AGRICULTURAL PRODUCTION FUNCTIONS Earl O. Heady ProfessorofEconomics andC. F. CurtissDistinguishedProfessor ofAgriculture, IowaStateUniversity John L. Dillon Research Officer, Commonwealth Scientific and Industrial ResearchOrganization, Melbourne, Australia Iowa State University Press, ^Ames, Iowa © 1961 by The Iowa State University Press. All rights reserved. Library of Congress Catalog Card Number: 60-11128. LITHOPRINTED IN THE UNITED STATES OF AMERICA BY CUSHING-MALLOY, INC., ANNARBOR, MICHIGAN, 1961 Preface The application of formal productionfunction concepts in agricul- tural research is a relatively recent development* The area of analysis was initiated by W. J. Spillman and other pioneer economists and physi- cal scientists in agriculture. However, economists andphysical scien- tists fairly well “went their own way* and several decades went by with little co-ordinated research in the prediction of physical production and in economic application of the results. However, with increased com- mercialization of agriculture and greater economic literacy of farm operators there is needfor design of physical and biological research so that the results can have greater economic interpretation and appli- cation. This text, considered as one in agricultural science rather than purely in economics, summarizes certain concepts and methods relat- ingto the prediction and use of agricultural production functions. It also reports, in summary form, the results from selectedproduction function studies. Emphasis is on concepts, principles, and methodolog- ical results. Practical applicaj£.cjn is considered to be a second step in communicatingprinciples and predictionfor farmer use. However, a few illustrative examples are includedto indicate how complex data and concepts can be interpreted and presentedfor practical use. The studies reported are largely those resulting from co-operative research efforts at Iowa State University over the past decade. These studies include an important group of products and are drawn together in this monographfor use of other scientists in economic and biological phases of agriculture. Numerous studies represent pioneering efforts in predicting production surfaces and in adapting them for economic in- terpretations. The designs explained are not suggested as optimum for the purposes, but are simply those which appeared to be appropriate at the time a particular experiment was initiated or which were consistent with the time and researchfunds available. Predictions for some of these experiments, if designs such as those summarized in Chapter 5 might have been used, couldhave been improved. Subsequent research has employedthese designs or modifications of them. Functionalforms are not presented as those most appropriate under all environmental conditions. Rather, they are usedto illustrate the types of relation- ships and recommendations which result when particular forms of y vi PREFACE functions are employed, or serve most efficiently under a given set of conditions. The studies of empirical production functions relating to crops, livestock, andpoultry were possible only because of the scientific in- terest, knowledge, and professional abilities of Damon V. Catron, John T. Pesek, Stanley Balloun, and Norman L. Jacobson. Along with Gordon C. Ashton, Roger Woodworth, Solomon Bloom, Vaughn C. Speer, John A. Schnittker, William G. Brown, Robert McAlexander, Dean E. McKee, Joseph Stritzel, John P. Doll, Gerald W. Dean, Harold O. Carter, C. C. Culbertson, Owen W. McCarthy and R. P. Nicholson, these persons should be considered as co-authors of this monograph. We have only drawn together the numerous studies conducted by these persons, add- ing a few chapters of interest to those conducting similar studies and listing a collection of reading pertinent to agricultural production func- tions. Accordingly, authorship of the persons named above is recog- nized in the appropriate chapters and we should more properly be con- sidered as editors of this monograph. Too, relative to the array of farm-firm production function estimates presented in the final chapter, we are grateful to Takashi Takayama, K. S. Suryanarayana, W. Darcovich, T. Godsell, G. D. Agrawal, Y. Wang, George Mason, and especially, Lennart Hjelm, Eje Sandqvist, and Yair Mundlak, for sup- plementary information willingly given. The authors hope that this text will serve to stimulate more and improved research in agricultural production functions. Too, we hope that it serves as foundation for further co-operative effort among per- sonnel of biological, physical, and economic sciences. The concepts, principles, and quantities presented are those relevant bothfor greater scientific knowledge and more efficient and practical use of certain ag- ricultural research. In drawing together the set of studies conducted at Iowa State Uni- versity, we do not depreciate or overlookthose which have been com- pleted or are underway at other research organizations and institu- tions. Time, space, and publication costs posed restraints which could not be eliminated at the time of preparingthis monograph. Earl O. Heady John L. Dillon May, 1960 Contents 1. Development of Production Function Studies 1 2. Economic Applications 31 3. Forms of Production Functions 73 4. Data Analysis for Production Function Estimation • . 108 5. Data Collection for Production Function Estimation 142 6. Economic Specification of the Production Function 195 7. Miscellaneous Empirical Problems Relating to Estimation of Production Functions 218 8. Pork Production Functions for Hogs Fed in Drylot 266 9. Pork Production Functions and Substitution Coefficients for Hogs on Pasture 302 10. Production Functions, Least-cost Rations, and Optimum and Optimum Marketing Weights for Broilers 330 11. Least-cost Rations and Optimum Marketing Weights for Turkeys 374 12. Milk Production Functions and Marginal Rates of Substitution Between Forage and Grain 404 13. Production Functions and Substitution Coefficients for Beef 452 14. Crop Response Surfaces and Economic Optima in Fertilizer Use 475 vii viii CONTENTS 15. Surfaces, Isoquants, and Isoclines From Fertilization 526 16. Functions for Fixed Plants and Other Farm Situations 554 17. Comparison of Production Function Estimates From Farm Samples Over the World 585 Bibliography 645 Index 665 CHAPTER 1 Development of Production Function Studies T HIS BOOK deals with agricultural productionfunctions. Its pur- pose is to summarize certainconcepts, empirical methods, and quantitative researches which relate to or have been derivedfor farm production functions. It covers both physical production functions based on experiments with crops and livestock andfirm production functions based on cross-sectional or time series samples. In particu- lar, it emphasizes research completed in these areas at Iowa State University. USE OF PRODUCTION FUNCTION CONCEPTS The production function is a concept in physical andbiological sci- ence. However, it was largely developed and, until recently, used mainly by economists. Historically, refinements in concepts relating to production functions grew out of economics probably because ofthe following reasons. (1) The nature of production functions is important in economic development and in determiningthe extent to which national products can be increased from given resource stocks. (2) The magni- tude of production coefficients serve as the base for determining opti- mum patterns of international or interregional trade. (3) The concept is basic to certaintheories inthe functional distribution of income. The conditions under which a total output can be imputedto the factors from which it is produced with the product just exhausted depends on the nature of the productionfunction. (4) The productionfunction pro- vides half or one of two general categories of the data needed in deter- mining or^specifyingthe use of resources and the pattern of outputs which maximize firm profits. (5) The algebraic nature of supply func- tions rests, in large part, upon the nature of the productionfunction. Research workers in the physical andbiological sciences of land- grant colleges andthe United States Department of Agriculture have long conducted research providing information onthe nature of agricul- tural production functions. However, historically, this research was designed and conducted somewhat apart from the formal concept of production functions represented by regression equations. More typi- cally, research was designed onthe basis of discrete phenomena 1 2 DEVELOPMENT OF STUDIES whereintwo or afewtreatments were usedto provide point estimates of crop or livestock output resulting from input of factors (materials representingtreatments) In some instances, although not designedfor . these purposes, the data were sufficient for deriving simple regression equations or input-output curves. More frequently, the experimental designs and statistical procedures used have only allowed indication of whether mathematically significant differences exist between the yield or output level of two or three discrete treatments or input levels. From these differences couldbe computedthe relative profitability of the fewtreatments or inputs. However, it was generally impossible to apply refined economic principles in determining the most profitable level of output and input, or the most profitable combination of inputs for a specified output. These designs andapproaches have proved useful inthe past and may continue to do so under certain conditions. In many cases, re- search workers in biological fields have been concernedonly with esti- matingthe output from a specific quantity of new material which serves as an innovation. Here the goal of the research oftenhas been to answer the question: Does the material or resource, used at any level whatsoever, give a response? Much early research onfertilizer fell inthis framework. In some cases, the practice or treatment under consideration represents a resource or material of discrete and limi- tational nature. Artificial insemination is an example. Here there is no important question of “dosage” and aformal production function ap- proach is inappropriate. For other materials used in production, the phenomena under consideration could have been estimated as a con- tinuous function, but there was little needto so represent it. For ex- ample, fertilizer rates recommendedto farmers, based on afewpoint estimates, have not always been as high as those which would maximize profits in the classical sense of afarm operating with unlimited capital under static conditions. However, because farmers operate in a decision-makingframework of uncertainty and limitedfunds, they often have failedto use fertilizer inputs even as large as those recommended onthe basis of trials which include only afew discrete treatment levels. Afinal reason might be given to justify experiments designed to give point estimates ofyield or outputfrom a few discrete input levels: the results provide points similar to the “straight line” segments assumed in linear programming. The optima selected withinthis framework of assumedphysical relationships always fall “on the corners,” repre- sented by the point estimates. However, refinements allowing optima to fall betweenthese “corners” may not be important where price un- certainty is so great thatex poste accuracy in decisions can never be attained. There are many biological and physical areas, however, where con- tinuous relationships are involved andthe data lendthemselves to for- mal productionfunction analysis. Also, in many of these areas, recom- mendations to farmers could be made with greater economic meaning if the experimental design and statistical analysis were of a form to allow DEVELOPMENT OF STUDIES 3 prediction of the production functions involved. Inthe past, these pro- cedures were not often used for several reasons. (1) The scientists conductingthe research often have used criteria other than economics in interpretingtheir findings and in making recommendations to farmers. Inthe past, for example, the criteria used oftenhave been rations which gave (a) the largest gain per pound of feed, (b) the great- est output per cow or hen, or (c) the fastest daily gain. The most prof- itable output or resource input is seldom identical with these maxima or minima. However, even where the objective is prediction of a phys- ical maximum or minimum, the exact quantity can be estimated more accurately where the data, if of appropriate nature, are usedto esti- mate the regression equations representing the productionfunctions in- volved. Derivatives then can be computed and equated to zero, with the appropriate magnitude of input then derived. (2) The statistical methods serving as a guide for research were based on early biological procedures which supposedthe datato be discrete phenomena most ap- propriate for point estimates. Early texts on experimental design and statistical analysis emphasizedthese methods as appropriate for ex- periments in biological sciences. (Early emphasis on continuous rela- tionships and regression analysis, as characteristic of physical data representative of production functions, was made especially by econo- metricians.) (3) Many physical scientists have not been acquainted with productionfunction concepts andthe economic principles which define profit maximization or cost minimization. Since they were par- ticularly concerned with conditions of profit maximization and compet- ing economic alternatives within the farm or firm, economists have been concernedwith marginal products, marginal rates of substitution, isoquants, and isoclines; quantities which are derivedfrom continuous productionfunctions. Inthe past decade, however, an increasing number of physical and biological scientists in agriculture have become acquaintedwith pro- duction function concepts. They have'become interested in interpreting basic relationships of nature accordingly and in using them to make recommendations to farmers. This increased interest and activity partly grows out of scientific advance and is parallel to that inthe field of agricultural economics. Generally, newfields of research start with a less formal and precise set of concepts and models. Over the dec- ades, continuous scientific investigation and thought provide refine- ment to the theories, concepts, and models which serve as the basis of designs in experiments and of principles in decision making. Pioneer agricultural economists, like other agricultural scientists, also were concerned with physical relationships which relate to productionfunc- tions. Without a well-defined set of concepts, they derived principles termed “factors affectingfarm profits.” The “factors affectingfarm profits” principles implicitly supposed certain conditions in respect to physical production functions. Generally, these early findings in agri- cultural economics served sufficiently in leading farmers to improve their farming operations. But these less formal principles also have, 4 DEVELOPMENT OF STUDIES withthe advent of time and improvedknowledge by bothfarmers and research workers, been replaced by more exact principles of profit maximization; principles which directly require knowledge of the rele- vant production and price quantities. Given recognition ofthese princi- ples, both physical scientists and economists will have greater future interest in quantities derivedfrom production functions. While the re- lationships represented by production functions are physical phenom- ena, economic principle is involved when recommendations to farmers, or decisions by them, relate to the quantity or mix of resources and products to be used and produced respectively. Accordingly, more co- operative research and educational activities can be expected among economists, agronomists, engineers, and animal scientists. NEED FOR PRODUCTION FUNCTION RESEARCH This text emphasizes production function estimates of the two fol- lowingtypes: (a) biological functions derivedfrom experiments where the plant is fixed at the magnitude of an animal or acre and (b) farm pro- duction functions derived from samples where plant size, as measured by “fixity” of a particular resource, is either fixed or variable. Per- haps the most appropriate use of biologicalfunctions is that of guiding farmers intheir individual decisions. However, these same data can be of extreme use for certain purposes of policy and economic develop- ment. For example, a nation such as India may have limitedfacilities for producing or purchasing fertilizer materials. With a given amount of fertilizer for annual allocationto agriculture, how much shouldbe distributedto various soil, geographic, and climatic areas? If the goal is to maximize the food product available from limitedfertilizer re- sources, productionfunctions derivedfor major regions and crops, with an indication of marginal coefficients, can provide a basis or guide for attainingthe goal. Similarly, the same type of information can be used during wartime or during any other emergency period when material shortages exist. On the other hand, farm production functions of the type discussed in this text probably best serve for the diagnostic pur- poses pointedout later. Inthis vein, they can provide general guidance for farmers’ decisions, credit policy formulation, readjustment of agri- cultural regions, etc. However, aside from studies based on highly re- fined samples, such as some mentioned in later chapters, they seldom can be usedto indicate “exact equilibrium use” of resources by an in- dividual farmer. With the passage of time, a greater knowledge level of farmers, and an increased commercialization of agriculture, there is increased need for experimental designs and research in biological fields which lend themselves to estimation of productionfunctions. Data can then be better adaptedfor economic interpretations and recommendations. An increasing number of farmers has knowledge of principles of profit maximization and wishto use fertilizer, feed, and similar resources in