for the particular product in question. In the case of table beef, for example, the immediate impact of. a continuous decline in income responsiveness of one percent annually, is to reduce retail prices by 1-2 percent and farm prices by 1.7 percent. This percentage decline increases over time. As in the previous experi- ment of coefficient change, the decline in profitability of fed beef production leads to a steady reduction in beef cow numbers, an increase in non-fed beef production and to an associated decline in non-fed beef imports. (c) Continuous Increase in Substitution Effects Evidence presented here and in other studies indicate an adverse shift of consumer preferences and tastes away from table beef. Factors cited have included alleged, undesirable health attributes of table Table 5.l6 Effect of a Continuous Decreasea in Income Slope Coefficients in Retail Meat Demand Curves: Slope Changes Applied Individually Hamburger Table Beef Demand Beef Demand Imme-c Short- Long- Imme- Short- [ong- Responseb diate Run Re- Run Re- diate Run Re- Run Re- Variable Units Impact sponsedsponsee Impact sponse sponsg__, percentage change from base level Retail Prices: ¢/lb. Table beef -l.2 -2.7 -4.4 0 .2 .l Hamburger beef 0 - .5 6.0 -l.9 -3.3 -l3.0 Pork 0 - .4 - .6 0 .l .2 Broilers O - .6 -2.3 0 .l .2 Farm Prices: ¢/lb. Fed beef -l.7 -3.3 -6.4 0 .3 -0 Non4fed beef -l.4 - .8 4.2 -2.0 -3.0 -l3.9 Pork 0 - .4 -l.l 0 .l .2 Broilers O -l.4 3.5 0 .2 .3 Beef cow numbers millions 0 - .3 -7.2 0 0 l.0 U.S. corn prices $/bu. 0 -l 0 -3.4 0 ' O 6 U.S. net exports mill. tons 0 - .3 1.0 0 0 - 1 feed grains U.S. beef imports mill. lbs. 0 — .6 -3.l .3 - .3 aOne percent linear increase. bReal prices in l967 dollars. c1933. ' d1934. 81991. 369 beef, a general preference for lean meats, the effect of official government reports citing the desirability of eating white meats or fish, or the convenience and strong promotion of certain fast foods such as hamburger beef and chicken pieces. Following the period of high beef prices, especially of table beef, during the middle 1970s, changes in consumption habits of traditional and heavy consumers of table beef, also contributed to the substitution of the other meats, and indeed non-meats, for table beef (see Chapter 2). The simulations presented in Table 5.17 reflect some of the quanti- tative effects of these types of structural changes occurring in the retail meat market over the past decade. In the first experiment, the coefficient of substitute product, hamburger beef, is increased only in other red meat demand relationships (table beef and pork) by one percent linearly over the projected period. The immediate impact of this struc- tural change on table beef and pork prices is a decrease of .5 percent and .1 percent respectively. In the next period the impact is a .9 per- cent and .2 percent, respectively. Over the longer term, the impact lessens for table beef but increases for pork. Two forces account for this diminished long-term effect on table beef prices. The first is the increased competitiveness of hamburger beef associated with rising real hamburger prices. The second is the effect of lower corn prices and feeder calf prices in moderating the decline in fed steer and heifer slaughter. Farm prices reflect these changes. As expected fed beef production falls along with beef cow numbers. Similar to previous simu- lations, the slaughter of breeding stock and lower feeder calf and corn prices, augment non-fed beef production, forcing down U.S. beef imports under the countercyclical policy. 370 Table 5.17 Effect of a Continuous Increasea in the Substitution Effect of Hamburger Beef on - Other Red Meat Demand and of Broilers on all Other Meat Demand: Substitution Effects Applied Individually Increase—in Hamburger Beef‘ Ifitrease in Broiler Coefficients in Other Coefficient in all Other Responseb Red Meat Equations Red Meat Equations Variable Imme- Short- Ebng- Ifi‘ - Shhrt- ’[bng- diate Run Re- Run Re- diate Run Re- Run Re- Impactc sponsed sponsee Impact sponse sponse ------------ percentage change from base level----------- Retail Prices: . Table beef - .5 - 9 - .l -l.4 -2.6 -3.9 Hamburger beef 0 - .2 3.9 - .7 -l.8 .8 Pork 0 - .2 - .7 -l l -l.8 -S.6 Broilers O - .l 2.0 O - .2 3.4 Farm Prices: Fed beef - .7 -l.l .2 -2.2 -3.5 -5.8 Non-fed beef - - .l - .4 3.7 -l.2 -2.3 .1 Pork - .2 - .3 - .9 -l.7 -2.8 -7.l Broilers O - .2 2.9 O - .2 5.0 Beef cow numbers 0 --.7 -4.3 0 - .2 -7.0 U.S. corn price 0 - l -2.0 - .2 - .6 -5.0 U.S. net exports 0 O 7 O O 3 feed grains U.S. beef imports 0 - .2 -l.O O - .5 -3.5 a“One percent linear increase. bReal prices in 1967 dollars. c1933. d1934. e1991. 371 This case of the effect of changes in the structural relationship of hamburger quantities on other red meats may be usefully juxtaposed with the case of increasing structural competition of broiler consump- tion on retail prices of all other meats. This scenario attempts to recreate, or perhaps more precisely, extrapolate an existing development in retail meat demand in the U.S. To date retail broiler prices have exhibited a strong downward trend in association with a consistent growth in levels of broiler consumption. Evidence of growing consumer preferences and tastes underscore this trend. For various reasons, only some of which can be fully substantiated, chicken has been promoted as a more healthful form of meat protein. It has been credited with a lower fat content although its production involves the extensive use of growth additives and antibiotics in feedstuffs (see Chapter 2). Convenience in preparation and purchasing as a ready-to-eat food have also contributed to its increasing share in the consumer meat budget and in overall meat consumption. The factors appear to have strongly influenced the struc- ture of the demand for other meats considered in this study. The immediate impacts of the one percent continuous increase in the broiler-quantity coefficient on table beef, hamburger beef and pork demands are a 1.4 percent, .7 percent and 1.1 percent decline in their respective prices. The short-run impact is approximately double these percentage reductions in price. Farm level declines are greater in each case. U.S. corn price and beef imports decline both in the short and long runs. Compared to the impact of the similar structural change con- sidered for hamburger or other red meats, the impact of the rising cross-effect of chicken meat on other meats is substantially greater. This is to be expected, first, because the impact is widespread and, 372 second, because of the large cross-coefficients of broilers versus ham- burger. In each of the simulations of structural change in retail meat demand in the U.S., the underlying scenario has been similar in impact although viewed from different perspectives. The recent history of the demand for table beef has been projected in terms of the implications of a rising own—quantity flexibility, declining income flexibility, and increasing cross-quantity flexibilities. In each case the result has been lower retail and farm prices for all products in the short-run but also in the long-run fOr table beef, lower corn prices, beef cow numbers, and generally, lower U.S. beef imports. 5.8.4 Impacts of a Proposed Dairy Herd Reduction Program Legislation was tabled in 1983 in the U.S. House of Representatives to implement a dairy cow herd reduction program. The objective broadly is to achieve a balance in the U.S. dairy market and hence reduce the large and growing annual cost of the current price support program. Details of the herd reduction program are not yet firm, however, in essence it is to provide incentives to producers to undertake dairy cow culling in order to decrease numbers more in line with effective demand. The presently proposed legislation provides for this program to be undertaken within one year. There is an amendment to this proposal, supported by the National Cattlemen's Association and others, proposing to postpone the program's implementation until 1984, to avoid the sea- sonal peak in cow and hog slaughter, and to temporarily suspend the 373 incentive if total federally inspected cow slaughter exceeds certain initial levels on a weekly basis. Certainly, the achievement of dairy market balance in the immediate term, 1983, can be expected to have a greater impact on the beef cattle market than if dairy herd reduction was carried out over a number of years. In terms of goals of market stability, there is strong reason to minimize the disruption of the beef and beef cattle markets by permit- ting the increase in dairy cow slaughter to be gradually absorbed into regular red meat supplies. A more obvious area where a one-year dairy cow slaughter program would be felt is U.S. beef imports. Most, if not all, dairy cow beef would compete directly with these imports, but would also be expected to bring about a direct reduction in the level of imports through the operation of the countercyclical beef import for- mula. In this section these two broad alternatives, namely to achieve dairy cow herd reduction in one year versus over several years, are examined. This analysis is by necessity, somewhat aggregative although the general order of magnitude of the impacts of the two policy alterna- tives should be obtained. In January 1, 1983 dairy cow numbers were 11,066 thousand head. The Spring 1983 report of the MSUAM (MSU 1983, p. 61) contains a 1983 estimate of milk produced per dairy cow of 1,250 pounds, giving U.S. milk production of 138,500 million pounds. Total supply including beginning commercial stocks and imports is estimated at 147,400 million pounds in 1983. Domestic consumption (utilization) is estimated at 127,300 million pounds including 2300 million pounds consumed on farms. 374 Supply is therefore some 14.5 percent or 18,500 million pounds in excess of demand in 1983. In terms of the annual average level of milk production per cow, this excess 18,500 million pounds represents 1,480 thousand dairy cows or 13.37 percent of the 1983 dairy cows herd. If it is assumed that all dairy cows produce the average amount of milk per cow, the slaughter of this excess number of dairy cows would bring supply into balance with demand, ceteris paribus. This assumption, however, may be quite unreal- istic as it is likely that farmers would slaughter their less productive cows before others. A distribution of the national dairy cow herd by productivity per cow would enable a more realistic assumption to be made. However, neither national nor useful regional distribution data are readily available. In a somewhat arbitrary fashion, therefore, the lower bound on the number of dairy cows slaughter needed to bring supply into balance with demand was the rounded figure of 13 percent.27 The upper bound on dairy cow slaughter was even more arbitrarily chosen at 2,213 thousand cows or 20 percent of the 1983 dairy cow herd. This percentage implies an aver- age productivity of culled cows of 8,360 pounds per cow or culled cows approximately two-thirds as productive as the national average for 1983 estimated in the MSUAM. In summary, a 13 percent and 20 percent dairy cow herd reduction implies the slaughter of 1,439 thousand and 2,213 2 7This is most certainly a lower bound. Nevertheless, a more de- tailed analysis may enter into the calculations, a producer parti- cipation factor. In this study, a 100 percent participation is implicitly assumed. 375 thousand cows, respectively. This slaughter may be carried out in one year or over a three-year period, 1983-85. To arrive at dairy cow beef production the number of dairy cows slaughtered were multiplied by the weighted average dressed weight of bulls and steers in 1981 (Table 5.18).28 When dairy cows are slaughtered and dairy cow beef enters the beef market over a three-year period, pro- duction is assumed to flow into the market in three equal portions. In terms of the mechanics of simulating the impact of this additional non- fed beef coming into the U.S. beef market, quantities were exogenously added to total non-fed beef production in the appropriate years. The one-year slaughter program can be expected to have a substan- tial immediate impact on the U.S. beef subsector (Table 5.19). The three-year slaughter program, as expected, is considerably less disrup- tive in the first year. However, over the longer term a much greater affect on retail and farm prices would be felt. The simulations of a one-year slaughter program, assuming a 13-20 percent dairy cow herd reduction, indicates a 4-6 percent reduction of table beef prices in the immediate term. Hamburger prices fall by 3—5 percent, with smaller declines in retail pork and broiler prices. With this one-year shock, the demand effects of available supply reductions in the following year, 7 cause a sharp increase in product prices. When the slaughter program is .‘ operated over a three-year period, the immediate decrease in retail table beef and hamburger beef prices ranges between 1-4 percent and 1-3 281981 was the latest available data (USDA. 19323). 376 .NNNNNNV (am: eN eoeNaNeoo NNNNN .umm; ocemzosu omo.NN mo ago; zoo ngmv mmmN .N acaacon co omamm N.NNN N.NNN NeN coo. N.NoN N Nose NoNeesNNN N.NNNN N.NNN NeN coo. NNoN N eN soNeNNNNN NeoNpozuoNa Noam zoo zeNao Nope» Nmm Nmm mnN ammaum .NNNsm .mzou No unmNmz ummmmga mmmco>< wounmez mNNN mmeN mucamzocu aumgmugmsmNm Nzou NgNao No Nonsaz om mmi. muNca vogougmaaNm new: zoo xtrmp mo mmoucoocoa moNoNNoa eoNNuavwm new: zoo ngac o>NumceoNN< Nova: comm zoo aNNoo No coNNmuoNN< use coNuozuoNa mp.m mpnmh Ei77 pom—u NNNNo NNNNe .NNNNNNN NNNN e. NuoNsa NNNNN .NNN N.N- N.N- N.N- N.N- N.N- N. - N.N- N.N- N.N- N.N- N.N- N.N- .NNNn NNNNNeN NooN .N.N Neon NeNNNmuouN N.N o. - N. - m. o o N.N o.n N.N m. N.N a. .e .NNNa NNNNNxm «oz .m.: N.N- N.N- N N.N- N.N- N.N- N.N- N.N- N.N- N.N- N.N- N.N- .NNNN NooNea eeeN .N.N N.N N. - N N.N N. - N N.N N. - N N.N N. - N NeoNNNNe Necessz xeN NNNN N.N- N.N- N.N- N.N- N.N- N.N- N.N- N N.N- N.N- N N.N- N.NNNNNN NHN- N.N- N.N- N.N- N. - N. - N.N- N.N- N.N- N.N- N.N N.N- N.NN N N- N.N- N.N- N.N- N N.N- N. - N.N N.N- N. - N.N N.N- Nose NoN-eoz N.N- N.N- N.N- N.N- N. - N.N- N. - N.N N.N- N. - N.N N.N- Nose NNN .NNxs "NouNcN scum NHN- N.N- N.N- N.N- N.N- N.N- N.N- N. N.N- N.N- N N.N- NNNNNNNN N.N- N.N- N. - N.N- N. - N. - N.N- N.N- N.N- N.N- N. - N. - sees —.m- o.m- m.- N.N- n. - N.N- m.~u m.~ —.m- n.—- m.— «.m- moon Nougansox N N- N.N- N.N- N.N- N. - N.N- N.N- N.N N.N- N.- N.N N.N- NNNN NNNNN .NNNN "NooNeN NNNNNN n -u u - -Noso— anon easy oncogu amaueoucua----------u-ucra- t--- . onemmm omeoqN Noume— oNcoqN oncomn NuaaaN omcoqN «Neon» NunasN uuNeomN uoNNoaw|.NonaN NNch ooNnaNNN> oNcoaNou -oz cam .0“ egg oNNNv -uz Nam -ua cam NNNNu -oz Nag -ox cam oNNNN -aa Nae -om can NNNNu -mcoa. autosm -aaIN -NNoN -Nsosm -92rN mcmul. NNosm -aea~ ago; -NNo;m -oeaw nmnil, .N NNN .NN NeocNNam NoNNN: NuNmo-c>mNsu¢ o» NagougaamNm-mmozlsop awrmnluo umoucaoson menu» oats» New» oco Nee): umNsnuuo Nowgmambm.3eu agree segues; coNuoavox ago: you ANNNo .m.= evacuate No uuNNeN NNNNNNNNN m—.m opaoh 378 percent, respectively. However, in the long-run, prices fall substan- tially more than long-run declines with the one-year program. With the overall profitability of beef production reduced by the inflow of dairy cow beef onto the meat markets, beef cow numbers fall. The short-run impact is greater in the one-year program although, as expected, the long-run response is greater in the three-year program. Corn prices seem to fall more in the one-year program, both in the short and longer runs. Perhaps the most consistent and substantial impact of this dairy cow herd reduction program falls on U.S. beef imports. Again the trade-off is between greater short-run declines in imports with the one-year slaughter program versus greater long-run declines with the three-year program. However, the disruption and market instability caused by the one-year program is likely to impose a greater cost on the subsector than when dairy cow slaughter is spread over a longer time period. Certainly in the latter case greater opportunity exist for market planning to cope with adverse long-run effects. 5.8.5 Effect of Changing Income Growth in World Markets In recent years, the developing country markets, the Soviet Bloc and mainland China have accounted for almost all of net imports of world wheat, almost 60 percent of world coarse grains and up to 40 percent of soybean net imports. To this group of world markets, the U.S. provides around 50 percent of wheat, 75 percent of coarse grains and more than 60 percent of soybean net imports. In the short—run net imports are highly variable depending on production. In the long-run the income, 379 population and productivity growth are the major determinants. When these regions are experiencing high economic growth their import demand for wheat and coarse grains rises. Depending on the level of U.S. livestock and livestock product prices, this import demand competes with the U.S. livestock subsectors for feedgrains. It may be expected, there- fore, that changes in the rate of income growth of these grain—importing regions will influence economic conditions in the U.S. beef subsector. Simulations of high and low income growth scenarios fbr developing markets, the Soviet Bloc and mainland China are presented in Table 5.20. Income growth rates for these regions were increased or decreased by three percent, relative to base level assumptions given earlier in Table 5.6, to give the high and low growth scenarios. The short-run impact of low 'world' income growth on U.S. prices of meats is rather small, although the long-run decline in real prices becomes substantial. The impact is greatest for heavy feedstuff consum- ing industries, namely, pork and broilers. The greatest impact, logi- cally, is on corn prices and U.S. net exports of feedgrains. For exam- ple, corn prices decline by one percent and 11 percent in the short and long runs, respectively, under the low income growth scenario. The low corn prices and reduced feedgrain exports reduce feed costs to U.S. livestock producers. Production increases in response to these incen- tives and hence prices fall. Notably, it appears that these changes in income growth in major world grain importing regions has a negligible impact on U.S. non-fed beef imports. 380 NNNNNNou NomN :N mouNNN NNomn .m.m NNNNN c. NoNNcNm =o>Nm wee—Nasamma No>oN omen o» o>NNNNoN .Neoogon m an uoNNoNuav ecu uoNNmNoep ago: NconoN smog» New Nouns guzocm osooeNN N N N.- N .NNN .22. N285 NooN N.N N.N m.N N.N- N.N- Neon .e .NNNE NNNNNmuoou NNNoNxm uoz .m.= N.NN o o.NN- N.N- .Na\» ooNNN :Noo .m.= N.N- o N. o NNoNNNNE Newcasz zoo Noam m.e 0 p.01 5.- Ncmpwocm 66 o o.mt ¢.l xLoa N.N N. N.N- N Noon Noe-No: —.N o m.~n m.n moon pom .NN\N "NooNNN scam m.m o . r .u mNoNNon m.¢ o m.m: m.n xgom m.p o m.- p.. mama gmmgaaEm: N. O ~.—i D I FQQQ OPAMH . .n—\$ "mmUrL& Prmvmm Iii Na>oN oNNN soc» manage unauemocoa LII; oNNNaNmm com 11- ; o N NN> «NeoNNmm Nag-mead a“ - EN mmcmqwom NoneN NNca N Na N :1. m «Nosm cam mean cam-Neogm rt, guzogm «soucN gmNz suzogw oeoucN zed. .uiltzu NNNNN NNNNNNNz Nan uoNN NmNsom .Namxgaz NNNNNNN>NQ NN agpxogm meoocN NNN: NNNNN> so; No Numywm o~.m oNNNN 381 5.9 Summary of Main Points Preferred continuous time-varying parameter models of the four major retail meat categories, estimated in the previous chapter, were integrated into the retail meat demand component of the Michigan State University Agricultural Model (MSUAM). This U.S. retail meat model, together with other model components of the U.S. livestock and grains sector and world wheat and feedgrains regions, were used in simulations; of various policy and economic scenarios of national and world agricul- tural markets. Scenarios chosen reflected specific structural changes which may potentially impact these markets. Particular emphasis was given in those simulations to the impact which these structural changes may have on U.S. imports of processing quality beef. An overview of the MSUAM and its structure were given. Details of the retail meat demand component with its linking and transmission equa- tions were also presented. Comparisons of actual values with simulated results fer the endogenous variables over the 1975-82 period provide one method of validation of the estimated model. ,Validation also involved an assessment of dynamic simulation experiments of impacts on key model variables. Projections to 1991 suggest rising real prices for broilers at retail. This represents a significant turnaround in the historical decline in real broiler prices and indicates a declining comparative advantage of broilers relative to beef at retail. This loss in advan- tage is illustrated by the stabilizing of retail broiler consumption per person and the upward trend in table beef and hamburger beef consumption through the late 1980's. 382 Structured into the base model projections are reduced planted acreage and above average grain yields to reflect the newly instigated Payment-In-Kind (PIK) Program. Contrasting these base model results were simulated drought effects on U.S. feedgrain yields, 20 percent lower in 1983. This raised corn prices 13 percent in that year. Higher feed costs led to a shift from feedlot produced beef towards range and pasture produced beef. Beef cow numbers decline in response to the dec- lining profitability of beef cattle production and in response to the increased slaughter of cows. The one-year impact from 20 percent lower feedgrain yields, causing increased cow slaughter and increased produc- tion of processing quality beef, brought about only a small reduction in U.S. beef imports. Retail prices increased in the short-run, especially for feed grain and soymeal intensive products, pork and broilers, and then decreased in later periods. The impact of a 50 percent one-year increase in the 1983 adjusted beef import quota was relatively small. A similar conclusion was reached in other studies. The immediate impact is to decrease table beef prices by 4.6 percent and fed beef prices by 7.3 percent. Ham- burger prices fell by 3.8 percent while non-fed beef prices fell by 4.7 percent in the initial year of impact. In the following years, the results from both a one-period impact and a permanent increase in beef imports show a larger price effect on processing quality beef prices than on table beef prices. The long-run specification of fed beef pro- duction, which responds primarily to changes in cattle inventories, causes‘an apparent underestimation of fed beef production and hence an overestimation of the short-run price response. 383 A sustained 10 percent increase, in retail demand for meats, applied on an individual product demand basis, increased prices in the immediate term given fixed supplies available. Retail prices of compet- ing products also rise, temporarily. For example, the rise in fed beef prices increases the corn price and feeder calf price which subsequently dampens fed beef slaughter and supplies available for consumption. With lower quantities available, retail prices of competing products also rise. The short-run response is a decline in competing products. An increase in table beef demand causes beef imports to rise initially then to decline over the longer term. An increase in demand for hamburger beef creates an opposite effect. A sustained increase in real dispos- able income also has a substantial positive effect on retail and farm prices for meats, the impact being greatest in latter periods. Imports rise initially as cow numbers are being increased, but then trend down- ward. The effects of structural changes in the demand parameters of retail meats in the U.S. were characterized by three major directions of change. These structural changes followed from results contained in the previous chapter and in large part reflect a continuation of present trends in structural change in retail demand. These developments are first, a rising direct flexibility of demand for beef, in particular table beef; second, a down trend in income flexibilities for all beef; and third, an increasing cross-effect on other red-meats from hamburger beef demand and an increasing cross-effect on all other meats from broiler demand. In each case the result has been similar although the effect has been viewed from different perspectives. The result has been lower retail and farm prices for all products in the short-run but also 384 in the long run for table beef, lower corn prices and beef cow numbers, and generally, lower U.S. beef imports. An analysis is presented of the effects on the livestock sector of a dairy cow slaughter program to achieve balance in the dairy market in one year or over three years. The one-year program is in the immediate term substantially more market destabilizing than the three-year pro- gram. However, over the longer term a much greater effect from the three-year program on retail and farm prices would be felt. Assuming a 13-20 percent dairy cow herd reduction, the one—year program causes a 4-6 percent reduction of table beef prices in the immediate term. Ham- burger prices fall by 3-5 percent, with smaller declines in retail pork and broiler prices. Hamburger beef prices continue to fall in later years also. The program operated over a three-year period results in immediate decreases in retail table beef and hamburger beef prices of 1-4 percent and 1-3 percent, respectively. However, in the long run, prices fall substantially more with the one-year program. Perhaps the largest impact falls on U.S. beef imports. With the one-year progran beef imports fall in the first couple of years by between 2-5 percent. With the three-year program beef imports fall by between 3-7 percent in the longer run. Finally, simulations were made of high and low income growth scenarios for developing markets, the Soviet Bloc and Mainland China. The short-run impact, for example, of low 'world' income growth on U.S. meat prices is rather small, although the long-run decline in real prices becomes substantial. The largest impact logically is on corn 385 prices and U.S. net exports of feed grains. A negligible effect was felt by U.S. non-fed beef imports. CHAPTER 6 CONCLUSIONS The U.S. beef market and indeed the market for other major meats, pork and poultry, have been subject to substantial structural change over the past several decades. These changes have occurred in nearly all segments of the beef subsector. For example, the development of a hybrid corn industry and other developments such as price support pro- gram in grain markets placed downward pressure on feedgrain prices and encouraged a large-scale grain-fed beef production system. Developments such as boxed beef have not changed the consumer product but have facil- itated major changes in the way much of the beef is distributed and marketed. At retail, changes in consumer lifestyles and habits and in consumer attitudes to health and convenience of foods, and the growth in fast food outlets each have had their impact on the demand for meats. Traditional methods of marketing of meats have come increasingly under pressure to develop or adapt to accommodate these transitions. More- over, the strong downturn in beef consumption since 1976 record levels has emphasized the need for greater scrutiny of these marketing methods. Against this trend is the continued strong upward trend in the con— sumption of poultry meats, particularly broilers. During 1976-82 these gains in market share were largely at the expense of red meats. The steady decline in real chicken prices has been achieved by substantial production and distribution efficiences in that subsector. Notably, however, the widening spread between beef prices and chicken prices has levelled off. Over the past several years retail broiler prices have 386 387 remained at around 30 percent of retail beef prices, suggesting possibly that these efficiency gains in a highly industrialized broiler industry have begun to diminish. At least, the comparatively traditional beef industry should not continue to lose competitive ground to chicken in terms of relative prices, as it has for the past several decades. With this background of considerable change in the U.S. markets for meats, structural models of retail demand were estimated. A central focus of these estimated models was the testing of the hypothesis of constant structural coefficients of retail meat demand. A major conclu- sion of this study, supported by considerable evidence, is that a con- stant parameter formulation for the retail demand fUnctions for table beef, hamburger and broilers is not appropriate and is likely to result in misleading structural coefficients of retail demand. A notable and consistent exception throughout this analysis was aggregate pork demand for which the null hypothesis of fixed coefficients was accepted in each parameter variation model tested. It is possible that significant parameter variation has occurred for the individual heterogeneous pork products, such as bacon and sausage meats. Structural changes in regression coefficients were identified and quantified using graphical analysis, qualitative shift and interaction variables, linear and cubic spline functions within a discontinuous time-varying switching regression model, and Legendre polynomials within a continuous time-varying parameter model. Annual models of retail market demand for table beef, hamburger beef, pork and broilers were specified as price dependent on the assumption that supplies available for consumption determine retail prices. This assumption within an 388 annual model is less satisfactory for broiler demand. However, for overall model consistency and compatibility within the Michigan State University Agricultural Model (MSUAM), used in this study, all retail meat demand equations were estimated as price dependent market models. The appropriate measure of demand responsiveness, therefore, is flexi- bility of demand. Linear and log-linear forms were used in ordinary least squares estimation of demand parameters of each of the four retail meats over the data period 1950-82. Considering the change which the retail market has undergone over the past several decades since 1950, it is perhaps not surprising that the same combination of traditional variables has not explained the at times complex and dynamic set of forces influencing retail prices over this period. It is less believable that the individual influences of these variables have remained constant throughout these years. Compared to pre-197O analyses of U.S. meet demand, the retail cross effects of hamburger beef consumption on table beef price is larger and the cross effect of table beef consumption on hamburger beef prices are smaller in this study. This means that hamburger beef has become a closer substitute with table beef and has achieved greater favor in many consumers' minds. Advertising, promotion and other forms of market and consumer education of substitute products, like hamburger beef, will make them even stronger substitutes. This will tend to make the direct flexibility of demand for table beef rise further. On the other hand, as seems the case for hamburger beef, successful attempts to product differentiate results in less substitutability, and hence, the direct flexibility of demand will tend to be lower. 389 The growth in the 5 to 24 age cohort, as a result of the baby boom of the 19503, has been a significant factor in augmenting the demand for hamburger beef. A significant negative relationship was identified between the older age cohorts and hamburger beef demand. On the other hand, the increasing purchases of food away from home showed no signifi- cant influence on beef demand and hence retail beef prices. The data series, for example, for hamburger beef prices or for aggregate food expenditures away from home may have been inappropriate variables when considering the influence of institutional (fast food restaurant) pur- chases on hamburger beef demand. The hypothesis of irreversibility of demand for beef is one testable hypothesis explaining systematic changes in the slopes of retail demand curves and in the derived flexibilities. This hypothesis, that direct flexibilities will tend to be more flexible during decreas- ing consumption phases and less flexible (inflexible) during increasing consumption phases, was accepted. Most conclusive results were obtained for all beef demand and for table and hamburger beef demands using changes in the cattle cycle as the underlying cause of these systematic changes in direct flexibilities of demand. Following from the close relationship between changes in the beef cattle inventory and retail consumption, this means that direct flexibilities of retail demand were higher during periods of beef cattle cycle or inventory upturns and lower in cycle downturns. The importance of managing beef cattle cycles is apparent since the greater the cycle amplitude, especially during the upswing, the more volatile retail prices will be in response to changing quantities of beef supplied. It was this volatility that had much to do with price induced shift in consumer preferences during the middle and 390 late 19703. The resultant changes in habits then had their impact on other consumption behavior patterns causing parameter shifts in other explanatory variables. The hypothesis that income responsiveness has been declining steadily over time, was not supported in this study. Income flexibili- ties fluctuated, albeit systematically, over the 33 year period. For instance, the income flexibilities of table beef and hamburger beef declined from 1977 to 1979, but have since shown a steady upward pro- gression. In recent years, the flexibility on income was higher during periods when incomes were decreasing. This agrees with the hypothesis that consumers (and retailers) are more responsive to decreases in incomes because of the tighter budgetary constraint imposed. Income flexibilities for broilers, however, have risen steadily over the - period. Regression coefficients and hence flexibilities of demand were also shown to vary significantly between the short run and the long run. Estimation of a price adjustment model indicated a smaller direct flexi- bility in the short run than in the long run. The impact of imperfect knowledge and institutional delays in the short run prevent complete and instantaneous adjustment of market prices to clear the market. Only after some time has elapsed in subsequent period do final adjustments in prices take place. It is argued in some detail that this result does not contradict the traditional notion of Marshallian short and long-run demand curves. Models discussed above permitted regression coefficients to change a small number of times and, therefore, allowed for a similar number of 391 changes in structure over that time series. The applications of linear and cubic spline functions and of Legendre polynomials permit the coef- ficients to vary more continuously over time. An important property of Legendre polynomials is the finality of coefficients property. This is a special property of orthogonal functions which in this case essen- tially means that the coefficient terms of the polynomial are indepen- dent of each higher-order term of the polynomial added when estimating higher degree models. Time varying parameter specifications of both splines and the Legendre polynomials were found to make a significant contribution beyond the classical model of fixed regression coefficients. In reality- shifts in demand from one structure to another are likely to be continu- ous and less likely to be abrupt. This continuity is more in accord with the inherent and psychic nature of people and more consistent with the role played by expectations in smoothing abrupt changes in market behavior. The notable and consistent exception to a time-varying param- eter formulation was the pork demand, i.e., the slope of the pork demand curve has not changed significantly over time. This is not to say that increases and decreases in demand have not occurred, but that the demand for pork has responded almost entirely to the relative prices of substi- tutes. This gives support to the hypothesized residual role pork fills in the buying habits of consumers, responding largely to relative prices and not to changes in tastes or preferences. Hence slope changes are not significant. It is perhaps not surprising that results indicate an integral relationship between variations in table beef and hamburger beef 392 production and consumption on the one hand and direct flexibilities for beef on the other. In the continuous time varying models employing Legendre polynomials, the pattern of flexibilities of beef characterizes production cycles over the 33 year period. This is consistent with the notion of irreversibility in beef demand and the nature of cycle-induced structural changes in retail demand. Over the past several years direct flexibilities for table beef have risen while for hamburger they have declined. The lower the direct flexibility, the greater the extent to which increases in quantities available per person are absorbed by demand and hence the smaller the price adjustment required to clear the market. This means that, recently, quantities of hamburger beef have been more readily absorbed by demand than have table beef or poultry meats. The implication is that the preference for hamburger beef has been growing vis-a-vis other meats. This position is reminiscent of the high beef price period of the early 19703 when ground beef became a closer substitute with table beef. Rising direct flexibilities for table beef and broilers indicates a weakening of demand. That is, as quantities available for consumption increase, effective demand is insufficient to absorb the increased quan- tities causing relatively large price adjustments to clear the market. This result for chicken is consistent with the observations that the market for chicken is becoming saturated and that the impact of gains from technology in the broiler industry, which have helped keep real prices down, may be slowing. More significantly, the substantial price advantage enjoyed by chicken over beef may be eroding. 393 It has been shown in this research that it may be misleading to compute a single coefficient that purports to show the response of prices to changes in quantities and income over an extended time period. These findings have important implications for the analysis and forecast of meat demand. For example, an analysis of the cost of a public commo- dity program, may depend largely upon the elasticity or flexibility of demand for that commodity. The accuracy and validity of estimated regression coefficients in terms of the time over which that coefficient is relevant, may, therefore, bear upon the overall quality and effec- tiveness of that government program. Preferred continuous time varying parameter models of the four major retail meat categories were integrated into a stochastic simula- tion model of U.S. livestock and feedgrains and world wheat and feedgrains markets. The historical analysis, indicating considerable systematic variation in the flexibilities of U.S. retail demand for meats, formed the basis of simulations of the effects of structural changes in demand parameters in the future. Simulated effects were characterized by three major directions of change, each of which represent a continuation of present trends in structural change in U.S. retail demand. The first structural change simulated was a rising own-quantity flexibility for table beef. This trend indicates a weakening of demand resulting in relatively large price adjustments in the market price to small changes in quantities, and reflects the negative quality changes which have led to a substitution of poultry meat for red meats. Second, separate simulations are made of the increasing cross effect of 394 hamburger beef and of broilers on other retail meats. Among red meats, hamburger beef, appears to have achieved greater preference among many consumers. This follows from its qualities of leaness and convenience in preparation. Hence, the hamburger beef quantity variable is simu- lated with an increasingly stronger substitution effect on the demand for other red meats. This scenario is juxtaposed with the case of increasing competition of broilers on the retail demand of all other meats. Third, income flexibilities for both table beef and hamburger beef have shown a relative decline since 1977. This downtrend is indi- cative of a decrease in the long-term growth rate of the U.S. industry. For each simulated change the coefficients in question were allowed to change linearly by one percent annually over the projection period In each case the direction of impact on U.S. agriculture has been similar, although the magnitude of the impact has varied when viewed from these different perspectives of structural change in retail demand. The result was lower retail and farm prices for all products in the short run but also in the long run for table beef, lower corn prices and beef cow numbers, and generally lower U.S. beef imports. The greatest impact occurred in the scenario of a rising cross-effect of chicken meat on other retail meat demands. This was expected because the impact is widespread and because of the large cross-coefficients of broilers. Other more general simulations were also conducted. Structured into base model projections was the newly created Payment-In-Kind (PIK) Program and the concomitant reduction in planted acreage and above aver- age yields. A simulated 20 percent decrease in feedgrain yields under 395 this program in response to a one year drought in 1983 raised corn prices 13 percent in that year. Resources shift from fed beef produc- tion into range and pasture fed production in response to higher feed costs. The declining profitability of beef cattle production led to an increase in cow slaughter. This in turn brought about an increase in production of processing-quality beef but less than a 2 percent reduc- tion in the U.S. beef import quota. This small response is not surpris- ing given the long legs in the quota formula. Subsequent rebuilding of the beef cow herd results in a 2 percent increase in imports over the longer term. A sustained 10 percent increase in retail demand for meats, applied on an individual product demand basis substantially increased prices in the immediate term, given fixed available supplies. Retail prices of competing products also rise, temporarily. The rise in fed beef prices, for example, increases corn prices and feeder calf prices which subse- quently dampen fed—beef slaughter and supplies available for consump- tion. With lower quantities available, retail prices of competing pro- ducts also rise. An increase in table beef demand causes beef imports to rise initially then to decline over the longer term. An increase in demand for hamburger beef creates an opposite effect. A sustained increase in real disposable income also has a substantial positive effect on retail and farm prices for meats, the impact being greatest in latter periods. Imports rise initially as cow herds are rebuilt, but then trend downward. The potential impacts of a proposed dairy herd reduction program have created considerable concern for some participants in the livestock 396 markets of the U.S. The program's aim, ostensibly, is to achieve bal- ance in the dairy market in one year or over three years. Instability in the livestock and feedgrain markets from the one-year program when implemented in 1983 was substantially greater in the immediate term than the three-year program. Over the longer term, however, there is a much greater effect on retail and farm prices. From the assumed 13-20 per- cent dairy cow herd reduction, the one-year program causes a 4-6 percent reduction of table beef prices in the immediate term. Hamburger prices fell by 3-5 percent with smaller declines in retail pork and broiler prices. Significantly, hamburger beef prices continue to fall also in later years. The three-year program results in immediate decreases in retail table beef and hamburger beef prices of 1-4 percent and 1-3 per- cent, respectively. However, in the long-run, prices fall substantially more with the one year program. Possibly the greatest impact falls on imports of beef into the U.S. With the one-year program beef imports fall in the first couple of years by some 2-5 percent. In the three- year program beef imports fall by 3-7 percent in the long run. The potential contribution which time-varying parameter models make to the understanding of commodity markets seem substantial. Certainly, it is intuitively appealing to allow regression coefficients to vary in a manner more in accord with the actual nature of economic and, indeed, with many other relationships. Application of such an approach has been demonstrated although clearly much work remains to be done. The appli- cations in forecasting and policy analysis seem extensive. Hence, the importance of undertaking such research should require less convincing. 397 A major emphasis given in this research is the role of beef imports into the U.S. The significance of this market remains of prime interest to major exporters such as Australia. It behooves policy analysts and decision makers in both the U.S. and, in particular, Australia to gain as complete an understanding as possible of inherent changes in the underlying structure of commodity markets. It is towards this under- standing that this research has been directed. APPENDICES APPENDIX A TABLES AND FIGURES 1398 .n.c vcm N.e canes com "mouoz NNN.NN NNN.NNN NNN.NN NNN.NN NNN.NN NNN.NNN NN.N NN. NNN. NNN. NNN.N- NNN. - NNN. - NNN. - NNN.NN NNNNNNN uuoNNouN NN.NN NNN.NN NNN.NN NNN.NN NNN.NN NNN.NN NN.N NN. NNN. NNN. NNN.N- NNN.N- NNN. - NNN. - NNN.NN NNNNNNN Naoa NNN.NN NNN.NN NNN.N NNN.NN NNN.NN NNN.NN NN.N NN. NNN. NNN. NNN.N- NNN. NNN.N- NNN.N- NNN.NN NNNNNNNN NooN uoNussaoN NNN.NN NNN.NN .NNN.N NNN.NN NNN.NN NNN.NN NN.N NN. NNN. NNN. NNN.N- NNN. NNN.N- NNN.N- NNN.NN NNNNNNN Noun NNNNN .3.N Nm er NoNN NNNNN ozaoN NNNNNN NNzNNoN NNNNNzoN NNNNNNN aeouoeoo ooNuN NNoooN NoNoNNoN aways: ddbmuom won newumismaoo owumuaoa mounmuwe> unconceanwfl unpmuww> acovcococ NmaHIOnoH powwow ecu uoso moumawumn Hove: omen “venue: couscouwoa woueaowca huucaaumm mNHocuHoN «can: use: you ecowuesuu cameos dunno: voumlwumm N< uNaoe wNeeuaa< . 1399 Appendix Table A2 Generation Statements and Derived Data Series Used in Estimation of Fed and Non-Fed Beef Production and Table Beef and Hamburger Beef Consumption sewn RSLFISCT-SLCTIFISLCT s sewn SLSR-RSLFISCT*FISLS sewn SLHF-RSLFISCTRFISLH sewn SLCW-RSLFISCTtFISCW sewn Sins-RSLFISCTaFISLas S sewn sn- 5tsntowrusga. .001 g sewn HF- SLHF*DWFIH * 001 sewn cw- Steweowrlcwgt. .001 g sewn as- SLBS*0WFIBS «.001 sewn DWFICT-DWCT/DWFICTS sewn swaF-(g sn+ HF)*RDWFICT s sewn BFRBF-g CBF /0CBF s sewn swan- sneFtnanaF s sewn SLFSHS -nncr2 /(SLsn+5LwF) S sewn FsF- SHBFflRSL SH SH 5 sewn CFBF -RBFRBF*8FBF g sewn NFSHBF- swaF— FBF sewn oar-Dca -3FBF- gwrswse S sewn wstaFn-g sta *RBFRBF S sewn BFR- oa *RBFRBF sewn CMFBF -3wstaFn+goan S sewn DCFaFnc- cran /POC sewn DCNFBFRc-OCMFBFR/POPC $ 1.50 1.51 1.52 1.53 1.54 1.55 1.55 1.57 1.5. 1.5. 1..0 1.51 1.52 1.53 1.54 1.55 1... 1..7 1... 1... 1.70 1.71 1.72 1.73 1.74 1.75 1.7. 1.77 1.7. 1.7. 1.50 1..1 1..2 ISLFISCT 1.41474 1.3.777 1.40122 1.3.52. 1.374.. 1.3.13. 1.3.125 1.35.30 1.34212 1.32557. 1.32312 1.2.5.2 1.25041 1.24537 1.250.. 1.2345. 1.1.... 1.1.4.1 1.14.1. 1.1427. 1.11.34 1.11733 1.12107 1.12355 1.107.. 1.0.454 1.0.17. 1.07.3. 1.07.22 1.07453 1.0.545 1 5557.55 10145.1 12107.5 13034.2 12974.7 14175.9 19935.9 15591.5 13154.5 14153.7 14799.5 15145.5 15192.5 15149.7 15505.5 17517.2 15125.1 15555.2 15552.1 19059.0 19420.7 19575.1 15552.1 19952.0 15055.7 19127.5 19555.2 15722.5 17545.7 17915.1 17552.0 17415.5 2 1.74.52 1724.35 1..1..5 2.43.54 3453..2 3730..7 3.00.77 4145.33 4075.07 502...3 5534... 53.3.35 5.4..20 522..00 5453.41 7550.51 ..74..1 3.5...7 .55..25 .552.4. .534.51 .403..4 .557-.7 .541... 3.23.73 105.4.0 1231... 113.4.3 11.7..1 ..35.37 ...2.41 10117.. 10475.5 3 5754.15 5755.27 775..02 .735-03 .2.5..5 .107.75 .41..27 52.5.73 5212.37 5.50.34 5345.43 5523.27 53.5.71 5707..3 32.3.40 7555.43 5.3...5 5.23..4 5.57.1. .1...1. 5430.42 .04..40 .322..7 7515.55 11705.7 10751.0 ...1.33 .550.21 5.55.2. 53.3.45 .702..2 7411.24 4 OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO uoNuou-usoaoquSQLL. 1.50 1.51 1.52 1.53 1.54 1.55 1.55 1.57 1.5. 1.5. 1..0 1.51 1.52 1.53 1.54 1... 1.57 1.5. 1.5. 1.70 1.71 1.72 1.73 1.74 1.75 1.7. 1.77 1.7. 1.7. 1.50 1..1 15.2 1.50 1.51 1.52 1.53 1.54 1.55 1.55 1.57 1.5. 1.5. 1.51 1.52 1.53 1.54 1.55 1... 1.57 1.5. 1... 1.70 1.71 1.72 1.73 1.74 1.75 1.7. 1.77 1.7. 1.7. 1.50 1..1 1.52 524.245 950.159 1225.09 1517.51 1757.12 1915.25 2029.99 2052.49 2720.55 2022.05 2225.20 2192.12 2505.92 . 2505.55 4120.45 4522.25 4995 . 22 9222.51 5424.70 5452.74 5222.52 5542.55 5014.14 5229.21 5554.59 7142.54 5555.50 5949.20 5551.55 5557.55 5151.72 5215.70 1 RSLFSHSH 4(N) .1..315 0.... 555555555555555 O 51.2... 5523.55 5530... 5.75.12 7217.3. 7.72.57 .35.... 5710.75 .517... 10555.3 10....1 11.37.0 12557.0 1.5.3.2 14737.7 1.7.2.2 15.3..3 171.2.2 1.3.4.. 15272.. 12734.5 15503.4 15.54.. 17102.0 1.1.1.. 1535... 15334.5 15117.5 2 1422.41 1.33.51 3033.42 352..3. 3475.75 4057.31 405.... 3725.35 3472.32 4123.53 4274.04 3.4..24 5011.15 4443.35 3.44.47 3.13.45 3711.0. 3237... 255.... 2355.04 2240.37 1553.13 1155.13 3535.7. 5055.2. 4.....3 4227.30 2525.55 1514.15 2505.74 3175.53 21.0-.3 3 205.107 ......OOIDOOOOI... ..500.2 ..54044 ..523.1 ..530.. ..54373 ..4.721 ..50730 ..510.7 ..51304 ..51747 ..57770 ..5550. ...4457 ..55451 .....33 ..70740 ....057 ..7034. ..73554 ...01.. ...075. ...3545 ...55.2 ...5504 ...4127 ....3.5 ......5 ...00.3 ...34.5 ...2175 ...5003 ...0..3 ...04.1 4 7505.24 735...7 5.55.05 5370.44 5515... 5455.52 5245... 5721.4. 4 oswir 5135.54 5502... 5431.5. 52.3.53 .422.52 ..71.71 .510... .55.... .503.4. 105...7 117.... 12553.0 12550.0 13.2... 150...7 14540.5 15750.4 15575.1 17131.1 172.7.7 15141.2 1.17..7 15545.4 17543.. 1..11.7 177.0.. 2015..3 20051.. 1.527.. 15005.1 17..3.4 15510.0 1.27..5 5 am. .7.2.51 .7.2715 .7.1421 .7.2502 .7..2.5 .753... .77.... .772.54 .7..441 .754052 .755045 .745354 .744..2 .7401.7 .73.430 .73...1 .740374 .74013. .740025 .73..14 .73..30 .740104 .735571 .740013 .73..37 .73.... .73.7.3 .73..02 .73..1. .73..52 .74005. .73.... .73.... 5 401 “..me _._.< “—0 :owuaEsmcou U.S.meoo 6:0 mmowLm meumm $0 Ho: I _.< wgzmwu xwflcwaa< Oon.wo 38303 330.3 noouom .3.— 5950A 000.1' 00.900 .................................................................................................... 2.. on. 600.0. O.w..b F'fl.'. 3.2. we. 00. Oh. H04 H55 $0. 95. NN. who 0&5 ch 9 an 5 Mb. ~p0.ao '05..» .no.ov va..vp .«0... 500.0» on. co. 3.3. me. «n . mm. no. no. «o. c? mm. mm. on . Nn. an.- 00h. aha. «00 «.0 «N0. «an. cvv .nfl. who. 00' 00h 0—0. '0 no .00 .u. #0 '0' 00¢ 00. .QO' .Ouw .avv 'p— anon! (1:931) punoa 13a 931130 1102 .uvcaoa :0HHHHa ««o.H «no: «ooH 0H wow; «0 ouuoaaH Hasuo< .auoac «can vauuafivu «Su «0 Ho>oH Hmauoa usu 5H anau .usmHga 0>HH 5H 5H 50H£3 .suoa uaouxo ucwHu3 nauouqo :H can uuuoaaH «can can aoHuuavoua 805m n .H0>0H Haauua :5 5H «moHa H. «.3 H. «.3 H. «.«o o «.8 «.on 65.3.3: 053» 1vuum uuuoa 1xm umz .m.= ««H. ««o.H HnH. ««n.H «nH. omn.H ncH. «n¢.H 0oo«.H .mnH .HHm uuuoaaH «own .m.= «.1 m.«« H.1 «.mH «.1 «.5H H.1 H.«H «.m .uao\m uuuHHoum «.1 «.o« o «.o« 01 «.«« H.1 o.¢« «.mH .uzu\m xuom «.1 H.o« o «.«« «.1 o.n« «.1 «.«« o.qH .030\» «can vmm1aoz . «.1 «.«« o H.«« «.1 «.om «.1 «.«« «.«« .uau\m «can not AuuoH 1H0n «omHv mmuHum such o «.«H o «.«H o o.«H o «.«H H.«H .mnH .HHm muuHHoum H. «.«« H. «.«« o n.H« o o.o« 5.5H .mnH .HHm 300m 0 «K o 9m H. 2.« o 0.5 4..» .2: .H3 wean awn—1:02 H.1 «.H« o.« o.aH o <.oH o «.«H o.oH .nnH.HHm «can can coHuusvoum A sham coo. new. Boo. mos. moo.1 mom. o coo. Nam. ofiuam younwaadm uuuHoz 1umoum oHuwm H.1 H.«c «.1 «.92 H.1 o.«« o H.w« «.«« mQOHHHH: uuuonaaz :00 «won H0>0H moan H0>0A H0>0H moan Ho>ua Ho>0A 055m H0>oH H0>04 000m H0>~H Hmauu< muHcp moHnaHua> Ioum nuance «can floum omcuso «nun scum oucasu «man aoum «mango umun mncoamoa ”ouuoan «ouuoaaH "ouuoaaH “uuuoaaH Ho>0A uowwwub Ho>0H nommHua Ho>og nommHuH Ho>0H uomeua HamH Ho1mmmH unauo>< «mmH «mmH «mmH mugoasm wmmm .m.= mo Hm>04 50mmwgh mg» .m> wagons” mmmm .m.= mo Hw>mg mmmm vmumzmc< we :owu553mm< as“ we mcoPuMUVPQEH m< 5.550 x_u=mag< 403 .muuoaaw moon :0 auoaa 0553 vouuafiva ecu m0 .mugwuoa HHuuumn ucouuoa oHH 5H H0>0H HowwHuu 0:95 o o.Hn H. «.H« c «.on o «.«« «.«« .094 auoHHoun «. «.«« a. «.«« o o.~n o «.«« «.«« .aAH xuom o o.o« «.1 «.c« n. «.o« q. «.«« e.m« .534 «can uomusnsuz c. «.«« «. o.«¢ H. 5.05 c «.«« m.Hn .mng «00m «Haas pcomuwm Hum :0Huaaanaoo «.1 «.«« H.1 m.«« «.1 o.¢« «.1 o.H« w.«« .onw muoHHoum «.1 «.«« «.1 «.«« «.1 «.H« H.1 H.H« «.oo .ono .xuom v.1 «.«o H.1 «.«« «.1 «.on o.1 c.mo m.Hn .ono «00m humusnam: «.1 «.«« c «.«« H.H1 H.«HH «.1 «.«« «.«« .nH\o «mom quua AauoHHoa o«mHv moUHum HHauox H0>0A 005a Hu>oH Hu>0q anon Ho>0a Ho>0H 005m Ho>oa H0>0a 055m Ho>0a Huauo< muHca aanqHua> abum oncogu «can Baum omcnnv anon aoum «wanna 055m scum omaasu «man uncoammm "auuoaaH «auuoman "uuuoaaH uuuuoaBH Ho>5H uumeus Ho>oH nomeuH H0>oa ummmaua Ho>oH nowmwua HmmH Hm1nmmH ouauu>< «mmH «mmH «mmH muuonaH «00m .m.: «0 0H0>oH ummeua 0:0 .m> muuonaH moon .m.: «0 H0>0H mwmm kumznv< m0 GOHunanmm< mnu m0 mGOHumoHHaaH AmwaaHuaoov m< oHnma xHucme< APPENDIX B SELECTED MSUAM SUBROUTINES (If... (155.! (5.5.5.5.... f! 4404 Appendix 8 Selected MSUAM Subroutines SUBROUTINE BEEFIJ.ITR) . cannon NY. NZ. NYHAX. uznAx. 5:541. NYRHAX. CRIT. 112515. 1151. I IVRFIN. INPARIIS). COHPON(ZO). EXDGYIZSO). DEFINYIZSO).. Z NFYIZSO).NLY(250).NFZI250). NLZIZSO). NAHYIZSO). NAHZIZSO). 3 EYIISO).PREVY(250).YYIZSO).Y(250.AO),ACTY(ZSO.AO).Z(ZSO.AO) INTEGER coupon. DEFINY. Exocv COHHON/ALIVlN/BFCOHTL.DCOHTL.DHEIFTL.SOHFSLL.SOHFFLL cannon lounnv/ ove368.ovs7.0v680n.nv690u.0v71.DV710u.ov72. UV720R.nv73.ov7u,u75.0v77.ov79.0v7905,DV7981. ano.ov800N.ov81.ova1ou.nv82,0vB20u.ov&30u.nnv7t. nov72.Rov73.nnvyh.ovasou.ov87ou.nvasou.ov910u. nvssou.ovsson.ovs7ou.0vagou.ova} cannon ITINES/ TIHE. TSQR. LOGT REAL LocT 8118381121138888838333888181333 run-5 888383"183383383:83‘8233383883881:883 .55555555555fittfitfitfitttiitfitfiifittiititiititttttfitttitQ§*******5***5*t 5 THIS SUBROUTINE PROVIDES oonESTlc (U.S.) RETAIL NEAT DEMAND 5 5 RELATIONSHIPS AND THE NECESSARY LINKAGE/TRANSMISSION EQUATIONS 5 5 WITH LIVESTOCK DEMAND AND PRODUCTION RELATIONSHIPS IN “XHODL”. * 5 5 55.555555555555555.5.5555555555555555fittttitAttttttfitttitfitfifitfititt55 AAAAAAAAAAAAAAAAAAAAAAAAA FED .EEF DEMAND CDHPONENT ARAAAAAAAAAAAAAAAAAAAAAAA EQUATION YYIZOI) IS RATIO OF FED STEER t HEIFER SLAUGHTER TO TOTAL STEER 5 HEIFER SLAUGHTER : THIS TRANSHISSION EQUATION PROVIDES LINK HITN DOHESTIC CONSUHPTION 0E FED 5 NON-FED BEEF. IFITIHE .5E. 83 .AND. 11(201) .GE. .97) YYI201)- .97 IFIEXDCYIZOI) .EQ. O) RSLFSHSH CONSTANT CORN PRICE FEEDER CALF PRICE RATIO LAGGED tYYIZOI)-.27575-.IIZSGAfiY(73.J-I)-.OOAZS*Z(3I.J-I)+I.OOZ*Y(20I1J'II 1111:1100 FACTOR:::::: 1*EYIZOI) 5-.051*DV83 IDENTITY YVIZOZ) IS DONESTIC CONSUMPTION OF FED 5EEF.RETAIL.PER HEAD. FACTOR .75 IS FOR CONVERSION OF CARCASS TO RETAIL HEIGHTS. IF (EXOCY (202) .EQ. O) DCFBFRC FBEFQT RSLFSHSH Pore SYYIZOZ)-((Y(7.J)*Y(ZOI.J))*.7A)/Z(ZOI.J) IDENTITY VVIIOS) IS DON. CONS. OF NON-FED DEEF,RETAIL.PER HEAD. FACTOR .72 IS FOR CONVERSION OF CARCASS TO RETAIL HEIGHTS. IFIEXOGY (203) .EQ. O) ocurach NFBIHT NFBFQT FBEFQT RSLESRSH POPC 511(203)-((v(83.2)+v(9.J)+(v(7.J)*(I-v(201.J))))*.7z)/2(201.J) IDENTITY YYI2OA) IS DON. CONS. OF PORK.RETAIL.PER HEAD. FACTOR .70 IS FOR CONVERSION OF LIVE TO CARCASS HEIGHTS. FACTOR .90 IS FOR CONVERSION CARCASS TO RETAIL HEIGHTS. 405 IFIEXOGYIZOA) .EQ. D) DCPKRC PORKQT POPC .VYIZOAI‘IY(6.J)*.70*.90)/ZIZOI.J) C IDENTITY VYIZOS) IS DON. CONS. OF 5RDILERS.RETAIL.PER HEAD. C FACTOR .93 IS FOR CONVERSION OF CARCASS TO RETAIL HEIGHTS. fl IFIEXOGYIZOS) .EQ. O) c DDSRc CNIKQT POPC 8YY(205)'(Y(II.J)*.93)/Z(201.J) IDENTITY 2(203.J) 1S DDR.EDNS. OF OTHER RED HEAT.RETAIL.PER HEAD. c c DCORHRc-DCVLRC+OCLHRC+DCPKRC .HHERE DcVLRc 1S Don.cous VEAL.RETAIL. 5 PER HEAD SET CONSTANT AT 677.71 HILL. LBS.(AV.1973-1982).AND HHERE c DanRc 1S 005.cous LAH8.RETAIL.PER HEAD SET CONSTANT AT 366.89 HILL. 5 LDS. (Av. 1973-1982). 846.63 15 THEIR Av. . c DcoRch DCPKRC 2(203.J)-811.63/2(201.J)+1(201.J) c' IDENTITY ZIZOh.J) 15 DON.GDNS.POULTRY NEAT.RETAIL.PER HEAD. c ODH.CDNS.TURKEY.RETAIL.PER HEAD:DCTKC-(.95*TURKQT)IPOPC c DDR.DDRS.0TRER CHICKEN.RETAIL.PER HEAD 1s SET CONSTANT AT 685.7 5111 c 155. (AV. 1973-1982). 5 ocean DCBRC DCOCHC DETAc . ZIZOA.J)-Y(205.J)+655.7/Z(20I.J)+(.95*YIIO.JI)/Z(20I.JI c0555....500.555...5....500.505.55.550.0.505.055.05500550000 000000000000 C .ASE HODEL EQUATION --TA5LE DEEF 9 IFIEXOGYIZOG) .EQ. O) a 811(206)-136.577-l.36217*Y(202.J)-I.71667*Y(203.J)-.29505*Z(203.J) 5 5 -1.6868252(201.J)+.0556735z(202.4) ' CCOCOOOOOC ...... .05... 5555555555 5.5.5.... ...... 5.5... 5555555 ......500550 C EQUATION VYIZOG) IS RETAIL FED .EE PRICE .REAL. C LEGENDRE POLYNOHIALS:PTIF5-PTI*DCF5FRC'Z(206.J)*YY(202)3 PTZFl-PT2*DCFBFRC-Z(207.J)*YY(202)a PTIDICR-PTI‘OICR-Z(206.J)*IIZOZ.J). >>>>>>>> TEST ON LEGENDRE PDLYNOHIAL>>>>>>> )FITINE .DE. 83) 2(206.J)-.875 IFITINE .GE. 83) 2(207.J)-.265625 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>S>>>>>> 5'10“" 1'... 1E(TIRE .E0. 83) TV(202)-48.61 IFITIHE .E0. 8») 11(202)-45.67 1E(TIHE .EQ. 85) YYIZOZI-h1.9A IFITINE .EQ. 86) 11(202)-bo.77 IFITIHE .EQ. 87) 17(2oz)-h).us 33328 ADJFDCO IFITIHE .GE. 53) ADJFD-(TINE-52)*.O209728 1E(Exocv(zo6) .E0. 0) 406 c RPBFR CONSTANT DEEDERc DCNFBFRC DEDRHRE 511(206)-2ua.21I-2.2152341(202.J)-I.9902541I203.J)-.196145 DCPNC ' DIcR ‘ PTIFB 5 42(203.J)-2.0972842(2OI.J)+.03271134z(202.1)+1.28011 c ETzrD 5 *(ZI206.J)*Y(202.J))'-75I415*IZ(207.J)*YI202.J)) C ************* ADO FACTOR TO DOHNHARD ADJUST RPDFR *fi**************** 5 -IO.7*DV53 ¢ EY(206) c PERCENTAGE CHANGE IN EARN PRICE OF EED-DEEE (EDEEPT). GHGEDPT-(IY(77.J)-Y(77.J-1))/Y(77.J))*Ioo C EQUATION VYI207) IS HARKETING HARGIN FED 5EEF. REAL. IFIEXOGYI207) .EQ. O) c NHIFR CONSTANT EDEEPT RHDER LAGGED 811(207)--.76678+.387886*YI771J)-.IO7338*CNGF8PT+.51698*Y(207.J-I) c DCF.FR(AGGREGATE) 8+.0006426*(Y(202.J)*2(20|.JII c NFVIFR-RPIFR-HHBFR . THEREFORE NFVIFR-Y(206.J)-Y(207.J). C EQUATION YY(205) IS 5Y°PRODUCT VALUE FED 5EEFICNOICE 5EEF).REAL. C THESE IY-PRODUCT VALUES ARE ALSO USED FOR NON‘FED 5EEF. IF (EAOGY (205) .EQ. O) C 5PV5FR CONSTANT RPBFR RP5FR LAGGED GYYIZO5)-'4.93I05+.I97844*Y(206.J)-.D795649*Y(205.J’I) I . GP 5 +.46657552(6.J) IDENTITY 71(77) IS TARA PRICE OF EED 5EEE. FACTOR 2.42 IS EDR EDHVERSIDH 0F RETAIL HEIGHT EQUIVALENTS TO EARH PRODUCT HEIGHT (SEE LPS-ISDZ). non IFIEXOGYI77) .EQ. 0) EDEEPT NEVDER DPVDER 811(77)-(Y(206.J)-Y(207.J)+Y(208.J)I/2.42 fl AAAAAAAAAAAAAAAAAAAAAAAAAAAAA C C NON-FED .EEF DEMAND COHPONENT C AAAAAAAAAAAAAAAAAAAAAAAAAAAAA Coo-5555555555555o555550555555505550.555555555555555555 55555 55050000000 C 5ASE EQUATION -- HAHBURGER BEEF * IFIEXOGYIZOS) .EQ. O) * GYY(209I-I03.59-I.OI23I*YEEOZ.JI'I.74955*Y(203.J)-.30I59*ZI203.J) * G '-I.I6762*Y(2OS.J)+.0400658*Z(202.J) c5050555555055505555505...55.555.555555.55.0555.55.05.5055005505550.55.. C EQUATION YYIZOSI IS RETAIL NON-FED (HAHBURGER) PRICE.REAL. C LEGENDRE POLYNOHIALS:PTINF.-PTI*DCNFDFRC-l(205.J)*YY(203)z C PTIDICR - PTI*DICR-Z(205.J)*ZIZOZ.J)3 004 ' 11"310 11411512.:- ...5- 1131*“3’3‘. ._ 1810" _‘5_1 1301.1”. ...:5555. 1 g. *1 .1 3.1-11‘ a 11105 AH z-iO'.) -'-~1'1 .151. .~ 1.2512324172455111“: . 1 umozm(1.501114115111111 - 11115147 .- 55:5: .5 fififif. f. f. (if... 323:: IFITINE IFITIHE IFITIHE IFCI'IHE IEITIHE ADJH5-O IF(TIHE 1407 .EQ. 83) 11(203)-21.85 .EQ. 84) 11(203)-20.29 .EQ. 85) 111203)-22.93 .E0. 86) YY(203)-27.0S .EQ. 87) YY(203)-32.)4 .GE. 53) ADJHD'ITIHE-82)*.OIOI658 IFIEXOGY (209) .EQ. O) RPHHIR CONSTANT DCFOFRC DCNFOFRC DCORHRC GYY(209)-54.5024-I.32579*Y(202.J)-I.51235*Y(203.J)-.059379 DCBRC DICR PTINFD 5 51(203.1)-I.OI68851(205.1)+.0160527Az(202.J)+.618259 PTIDICR 5 .(ZI205.J)*Y(203.J))-.OO735I9*(2(206.J)*Z(202.J)) + EYIZOS) PERCENTAGE CHANGE IN BEEF COH PRICE. CHGDCP-(IY(78.1)-v(78.J-I))/¥(78.J))5100 EQUATION YY(2IO) IS HARKETING HARGIN NON-FED .EEF.REAL. IFIEXOGYIZIO) .EQ. O) HHHNIR CONSTANT 5FCOHP HARGIN LAGGED GYY(2IO)O*7.405I2+.902215*Y(78.J)-.I74959*CHG5CP+.592965*Y(2IO.J-I) NEVHHDR-RPHRDR-NHHHDR. THEREFORE. NEVHHDR-v(209.J)-v(210.1). IDENTITY YY(75) IS FARR PRICE OF NON-FED 5EEF(COHHERCIAL COHS-OHAHA) FACTOR 2.4 IS FOR CONVERSION OF RETAIL HEIGHT EQUIVALENTS TO FARH PRODUCT HEIGHT (SEE LPS-ISDZ) THE SAHE IY-PRODUCT VALUES FOR FED .EEF ARE USED FOR NON-FED 5EEF. IFIEXOGYI75) .EQ. O) 5FCDHP NFVHHBR 5PVDFR 511(78)-(v(209.4)~1(2IO.J)+1(208.J))/2.b AAAAAAAAAAAAAAAAAAAAA PORK DEHAND COHPDNENT AAAAAAAAAAAAAAAAAAAAA EQUATION YYIZII) IS RETAIL PORN PRICE.REAL. IFITIHE IFITIHE IFITIHE IFITIHE IFITIHE ADJPl-O .EQ. 53) YYI204)’53.IZ .EQ. 54) YYI204)-56.74 .EQ. 85) YYIZOA)'59.I9 .EQ. 86) YYIZOAI'55.84 .EQ. D7) YYIZOb)-58.90 IFITIHE .GE. 53) ADJPN-(TIHE-DZ)*.OOOAOSGD7 IFIEXOGYIZII) .EQ. D) RPPKR CONSTANT ' DCEDERC DCNEDERC DCPKRC 1408 GYY(2II)-92.059I-.333509*Y(202.J)-.350867*Y(203.J)-I.04475 DCDRC DICR 5 *Y(ZON.J)-I.5206I*Y(205.J)+(.0405687-ADJPK)*Z(202.J) C PERCENTAGE CHANGE IN FARM PORK PRICE. CHGPKP-((Y(76.J)-Y(76.J-I))/Y(76.J))*IOO C EQUATION 11(212) IS MARKETING NARGIN PORN.REAL. IF(EXOGYIZIZI .EQ. O) C MMPNR CONSTANT PORKPT MMPNR LAGGED GYY(2I2)-IO.7757+.SOIIII*Y(76.J)-.09423*CHGPKP+.135755*Y(2I2.J-I) C DCPKRIAGGREGATE) 8+.00044585*IY(204.J)*Z(ZOI.J)) C NFVPRR-RPPNR-HHPRR. THEREFORE. NFVPRR-V(2II,J)-1(212.J). C IY-PRODDCT VALUE PORK.REAL. IF(EXDGY(213) .EQ. O) C .PVPNR CONSTANT RPPKR . RPPKR LAGGED CPI GYY(2I3)-.ZOO3I7+.O78322*Y(2II.J)-.0242294*Y(21I.J-I)-.$I.54*Z(6.J) IDENTITY YY(75) IS FARM PRICE PORN.REAL. FACTOR 1.73 IS FOR CONVERSION OF RETAIL HEIGHT EQUIVALENTS TO FARM PRODUCT HEIGHT (SEE LPS-I952). flflfl IF(Ex051(76) .EQ. O) C PDRNPT NFVPIIR DPVPIIR 8YY(76)-(Y(ZII.J)-Y(212.J)+Y(213.J))/I.73 AAAAAAAAAAAAAAAAAAAAAAAA C C DROILER DEMAND COMPONENT C AAAAAAAAAAAAAAAAAAAAAAAA c.555555555 ...... .505... 55555 5 555555555555555 5 .......... 5 55555 0 ........ 5 C 5ASE'EQUATIDN -- 5RDILER 5 IF(TIHE .GE. 6h) DVCONST-I * IF(EXOGY(2II1) .EQ. O) 5 GYY(2I4)-89.5305-.381576*Y(202.J)-.42859*Y(203.J)-.2|6839*YI204.J) t 5 ~2.AI336*Y(205.J)+.0240165*Z(202.JI-26.2127*DVCONST 5 5 +.958Ib3aDVCONST51(205.1) c.5550055055500500555... 00000000 C EQUATION YY(2I4) IS RETAIL DROILER PRICE.REAL. C LEGENDRE POLYNOMIALSzPTIDR-PTliDCBRC II'l(20651I)*YY(205): C PTIDICR-PTI*DICR-Z(2065J)*Z(202.J). 5* ADJ-O . 5* IF(TIME .EQ. 53) ADJ-.63 *5 IF(TIME .EQ. 54) ADJ-.68 *5 IF(TIME .EQ. 85) ADJ-.71 5* IF(TIME .EQ. 55) ADJ-.74 *3 IF(TIME .EQ. 57) ADJ-.73 .409 ADJ-O IF(TIHE .CE. 83) ADJ-(TIHE-DZ)*.OI39I736 IF(EAOCYIZIA) .EQ. O) RPBRR CONSTANT DCFDFRC DCNFRFRC DCPKRC SYY(2IA)-65.232I-.33I5I3*Y(202.J)°.AOIZIS*Y(203.J)-.ZOI923 DCBRC DICR PTIDR C *YCZOA.J)-2.0DA7A*Y(205.J)+.O30l§88*2(202.J)+.693*(Z(206.J) PTIDICR S *Y(2OS.J))-.OO79AA7*(2(206.J)*ZI202.J)) * EYIZIA) PERCENTACE CHANGE IN EARN BROILER PRICE. CHNGIRP-(IY(80.J)-Y(80.J-I))/Y(80.J))*IOO EQUATION YYIZISI Is NARNETINC NARCIN IROILERS.REAL. IF(EXOGY(216) .EO. O) RRORR CONSTANT CNINPT HHIRR LACCEO tYY(2I5)-9.30512+.15I686*Y(80.J)+.06899*CHNGBRP+.63596*Y(2|5.J-I) OCORR(AOCRECATE) A - S‘.DOOBI35*(Y(205.J)*Z(ZOI.J)) PVDRR-RPDRR-HHDRR. NOTECEVDRR-CNIKPT-FARH PRICE DROILERS. IDENTITY vv(80) Is EARN PRICE OROILERs.REAL. IF(EXOGY(BO) .Eq. 0) CNINPT RPBRR HHDRR CYY(DO)-Y(ZIA.J)°Y(215.J) RETURN END 411) SUDROUTINE CPIFOODIJ) COHHON NY. NZ. NYHAX. NZHAX. NCHAX. NYRHAX. CRIT. IYRDAS. IYRI. I IVREIN. INPARIIs). CONPON(2O). EXOGY(250). DEFINY(2 0) g :ziggg?);::;:igggiuss£250;. NLZ(250), NAHY(250). NA:Z(550). O I 2°.¥2°.“. O O I .EéflTEGER CONPON. DEFINY. EXOGY 5 ( 5 O) “CTY(25° “0) 2(250 “0) E FARR VALUE: FVIF--2.hh2+2.208*¥(77.J)*Z(6.J) NARNET SPREAO: NSSE--IO.350+AI.8*2(6.J) PORN EARN VALUE EVPN-7.238+I.AssaV(76.J)a2(6.J) NARNET SPREAO: HSPK--9.I95+39.3*Z(6.J) IROILERS EARN VALUE: FVBR--.573+I.AA7*Y(BO.J)*Z(6.J) NARNET SPREAO: NSBR-8.I82+II.066*2(6.J) ECCS EARN VALUE. FVEG--.h96+I.003*Y(82.J)*Z(6.J) NARNET SPREAO: HSEG-Iz.67h+8.261fi2(6.J) TURKEYS EARN VALUE: FVTK-3.368+I.I}O*Y(8I.J)*Z(6.J) NARNET SPREAO: HSTK-IZ.585+IO.I97*Z(6.J) OAIRV PROOUCTS EARN VALUE: FVHBDP-5.525+I1.606*Y(79.J)*Z(6.J) NARNET SPREAO: NSNSOP-I8.AA8+5A.AE7*2(6.4) CEREAL C BAKERY PRODUCTS EARN VALUE: FVHBCP-h.696+9.225*¥(75.J)*2(6.J) NARNET SPREAO: HSHBCP--IO.6I5+I08.77I*Z(6.J) EATS A OILS EARN SPREAO: FVHBFOII.736+.590*Y(72.J)*Z(6.J) NARNET SPREAO: ASHBFO-.957+18.23I*Z(6.J) RETAIL PRICE VALUES OE EACN PROOUCT RPEE-EVRE+NSRE RPPN-EVPN+NSPN RPOR-EV3R+NSOR RPEc-EVEC+NSEC RPTN-EVTN+NSTN RPNaOP-EVNSOP+NSNROP RPNOCP-EVNRCP+NSNOCP RPnaro-EVNREO+NSNSEO VRlF-RPBF*O.8595 VRPK-RPPK*O.607O VRRN-VROE+VRPN VRIR-RPBR*O.A065 VRTK-RPTK*0.09IO VRPNoVRRR+VRTII RVHBEG--I.588+.h683*RPEG RVEST-VRRN+VRPN+RVNREC+RPNOOP+RPNOCP+RPNREo RVSTIDx-RVEST/h33.7l 2(77.J)--.02662+I.OI612*RVSTIDX RETURN ° ENO . -INTN .AANIU ; Zr‘ISHOD IQ! w I.=: Hamm ‘CI”V.:3?SIVV . TAM} .VNI‘”I‘ rev-EOE. I‘ .‘I‘POLE ”u r '11.!!! D I .‘w’RA '«JAIUT 3 - I' 14A! 9 1 r."‘..‘ 6'“! I I.“ ‘ II“. 3 I. am I Pfi‘i 'lI-LO } "'N ".1: 7 I am I . 'I ; Guy 3 j! '. J 3 -.".AI ." .L4 3 -‘.r ..-)R 1 mm 1’\:u .-.aoII “ .,._I.031I | ‘ ‘ "’13,?“ " ,_ «new ' .Iwiall1l ’. . . ‘ mica?" 1LI~1I§~~ '.-J.:‘ ' n ilflfl‘lOin -. ~ \Ja‘vfiM‘MW ’ APPENDIX C LISTING OF DATA USED IN RETAIL DEMAND ESTIMATIONS I050 I051 1052 1053 1054 1055 1050 1057 1050 1050 1000 1001 1052 1053 1005 1000 '1007 1000 1000 1070 1071 1072 1073 1074 1075 1070 1077 1070 1070 1000 1001 1002 1050 1051 1052 1053 1054 1055 I050 1057 1050 1050 1000 1051 1002 1003 1004 1005 I005 1007 1000 1070 1071 1072 1073 1074 1075 1070 1077 1070 1070 1000 1001 1002 §§§§§§§§§§§§ E§$§$§$§$§2338388882 §§§§§§§§ "04.2000 00.1000 74.4000 05.7000 07.2000 07.0000 05.4000 111.000 155.000 100.000 155.000 140.000 143. I44. I30. 152. 175. 411 Appendix C 2130.14 2211.23 2220.15 2254.02 2324.02 2373.20 2507.00 2027.71 2732.01 2005.40 2000.30 2022.00 2001.07 3024.02 3110.05 3270.02 3104.30 3100.20 3244.50 3313.20 3305.42 3300.30 3271.04 3271.40 3270.10 3 AOE5T024 30.0300 30.5300 30.0000 31.0000 31.1300 31.4200 31.0000 32.0200 32.4500 32.0300 33.3000 33.0000 34.4000 35.0000 35.5000 30.1300 35.0000 37.0000 37.3000 37.5500 37.7000 37.7700 37.3500 37.0700 30.0200 30.0000 30.3000 35.0000 35.3400 34.7000 34.0000 33.3500 32.0000 0 10051.1 11451.0 11010.4 12150.0 13035.4 13074.5 14101.0 14011.1 . 15300.0 15004.5 15320.2 15003.5 17004.0 17720.0 10001.2 10113.0 10700.0 20372.3 20012.0 10235.4 17405.2 17250.4 17570.4 17750.0 4 Listing of Data Used In Retai] Demand Estimations AGE25T004 0.0000000000000000 50.2500 50.0200 40.7500 40.4300 40.0400 40.0300 40.2100 47.7000 47.1400 40.5000 40.1300 45.0200 45.1700 44.0000 44.4500 44.1700 44.0200 43.0000 44.1300 44.1000 44.1200 44.0000 44.4000 44.0200 40.1000 45.4200 45.0000 40.1000 40.5000 47.0000 47.4200 47.0500 40.2700 3313.75 3030.00 4100.07 3770.30 4050.10 4301.11 4275.10 4522.07 5514.45 5001.12 5200.30 0400.40 7207.00 7070.01 7013.02 0703.70 0523.30 10201.3 10004.7 11077.7 11705.0 12720.1 12125.0 11205.4 0005 70.4000 00.4000 07.0000 70.0000 00.4000 50.4000 57.0000 71.0000 02.1000 04.0000 02.1000 00.3000 03.7000 5430.51 5255.74 0352.02 5004.71 0733.23 0004.55 5335.01 - 5050.71 5030.00 5572.45 5700.35 5004.00 0100.14 0147.35 5000.20 5723.10 5421.54 5205.01 5200.54 5000.75 4750.21 0010.57 0300.03 0903.40 0201.00 0502.32 5423.03 5003.44 0231.54 5032.02 0 1050 1051 1052 1053 1054 1055 I057 1050 1050 I000 I001 I002 1053 1004 I055 1000 1007 1000 1000 1070 I071 1072 1073 1074 1075 1070 1077 I070 1070 1000 1001 1002 00000 000000' ocrxg . 1144.44 442.444 4711.42 1241.44 1744.44 '''' 1°. """ . 444.444 444.444 14124.4 1472.44 1742.44 271,383 . 1414.74 444.424 14272.4 1744.44 1444.44 714.444 . 1244.74 447.724 4244.44 1417.44 1447.44 741.444 . 1427.24 444.774 4411.24 2177.44 1442.44 442.444 . 1244.24 444.244 14444.7 2244.44 1214.44 414 444 . 1244.24 444.444 14244.4 2444.44 1171.44 444.444 . 1242.47 424.474 4471.44 2222.44 1474.44 444.444 . 444.744 427.144 4442.24 2744.44 1442.44 1414.44 . 444.474 742.744 14441.2 2444.44 1444.44 1442.44 . 424.124 744.424 14724.4 4144.44 742.444 1444 44 . 441.174 414.444 4244.44 4444.44 722.444 1224 44 . 444.424 444.424 14444.7 4724.44 742.444 1242.44 . 744.444 424.444 11244.2 4444.44 441.444 1244 44 . 412.124 444.474 11424.1 4224.44 442.444 1244 44 . 422.444 422.244 14444 4 4444.44 714.444 1424 44 . 724.424 444.244 14421 4 4144.44 441.444 1444.44 . 424.444 442.444 11474.4 4221.44 742.444 1444.44 . 441.244 444.424 12141.4 4447.44 774.444 1447.44 . 427.474 .447.244 12444.4 4424.44 711.444 1442.44 . 444.444 444.414 12444.4 7422.44 724.444 1412.44 . 444.744 472.724 14444.4 7442.44 774.444 1747.44 . 272.444 441.744 12442.4 7424.44 747.444 1444.44 . 214.444 442.444 12414.4 7744.44 444.444 1774.44 . 442.444 422.244 12444.7 7444.44 744.444 1444.44 . 744.244 244.444 14414 2 7442.44 714.444 1412.44 . 712.474 244.424 11442.4 4414.44 414.444 1442.44 . 447.424 244.444 12174.4 4444.44 444.444 1444.44 . 441.444 224.744 12224.4 4442.44 441.444 2414.44 . 244.444 244.444 14227.4 14444.4 444.444 2244.44 . 244.444 224.444 14444.4 14444.4 444 444 2274.44 . 244.224 221.447 14742.7 11442.4 747.444 2444.44 . 274.174 244.144 12414.7 11424.4 721.244 2444644 1 2 2 4 412 Sources of Data Used in U.S.Am Model (a) (b) (e) (d) (e) (f) (s) U.S.D.A., Livestock and Meat Statistics, Supplement for 1981 (and earlier supplements), Statistical Bulletin N4. 522. U.S.D.A., Livestock and Meat Situation, various issues. U.S.D.A., Livestock and Poultry Situation, various issues. U.S. Government Printing Office, Economic Report of the President, February 1983. Department of Commerce, Statistical Abstract of the United States, various issues. U.S.D.A., Agricultural Statistics, various issues. U.S.D.A., Poultry and_§gg Situation, various issues. LITERATURE CITED LITERATURE CITED Agnew, D. B. (1979), "The Outlook for Hamburger," Livestock and Meat Situation, LMS-226, USDA, washington, DC, pp. 26-38. Allen, J. W. and Pierson, T. R. (1980), Current and Future Role of Meat Packaging in Supermarket Operations, 'Michigan State University, East Lansing, (unpublished). Allen, J. W. and Peirson, T. R. (1982), Meat Operations: Their Backs Against the wall, Agricultural Economics, Staff Paper 82-102, Muchigan State University, East Lansing. American Meat Institute (1982), Meatfacts 1982 Edition, AMI. Anon (1981a), "No-frills Food: New Power for the Supermarkets," Business week, Anon (1981b), "Restaurant Chains: Surviving a Stagnant Market," Business week, May 4, 1981, pp. 106-107. Anon (1981c), "McDonald's: The Original Receipt Helps it Defy a Downtown," Business week, May 4, 1981, pp. 161-162. Arzac, E. R. and Wilkinson, M. (1979), "A Quarterly Econometric Mbdel of United States Livestock and Feed- grain Markets and Some of its Policy Implications," American Journal of Agricultural Economics 61(2), pp. 297-308. Bain, R. (1977), An Econometric Model of the United States Beef Market, Beef Research Report No. 20, Bureau of Agricultural Economics, Canbera. Baker, T. G. (1978), An Econometric and Financial Projections Model of the U.S. Farming Sector, Ph.D. Thesis, Michigan State university, East Lansing. Barrett, J. P. (1974). "The Coefficient of Determination-Some Limitations," The American Statistician, 28(1), pp. 19-20. 413 414 Beef Industry Council of the National Livestock and Meat Board (1982), Consumer Marketing, Advertising and Promotion Plan prepared by Management Horizons, Inc., unpublished. Belsey, D. A. (1973a), "On the Determination of Systematic Parameter Variation in the Linear Regression Model," Annals of Economic and Social Measurement, 2, pp. Belsey, D. A. (1973b), "The Applicability of the Cameron Filter in the Determination of Systematic Parameter Variation," Annals of Economic and Social Measurement, 2, pp. 531—533. Block, D. (1977), The Meat Board Meat Book, Benjamin Company, New York. Box, M. J. (1965), "A New Method of Constrained Optimisation and a Comparison With Other Methods," Computer Journal, 8, pp. 42-52. Box, M; J., Davis, D. and Swann, W. H. (1969), Non-Linear Optimisation Technique, Imperial Chemical Industries Mono- graph No. 5, Oliver and Boyd, Edinburgh. Breimyer, H. F. (1955), "Elmer working: The Demand for Meat: A Review," Agricultural Economics Research, 7(3), pp. 73-77. Breimyer, H. F. (1961), Demand and Prices for Meat: Factors Influencing Their Historical Development, Technical Bulleting No. 1253, USDA. Brock, J. L. (1978), Unpublished notes from CPC International Study of Future Trends in Food Distribution, MSU, February 1978. Brown, T. M. (1952), "Habit Persistence and Lags in Consumer Behavior," Econometrica, 20(3), Brown, A. and Deaton, A. (1972), "Surveys in Applied Economics: Models of Consumer Behaviour," The Economic Journal, 82(328), pp. 1145-1236. Burbee, C. R. (1978), "Economic Impacts of a Ban on Selected Animal Drugs," National Food 351.133., “FR-5. Washington, DC, pp. 42-44. Burbee, C. R. (1980), Feeding Antibotics to Farm Animals: Benefits, Risks and Alternatives," National Food Review, NFerZ, pp. 21-23. 415 Burbee, C. R. (1982), "Beef Grade Alternatives," National Food Review, MFR-19, pp. 17-21. Bureau of Agricultural Economics (B.A.E.), (1981), Meat: Situation and Outlook 1982, AGPS, Canbera. Bureau of Agricultural Economics (1975), "Developments in the Japanese Beef Market: Their Implications for Production Systems in the Australian Beef Cattle Industry," Beef Research Report No. 17, Canbera. Burk, M. C. (1961), Measures and Procedures for Analysis of U.S. Food Consumption, .Agricultural Handbook No. 206, USDA, weshington, DC. Buse, R. C. (1958), "Total and Elasticities--A.Predictive Device," Journal of Farm Economics, 40(4), pp. 881-891. Campbell, G. R. and Hayanga, M. L. (ed), (1978), Vertical Organization and Coordination in Selected Commodity Sub- sectors, N. C. Project 117, werking Paper Series, WP-ZO, August 1978. Chambers, R. 6., Just, R. E., befitt, L. J. and Schmitz, A. (1978), International Markets in Diseguilibrium: A Case Studygof Beef, working Paper No. 69, Giannini Foundation, University of California. Chang, H. (1977), "Functional Forms and the Demand for Meat in the united States," Review of Economics and Statistics, 59(3), pp. 355-359. Chavas, J. P. (1982), A Study of Structural Change: The Demand for Meat, paper presented to the American Agricultural Economics Association Conference, Logan, Utah, August 1982, and published as "Structural Change in the Demand for Meat," American Journal of Agricultural Economics, 65(1), pp. 148- 153. Chow, G. C. (1960), "Tests of Equality Between Sets of Coefficients in Two Linear Regressions," Econometrica, 28, pp. 591-605. Christensen, T. H. (1979), A Simultaneous Predicitve Model of the United States Soybean Industry, M.S. Thesis, Michigan State University, East Lansing. Claffey, B. (1982), "Use of an Adult-Equivalent Scale in Meat Demand Analysis," Livestock and Meat: Situation and Outlook, EMS-246, USDA, July 1982, pp. 18-19. Colman, D. and Miah, H. (1973), "On Some Estimates of Price Flexibilities for Meat and Their Inter- pretation," Journal of Agricultural Economics, 24(2), pp. 353-367. 416 Conable, D. B. (1980), "Changes in U.S. Import Law," Foreign Agriculture Circular, Supple— ment, USDA. Cooley, T. F. and Prescott, E. C. (1973a), "Tests of an Adaptive Regression Medel," Review of Economics and Statistics," 60(2), pp. 248-256. Cooley, T. F. and Prescott, E. C. (1973b), "An Adaptive Regression Model," International Economic Review, 14, pp. 364-371. Cooley, T. F. and Prescott, E. C. (1973c), "Varying Parameter Regression: A Theory and Some Applications," Annals of Economic and Social Measurement, 2, pp. 463-473. Cornell, L. D. (1976), "The Impact of Death Duties on Sheep Farms in New South wales," Quantitngeview ongricultural Economics. Crom, R. (1970), A Dynamic Price-Output Model of the Befief and Pork Sectors, Technical Bulletin 1426, USDA. Cromoco, D. H. (1982), An Econometric Simulation Study of the Effect of Exchange Rate Over- valuation on Brazilian Agriculture, Ph.D. Thesis, Michigan State University, East Lansing. Daly, R. F. (1956), The Long-Run Demand for Farm Products, Agricultural Economics Research 8(3). pp. 73-91. DeJanvry, A. C. (1966), Measurement of Demand Parameters Under Separability, Ph.D. Thesis, University of California, Berkeley. Dixon, B. L. and Martin, L. J. (1982), "Forecasting U.S. Pork Production Using a Random Coefficient Model," American Journal ongricultural Economics, 64(3), pp. 530-538. Douglas, E. (1967), "Secular and Cyclical Changes in the Demand for Components of a Product Cluster," Review of Economics and Statistics, 49(1), pp. 63-76. FAO (various issues), Production Year Book, United Nations, Rome. Farrell, M. J. (1952), "Irreversible Demand Functions," Econometrica, 20, pp. 171-186. Ferris, J. N. (1974) Development of a Prediction Model for Forecasting Annual Prices and Production of Cattle, Staff Paper, No. 74-2, Michigan State university. 417 Ferris, J. N. (1981), Cattle Outlook: 1981-82, paper presented to Midwest Agricultural Out- look Conference, Champaign, IL, August 1981, Staff Paper No. 81-57, MHchigan State University. Ferris, J. N., Ross, J. W., and Perraut, E. E. (1983), An Analysis of the PIK.Program, WOrking Paper, Department of Agricul- tural Economics, Michigan State University. Fisher, B. S. (1979), "The Demand for Meat--An Example of an Incomplete Commodity Demand System," Australian Journal of Agricultural Economics, 23(3) pp. 220- 230. Foote, R. J. (1958), Fox, Fox, Fox, Fox, Fox, Analytical Tools for Studyinngemand and Price Structures, USDA, Agriculture Handbook No. 146. K. A. (1953), The Analysis of Demand for Farm Products, Technical Bulletin No. 1081, USDA. K. A. (1954), "Structural Analysis and Measurement of Demand for Farm.Products," Review of Economics and Statistics, 36(1), pp. 57-66. K. A. (1955), "Changes in the Structure of Demand for Farm Products," Journal of Farm Economics, 37(3), pp. 411-428. K. A. (1958), Econometric Analysis for Public Policy, Iowa State University Press, Ames, IA. K. A. (1962), "The Study of Interactions Between Agriculture and the Nonfarm Economy: Local, Regional and National," Journal of Farm Economics, 44(1), pp. 1-30 Foytik, J. (1951), "Characteristics of Demand for California Plums," ilgardia, 20(20), pp. 407-527. Freebairn, J. W. and Gruen, F. H. (1977), "Marketing Australian Beef and Export Diversification Schemes," Australian Journal of Agricultural Economics, 21(1). Freebairn, J. S. and Raussar, G. C. (1975). "Effects of Changes in the Level of U.S. Beef Imports," American Journal of Agricultural Economics 57(4), pp. 676-88. Fuller, W. A. (1969), "Grafted Polynanials as Approximating Functions," Australian Journal onggricultural Economics, 13(1), PP- 35-46- 418 Fuller, W; A. and Ladd, G. W. (1961), "A.Dynamic Quarterly Medel of the Beef and Pork Economy," Journal of Farm.Economics, 43(4), pp. 797-812. Galbraith, J. K. (1979), The New Industrial State, 3rd ed. revised, Mentor, New American Library, New York. Galbraith, J. K. (1973), "Power and the Useful Economist," American Economic Review, 63(1), pp. 1-11 0 Gallo, A. E. (1981), "Food Spending and Income," National Food Review, NFR-l4, Washington, DC, pp. 2‘40 Gardner, B. (1981), "On the Power of Macroeconomic Linkages to Events in U.S. Agriculture," American Journal of Agricultural Economics, 63(5), pp. 871-878. George, P. S. and King, G. A. (1971), Consumer Demand for Food Commodities in the United States with Projections for 1980, Giannini Foundation Menograph No. 26, Californian Agricultural Experiment Station. Gintis, H. (1972), "Consumer Behaviour and the Concept of Soveignty: Explanations of Social Decay," American Economic Review, pp. 267-268. Goddard, E. W}, Martin, L. J. and Griffith, G. (1979), A.Medel of Australian Trade in Beef with North America, Wbrking Paper No. 8, Agriculture Canada. Goldfeld, S. M. and Quandt, R. E. (1973), "The Estimation of Structural Shifts by Switching Regressions," Annals of Economic and Social Measurement, 2, pp. 475-485. Goodwin, G. W. (1965), Cattle Feeding: An Analysis of Oklahoma's Opportunities, Oklahoma Agricultural Experiment Station, Processed Series P-488. Goodwin, J. W., Andhorn, R. and Martin, J. E. (1968), The Irreversible Demand Function for Beef, Oklahoma Agricultural Experiment Station Technical Bulletin, T-127. Green, H. A. J. (1978), Consumer Theory, revised edition, Academic Press, New York. Green, R., Hassan, Z. A. and Johnson, S. R. (1978), "Alternative Estimates of Static and Dynamic Demand Systems for Canada," American Journal of Agricultural Economics, 60(1), pp. 93- 107. 419 Griffiths, W. E. (1971), Estimation of Regression Coefficients Which Change Over time, Ph.D. thesis, University of Illinois. Griliches, Z. (1967), "Distributed Lags: A Survey," Econometrica, 35, pp. 16-49. Grunfeld, Y. and Griliches, Z. (1960), "Is Aggregation Necessarily Bad?", Review of Economics and Statistics, 42 (1) 0 Pp. 1-12 0 Gujarati, D. (1978), Basic Econometrics, McCraw Hill, New York. Haidacher, R. C., Graven, J. A., Huang, K. S., Smallwood, D. M., and Blaybeck, J. R. (1982), Consumer Demand for Red Meats, Poultry,yand Fish, ERS, USDA. Hamm, L. G. (1983), Changes in Food Distribution: Some Implications for Farm Policy and Income, Paper presented to Michigan Farm and Food Policy Conference, Lansing, February 1983. Harris, D. (1980), “Ainescription of the U.S. Beef Market: An Australian Perspective," Quarterly Review of the Rural Economy, 2(4), pp. 376-384. Harris, D. (1982), "Factors Affecting the Demand for Beef in the USA," Meat: Situation and Outlook 1982, APGS, Canbera. Harris, D. and Dewbre, J. (1983), "Medelling the Linkage Between the U.S. and Australian Beef Markets," working Paper 83-9, Bureau of Agricultural Economics, Canbera. Harvey, A. C. (1981), The Econometric Analysis of Time Series, Halsted Press, Wiley and Sons, New York. Hassan, Z. A. and Johnson, S. R. (1976), Consumer Demand for Major Foods in Canada, Economics Branch Publica- tion No. 76/2, Agriculture Canada. Hayenga, M; L., (ed.) (1978), Pricing Problems in the Food Industry: With Emphasis on Thin Markets, Menograph No. 7, North Central Research Project, NC-ll7. HinChy, Mo (1978), "The Relationship Between Beef Prices in Export Markets and Australian Saleyard Prices," ngrterlpreview of Agricultural Economics, 31(2-3), pp. 83-105. 420 Houck, J. P. (1965), "The Relationship of Direct Price Flexibilities to Direct Price Elasticities," Journal of Farm Economics, 47(3), pp. 789-792. Houck, J. P. (1966), "A Look at Flexibilities and Elasticities," Journal of Farm Economics, 48(2), pp. 225-232. Houck, J. P. (1974), "The Short Run Impact of Beef Imports on U. S. Meat Prices." Australian Journal offiégricultural Economics, 18(1), 60-72. Houthakker, H. S. and Taylor, L. D. (1970), Consumer Demand in the United States: Analyses and Projections, 2nd. ed., Harvard University Press, Cambridge, Massachusetts. Huang, C. and Raunikar, R. (1981), "Spline Functions: An Alternative to Estimating Income--Expenditure Relationships for Beef," Southern Journal of Agricultural Economics, 13(1), pp. 105-110. Hunt, J. P. (1972), The Contrasted effects of Quota, Autarky and Free Trade Policies on U.S. Beef Production and Policies, Ph.D. thesis, University of Minnesota. Intriligator, M. D. (1978), Econometric Models, Techniques, and Applications, Prentice-Hall, New Jersey. Johnson, R. W: M. (1978), "Meat Stabilization in New Zealand," In Kellogg Rural Adjustment Unit (KRAU), Stabilization in the Beef Industry, University of New Eangland, Armidale, Australia. Jones, J. and Weimer, J. (1980), "A Survey of Health-Related Food Choices," National Food Review, NPR-12, Washington, D. C. pp. 16-19. Kalman, R. E. (1960), "A New Approach to Linear Filtering and Production Problems," Transactions of the American Society of Mechanical Engineers, Series D, Journal of Basic Engineering, 82, 35-45. Kalman, R. E. and Bucy, R. S. (1961), "New Results in Linear Filterine and Prediction Theory," Transactions of the American Society of Mechanical Engineers, Series D Journal of Basic Engineeripg, 83, 95-108. Kelejian, H. H. (1974), "Random Parameters in a Simultaneous Equation Framework: Identification and Estimation," Econometrics, 42(3), 517-527. Kuester,J. L. and Mize, J. H. (1973), ,thflmization Techniques with Fortran, McGraw-Hill, New York. 421 Kiehl, E. R. and Rhodes, V. J. (1956), "Techniques in Consumer Preference Research," Journal of Farm Economics, 38(5), pp. 1335-1345. Kostecka, A. (1980), "Restaurant Franchising in the Economy," Restaurant Business, March 1, pp. 138. Labys, W. C. (1973), ,Dynamic Commodity Models;_ Specification, Estimation and Stimulation, Heath and Co., Lexington, Massachusetts, Ladd, G. W. and Suvannunt, V. (1976), "A.Mede1 of Consumer Goods Characteristics," American Journal of .Agricultural Economics, 58(3), pp. 504-510. Lancaster, K. J. (1966), "A New approach to Consumer Theory" The Journal of Political Economy, 74(2), pp. 132-157. Langemeir, L. and Thompson, R. G. (1967), "Demand, Supply and Price Relationship for the Beef Sector, Post Wbrld War II Period," Journal of Farm Economics, 49(1), pp.169-183. Lattimore, R. and DeGorter, H. (1980), Impact of Beef Imports on the Canadian Beef Market, Working Papers, Agriculture Canada. LeBovit, C. and Boehm, W. T. (1979), "Changes to Meet Dietary Goals," National Food Review, NRF-8, Washington, D. C. pp. 24-26. Lhuch, C. Powell, A. A. and Williams, R. A. (1977), Patterns in Household Demand and Savings, Oxford university Press. Longen, K. and Stucker, T. (1980), "1968-1978: More Fish, More Poultry, Less Red Meat," National Food Review, NPR-9, Washington, D.C. pp. 28-29. Lucas, R. E. (1976), "Econometric Policy Evaluation: A Critique," Journal of Mbnetary Economics, Supplement 3, pp. l9-46. Macauley, G. (1975), Regional Livestock Model, Research Division, Canadian Department of Agriculture, Ottawa, (unpublished). Main, G. W. Reynolds, R. G. and White, G. M. (1976), "Quarterly-price Relationships in the Australian Retail Meat Market, ,Qmarterly Review of Agricultural Economics, 29(3) pp. 193-211. 422 Manderscheid, L. V. (1964), "Some Observations on Interpreting Measured Demand Elasticities," Journal of Farm Economics, 46(1), pp. 128-136. Manetsch, T. J. and Park, G. L. (1980), System Analysis and Simualation with Applicationf to Economic and Social Systems, Part II, Department of Electrical Engineering and Systems Service, Michigan State University, East Lansing. Manetsch, T. J. and Park, G. L. (1981), System Analysis and Simulation with Application to Economic and Social Systems, Part I, Department of Electrical Engineering and Systems Science, Michigan State University, East Lansing. Marschsk, T. A. (1978), "On the Study of Taste Changing Policy," American Economic Review, Marshall, A. (1920), Principles of Economics, 8th ed., Macmillan, London. Martin, W. (1983), The Effects of Alternative Methods of Disaggregating U.S. Beef Con- sumption on the Estimated Price of Beef Imports, paper presented to the Australian Agricultural Economics Society Conference, Brisbane, February 1983. McCoy, J. H. (1979), Livestock and Meat Marketing, 2nd edition, AVI Publishing Co., Westport Connecticut. Meinken, K. W. (1955), The Demand and Price Structure for Wheat, USDA, Technical Bulletin, No. 1136. Mitchell, A. Mandel, T. F., Crosson, K. M. and Willard, A. R. (1980), Social Trends and Food Retailing, prepared by the Centre for the Study of Social Policy for the Coca-Cola Retailing Research Group, October 1980. Mitchell, D. O. (1982), A Description of the International Component of the MSU Agricultur- al Model, Staff Paper, Michigan State University, June 1982. MSU (1983), , MSU Agricultural Model Long-Term Forecast of U.S. and World Agricul- ture, Department of Agricultural Economics, Spring 1983. Mundlak, Y. and Raussar, G. L. (1979), Structural Change, Parameter Variation,_and Forecasting, Working Paper No. 76, Giannini Foundation, University of California. 423 National Cattlemen's Association (1982), The Future for Beef, A report by the Special Advisory Committee, reprinted from the Beef Business Bulletin, March 8, 1982. National Cattleman's Association (1983), Action needed to amend dairy legislation, Action Call, 29 July, 1983. National Restaurant.Association (1980), Food Service Trends, 2(9), pp. 1-20. Nelson, K. E. (1977), Economic Effects of the 1976 Beef Grade Chan es, Technical Bulletin, No. 1570, ERS, USDA. Nerlove, M. (1958), Distributed Legs and Demand Analysis for Agricultural and Other Commodities, USDA, Handbook No. 141. OECD (1980), Trade by Commodities. Series C, Paris. Office of the Federal Register (1980), "Proposed Rules}' Federal Register, 47(131), pp. 29685-29688. Papadopoulos, C. (1973), "Factors Determining Australian Saleyard Prices for Beef Cattle," ,Qmarterly Review of Agricultural Economics, 26(3), pp. 159-170. Pessemier, E. A. (1978), "Stochastic Properties of Changing Preferences," American Economic Review, 58(2), pp. 380-385. Phlips, L. (1974), Applied Consumption Analysis, North-Holland Publishing Company, Amsterdam. Pindyck, R. S. and Rubinfeld, D. L. (1976), Econometric Models and Economic Forecasts, McGraw-Hill, New York. Poirier, D. (1973), "Piecewise Regressions Using Cubic Splines," Journal of American Society, 68, pp. 515-524. Poirier, D. (1976), The Econometrics of Structural Chan nge, North-Holland, Amsterdam. Pollak, R. A. (1978), "Endogenous Tasts in Demand and Welfare Analysis," American Economic Review, 68(20), pp. 374-379. 424 Pope, R. Green, R. and Eales, J. (1980), "Testing for Homegeneity and Habit Formations in a Flexible Demand Specification of U.S. Meat Consumption," American Journal of ‘Agricultural Economics, 62 (4), pp. 778-784. Powell, M. J. D. (1965), "A.Method for Minimizing a Sum of Squares of Non-Linear Functions with out Calculating Derivatives," Computer Journal, 7, 303-307. Prest, A. R. (1949), "Some Experiments in Demand Analysis," Review of Economics and Statistics, 31(1), pp. 33-49. Purcell, W.D. and Nelson, K. E. (1976), "Recent Changes in Beef Grades: Issues and Analysis of the Yield Requirements," American Journal of_Agricultural Economics, 58(3), pp. 475-484. Raussar, G. C. Mudlak, Y. and Johnson, S. R. (1981), Structural Chamge,,Updatingmand Forecasting, Working Paper No. 130, Giannini Foundation, University of California. Reagan, R. (1983), Economic Report of the President, USGPO, washington, D. C. Reeves, G.W. (1979), Price Stabilization, Bilateral Trade and Institutional Constraints: The Case of Beef in Australia and the United States, Ph. D. thesis, University of California, Davis. Riedy, K. (1980), "Solving the Mysteries of the Food Grading System," National Food Review (NEE-12, Washington, D. C., pp. 19-20. Reimund, D. A., Martin, J. R. and Moore, C. V. (1981), Structural Changgin Agmiculture: The Emperience for Broilers, Fed Cattle and Processing_Veg£ables, Technical Bulletin N01 1648, USDA, Washington, D. C. Rizek, R. L. and Rockwell, G. R. (1970), Household Consumption Patterns for Meat and Poultry, Sprimgyl965, Agricultural Economics Report 173, USDA, Washington, D. C. Roberts, T. (1979), "Cancer-causing Chemicals in Food," National Food Review, NFR-6, washington, D. C., pp. 36. Roberts, R. K. and Heady, E. O. (1979), A Five-Commodity Econometric Simmlation'Model of the U.S. Livestock and Poultry Sector, Cad Report 83T, Iowa State University. Rockwell, G. R. (1959), Income and Household Size: Their Effects on Food Consumption, Report 340, washington, D. C. 425 Rojko, A. S. (1961), "Time Series Analysis in Measurement of Demand," Agricultural Economics Research, 13(2), pp. 37-54. Russell, L. B. (1979), "The Macroeconomic Effects of Changes in the Age Structure of the Population," in Ballabon, M..B. (ed.), Economic Perspectives: An Annual Survey of Economics, Vol. 1, Harvard Academic Publishers, London. Ryan, T. J. (1980), "A Note on Bias in the Estimated Effects of Beef Imports on U.S. Beef Prices," Australian Journal of Agricultural Economics, 24(1), pp. 60-64. Schmelzer, J. R. (1981), The Commercial Foodservice Sector: Trends in Growth and Market Structure, WOrking Paper Series, N. C. Project 117, WP-56. Schmitz, A., McCalla, A. F., Mitchell, D. O. and Carter, C. (1981), Grain Export Cartels, Ballinger Publishing Company, Cambridge, Massachusetts. Schuh, G. E. (1976), "The New Macroeconomics of Agriculture," American Journal of IAgricultural Economics, 58, pp. 802-811. Schultz, H. (1938), The Theory and Measurement of Demand, Chicago University Press, Chicago. Sexauer, B. H. and Mann, J. S. (1979), "Food Spending in Single-Person Households," National Food Review, NFR-7, Washington, DC, pp. 32-33. Shaffer, J. D. (1979), Food System Organization and Performance: Toward a Conceptual Framework, paper presented at the Allied Social Science meeting, Atlanta, Georgia, December 1979. Shepherd, G. S. and Futrell, G. A. (1982), MarketingTFarm Products: Economic Analysis, 7th edition, Iowa State University Press, Ames, Iowa. Silvestre, H. (1969), Demand Analysis: An Attempt to Develop a Methodology for Detecting the Points in Time where Structural Changes Took Place, Ph.D. thesis, Cornell University. Simpson, J. R. (1982), "The Countercyclical Aspects of the U.S. Meat Import Act of 1979," American Journal of Agricultural Economics, 64(2), 243-248. 426 Simpson, J. R. and Farris, D. E. (1982), The World's Beef Business, Iowa State University Press, Ames, Iowa. Singh, B., Nagar, A. L., Chaudhry, N. K. and Raj, B. (1976), "On the Estimation of Structural Change: A Generalization of the Random Coefficients Medel," International Economic Review, 17, pp. 340-361. Smith, H. (1937), "A Note on Time Elasticity of Demand," Economics. Smith, R. (1978) , "Ground Beef, Consumer Purchases and Satisfaction," National Food Smith, R. and Weimer, J. (1980), "Beef Grading: Meeting New Goals," National Food Review, NPR-7, washington, DC, pp. 30-31. Stafford, T. H. and Wills, J. H. (1979), "Consumer Demand Increasing for Convenience in Food Products," National Food Review, NPR-6, Washington, DC, pp. 15-17. Stanton, B. F. (1961), "Seasonal Demand for Beef, Pork and Broilers," Agricultural Economics Research, 13(1), pp. 1-14. Stinson, T. F. and Lubov, A. (1982), Segmented Regression, Threshold Effects, and Police Expenditures in Small Cities, American Journal of Agricultural Economics, 64(4), pp. 738-746. Stone, J. R. N. (1954), The Measurement of Consumers' Expenditure and Behavior in the United Kimgdom, 1820-1938, Vol. I, Cambridge University Press, Cambridge, England. Suits, D. B. (1957), "Use of Dummy Variables in Regression Equations," Journal of American Statistical Associations, 52, pp. 548-551. Suits, D. B. (1966), Statistics: An Introduction to Quantitative Economic Research, Rand McNally, Chicago. Sun, T. (1979), "U.S. Fish Consumption and Price," National Food Review, NPR-6, Washington, DC, pp. 18-19. Sun, T. Y. (1982), "Three Kinds of Frankfurters: Retail Demand," National Food Review, NPR-17, washington, DC, pp. 20-21. 427 Swamy, P. A. V. B. (1970), "Efficient Inference in a Random Coefficient Regression Medel," Econometrica, 38, pp. 311-323. Theil, H. (1971), Principles of Econometrics, John Wiley and Sons, Inc., New York. Tomek, W. G. (1965), "Changes in Price Elasticities of Demand for Beef, Pork, and Broilers," Journal of Farm Economics, 47(3), pp. 793-802. Tomek, W. G. and Cochrane, W. W. (1962), "Long-Run Demand: A Concept, and Elasticity Estimates for Meats," Journal of Farm Economics, 44(3), pp. 717-730. Tomek, W. G. and Robison, K. L. (1977), "Agricultural Price Analysis and Outlook," in L. R. Martin (ed.) A Survey of Agricultural Economics Literature, Volume 1, University of Minnesota Press, pp. 227-409. Tolley, G. S., Wang, Y., and Fletcher, R. G. (1969), "Reexamination of the Time Series Evidence on Food Demand," Econometrics, 37(4), pp. 695-705. Trapp, J. N. (1976), An Economic Simulation Model of the United States Agricultural Sector, Ph.D. thesis, Michigan State University, East Lansing. Traub, L. G. and Odland, D. D. (1980), "Convenience Food Cost Update," National Food Review, NPR-9, washington, DC, pp. 17-19. Tryfos, P. and Tryphonopolous, N. (1973), "Consumer Demand for Meat in Canada," American Journal of Agricul- tural Economics, 55(4), pp. 647-652. U.S. Bureau of Labor Statistics (various years), Consumer Expenditure Survey Series: Dairy Survey, 1972, 1973, July 1973-June 1974, Reports 448-l,2,3. USDA (1956), ' Food Consumption of Households in the United States, Household Food Consumption Survey 1955, Report No. 1, Washington, DC. USDA (1972), Food Consumption of Households in the United States, Household Food Consumption Survey 1965-66, Report No. 12, Washington, DC. USDA (1979), Conversion Factors and Weights and Measures: For Agricultural Commodities and Their Products, Statistical Bulletin No. 616. USDA (1981), Livestock and Meat: Outlook and Situation," LMS-240, May 1981. 428 USDA (1982a), Livestock and Meat Statistics: Supplement for 1981, Statistical Bulletin No. 522, October 1982. USDA (1982b), Food Consumption: Households in the United States, Nationwide Food Consumption Survey, 1977-78, Report No. H—l, Washington, DC. USDA (various issues), National Food Review, Washington, DC. U.S. International Trade Commission (1977), Conditions of Competition in U.S. Markets Between Domestic and Foreigg Live Cattle Meat Fit for Human Consumption, Publication 842, washington, DC. Uvacek, E. (1968), "A New Look at Demand Analysis for Beef: American Journal of Agricul- tural Economics, 50(5), pp. 1501-1506. Van Dress, M. G. (1980), "Fast Food Industry Growth," National Food Review, NPR-9, Washington, DC, pp. 35-37. Veblen, T. C. (1977), The U.S. Beef System: Research Opportunities, Stanford Research Institute, Project 4390, July 1977. Wailes, E. J. (1983), A Simulation Model of United States Agriculture, Ph.D. thesis, Michigan State University, East Lansing. Walsh, E. B. (1977), An Analysis of Realized and Potential Scanner Savings, Food Marketing Institute, Fall Conference, Atlanta, Georgia, October 1977. Ward, C. E., Henderson, D. R., and Hayenga, M. L. (1978), "Vertical Organization and Coordination in the Cattle-Beef Sub- sector," in G. R. Campbell and M. L. Hayenda (ed.) Vertical Organiza- tion and Coordination in Selected Commodity Subsectors, N.C. Project 117, Working Paper Series, WP-20, August 1978. Ward, R. W. and Myers, L. H. (1979), "Advertising Effectiveness and Coefficient Variation Over Time," Agricultural Economics Research, 31(1), pp. 1-11. Waugh, F. V. (1964), Demand and Price Analysis: Some Examples from Agriculture, USDA, Technical Bulletin No. 1316. Weeks, R. (1979), "The Effect of the Proposed U.S.‘Meat Import Law on U.S. Imports of Meat," Quarterly Review of the Rural Economy, 1(4), pp, 297- 305. 429 Weeks, P. and Turner, B. (1981), "Effects of the Proposed Canadian Meat Import Law," Quarterly Review of the Rural Economy, 3(3), pp. 232-239. Weimer, J. (1976), "Taste Preference for Hamburger Containing Textured Vegetable Protein, Wohlgenant, M. K. (1982), 'Structural Shifts in Demand for Meats: Taste or Quality Changes?," paper presented to the American Agricultural Economics Association Conference, Logan, Utha, August 1982. Wolf, C. (1983), GSIM Reference Manual, Department of Agricultural Economics, Michigan State University, East Lansing. Working, E. J. (1954), Demand for Meat, University of Chicago Press, Chicago, Illinois. Working, E. J. (1927), "What do Statistical Demand Curves Show?," Quarterlngournal of Economics, 41, pp. 212-235. Zellner, A. (1962), "An Efficient Method for Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias," Journal of the American Statistical Association, 57, pp. 348-368. Zellner, A. (1963), "Estimators for Seemingly Unrelated Equations: Some Exact Finite Sample Results," Journal of the American Statistical Association, 58, pp. 9770992. Zellner, J. A., Gallo, A. E. and Levey, B. (1980), "Alternative Measures of Inflation," National Food Review, Spring, Pp. 20-21. Sli- [UNN‘ERE-ITV HBV‘AH lklwflll I‘ll H ll IT! 3 1293 02364 7310