AN ECONOMETRIC ANALYSIS OF THE FEED-GRAIN LIVESTOCK ECONOMY OF GREECE By Vassilios Constantinou Kalaitzis A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1978 Ln., 5“ n 0 Q ABSTRACT AN ECONOMETRIC ANALYSIS OF THE FEED-GRAIN CATTLE ECONOMY OF GREECE by Vassilios C. Kalaitzis The purpose of this dissertation effort was to develop an econometric model capable of analyzing demand/supply conditions pre- vailing in the Greek feed-grain cattle economy over the period l951- l972. The dissertation itself presents a summary of the methodology and procedures used in developing five submodels incorporating the supply/demand conditions in five industries -- roughage, feed-grain, beef, veal and milk -- along with a subsequent predictability test for each submodel. Utilizing average annual data, each submodel examines domestic and import demand and domestic supply. The impact of a number of domestic macro-economic and agricultural policy variables on production and consumption can be studied through exogenous variables used as proxies for these factors. And, by means of domestic interaction components, the impact of various foreign policies and economic factors on Greek feed-grain cattle economy may also be calculated. More specifically this research attempts to portray realistically: Vassilios C. Kalaitzis a) impacts of feed production upon production of beef, veal and milk in Greece; b) impacts of increasing incomes on beef, veal and milk consumption. The method employed was ordinary least squares along with polynomial distributed lag equations to estimate supply response in beef and milk production. Supply response for both crop and livestock products was calculated by utilizing a finished product supply equation and a supply equation consisting of number of acres planted and yield per acre or number of animals slaughtered (milked) and average per animal quantity of finished product. In the absence of data beyond the year l972, the model's predicting ability was tested using as actual values the exogenous variables fOr the years 195l and l969. Predictions were then made for the years 1970, 197T and l972, and these values were compared with actual values fOr the same years, a comparison which enabled a judgment as to whether the model is reliable in making projec- tions. Rising incomes have been the main demand shifter in beef, veal and milk consumption in Greece, while policy variables, such as subsidies, have also played a crucial role in increasing short-run production of both crops and livestock products. This study indicates that if the production of beef, veal and/or milk is to be increased in the short-run, a high beef-feed, veal-feed and milk-feed ration is required. However, in the long-run in- creased output of these products depends on an abundant supply of feed-grain and roughage. Vassilios C. Kalaitzis This study also reveals that production and consumption of the aforementioned livestock products heavily depend on feed-grain imports and on imports of finished products of beef and/or veal. And, if a keen demand arises for feed-grain as a result of adverse production conditions in other areas of the world or as a result of high feed-grain prices in the world market, beef, veal and milk production in Greece will decline in both the short- and the long- run. Given unstable world conditions, then, a positive program to increase the output of feed-grain combined with an increase in the consumption of other than red-meats and a general policy en- couraging growth and development of the feed-grain cattle economy in Greece will be required if production expansion policies are to be carried out. ii AE OOANEI O HAIOZ MONAXA H PHZ KAPHO NA AQZH XPEIAZONTAI KAI AAAA HOAAA KAI HPO HANTOZ H PNQZH KQZTHZ HAAAMAZ THE SUN IS NOT ENOUGH FOR THE EARTH TO GIVE THE FRUITS MANY OTHER THINGS ARE NEEDED AND, ABOVE ALL, KNOWLEDGE COSTIS PALAMAS Dedicated to My Family and My Teachers ACKNOWLEDGMENTS The author wishes to express his appreciation for the en- couragement given to him throughout his work at the University of Thessaloniki, Greece and his studies at Michigan State University by Professor E. Papageorgiou and Professor G. Kitsopanidis in addi- tion to all the faculty members of the College of Agriculture and Forestry at the University of Thessaloniki. The author also wants to express his gratitude for the aid provided him throughout his studies at Michigan State University by .Professor A. Schmid, who served as his major professor during the ‘ course of his graduate studies. The valuable assistance provided by Professors J._Ferris (serving as Thesis Advisor), J. Bonnen, V. Sorenson, L. Manderscheid, M. Steinmuller, and B. Brown, who served on his Guidance Committee, and Professor S. Thomson, who served on his Thesis Committee, is also greatly appreciated. To his fellow students at Michigan State University, he offers his thanks for both the academic and nonacademic experiences enjoyed while in East Lansing. His gratitude also goes to the Ministry of Coordination, Office of Technical Assistance, Greece, for the financial support provided throughout his studies at Michigan State University. iv Special thanks also go to Mrs. Anne Cauley for her careful editing of the thesis and to Mrs. Noralee Burkhardt for her expert typing job. TABLE OF CONTENTS List of Tables ........................ List of Figures ....................... Chapter I INTRODUCTION AND STATEMENT OF THE PROBLEM ...... Introduction ................. ~. . . . The Problem ..................... Scope of the Problem and Study Hypotheses ...... Purpose of the Study ................. Research Objectives ................. General Research Objectives ........... Specific Research Objectives ........... Data Collection ................... Limitation of the Study ............... Significance of the Study .............. Outline of the Study ................. II THE AGRICULTURAL SECTOR IN THE WHOLE ECONOMY ..... General Views .................... Excess Labor Devoted to Agriculture ......... Agricultural Land .................. Farm Size in Greece ................. Capital in Greek Agriculture ............. Agriculture and Foreign Trade ............ The Geographical Pattern of Agricultural Trade . . III THE INDUSTRIES INVOLVED ............... The Cattle Industry . ' ................ Size and Composition of the Cattle Industry ..... Regional Patterns of the Cattle Industry ....... Cattle Breeds .................... Production Systems of Beef, Veal and Milk ...... Beef, Veal and Milk Production from the . Indigenous Herds ............ '. . . Beeffi Vgal and Milk from the Local Improved er ..................... Specialized Beef Production ........... vi Chapter IV Dairy Production ................... Livestock Productivity in Greece . .......... Regional Patterns of Growth in Livestock Productivity . . . . . . . . . . . . . . . . . Number of Cattle per Holding ............. The Typical Greek Livestock Farm ........... The Feed-Grain Industries .............. METHODOLOGICAL APPROACH ............... Diagrammatic Presentation of the Feed-Cattle Economy Inputs and Outputs ............ Further Diagrammatic Development of the Model . . Livestock Supply and Derived Input Demand ...... Supply and Derived Input Demand for Feed-Grain~ and Roughage ................... The Econometric Model ................ Single Equations vs. Simultaneous Equation System Model Design and Identification ........... DEMAND ........................ Consumption ..................... Consumer Choice ................... The Theory of Demand ................. Definitions and Demand Relationships ....... Demand Elasticities (own-price, income, cross) . . . . Demand Estimation .................. Factors Associated with Demand for Feed Grain and Roughage ................... Price Factor ................... Cattle Population ................ The Price of Related Goods ............ Cattle Feeding Practices and Management ..... Farm Income ................... The Availability of Feed-Grain and Roughage . . The Prices of Livestock Products ......... Empirical Estimation of Demand . . .......... The Variables Used ................ Endogenous Variables ............. Exogenous Variables ............. The Demand for Feed-Grain and Roughage ...... Imports of Feed-Grain ............ Factors Associated with Demand for Beef, Veal and Milk ................... Product Price ................ Consumer Personal Disposable Income ..... Population .................. The Price of Related Goods .......... vii Page 33 34 35 36 37 38 42 88 89 94 . 94 Chapter VI The Range of Foods Available to Consumers . . Consumer Tastes and Preferences ....... The Demand for Beef ................. Retail Level .................. Retail Demand for Beef ............. Retail Elasticities for Beef .......... The Relationship Between Farm and Retail Demand for Beef ................. Imports of Beef ................... The Demand for Veal ................. Retail Demand for Veal ............. The Equations Fitted .............. Elasticities of Demand for Veal at the Retail Level . Farm Demand for Veal .............. The Equations Fitted .............. Imports of Veal ................... Import Elasticities for Veal .......... The Demand for Milk ................. Retail Demand for Milk ............. Elasticities of Milk at the Retail Level Farm Demand for Milk ............. : Elasticities of Milk at the Farm Level SUPPLY ......................... Introduction .................... The Role of Prices in Supply Response ........ Formulation of the Models .............. Direct vs. Indirect Method of Estimation ...... Empirical Estimation of Feed-Grain Supply ...... Elasticities of Supply for Feed-Grain ...... Empirical Estimation of Roughage Supply ....... Elasticities of Supply for Roughage ....... Empirical Estimation of Beef Supply Response Price Elasticities of Beef .......... : Empirical Estimation of Veal Supply Response Veal Calves Slaughtered (VCSLt) Equation . Price Elasticities for Veal ........... Time Lag Consideration i ....... ' ....... General Discussion ............... Distributed Lag Models ............. Fitting the Polynomial Lag Model ........ Degree of the Polynomial and Length of Runs . . . Some Decision Criteria ............. Selecting the Polynomial Lag Model ..... Empirical Results of Distributed Lag Models . The Beef Cattle Slaughter (BCSL ) Equation Number of Cows Milking (NCMt ) E uation . . . : 2 Third Degree Polynomial ............. Summary . . . ', ................. ‘ viii 2T5 227 231 In. Chapter Page VII -TRENDS AND PROJECTIONS ................ 236 Trends in Domestic Demand and Supply of Beef and . Veal ....................... 236 J The Model as a Whole ................. 240 ‘ A. Roughage Submodel .............. 243 B. Feed-Grain Submodel ............. 244 J C. Beef Submodel ................ 244 i D. Veal Submodel ................ 245 E. Milk Submodel ................ 246 Projection of the Feed-Grain-Cattle Economy ..... 254 Sensitivity Analysis ................. 262 VIII SUMMARY, CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH ....................... 263 Summary and Conclusions ............... 263 Objectives of the Study ............. 263 The Problem ................... 264 Appraisal of the Results ............. 264 Demand Analysis ............... 264 Supply Analysis ............... 269 Beef Cattle Slaughtered ......... 272 Appraisal of the Research Method Used ...... 282 Limitations of the Models and Resultant Recommendations for Future Research ....... 283 Projection of the Feed-Cattle Economy ......... 285 Policy Implications ................. 286 Free Trade Option ................ 287 Self Sufficiency Option ............. 288 Joining the EEC Option .............. 292 APPENDICIES ......................... 299 BIBLIOGRAPHY ......................... 310 ix Table (JON 10 11 12 13 14 15 16 17 18 19 LIST OF TABLES Active Agricultural Population in Greece Labor Productivity Gross Product at 1963 Prices Distribution and Number and Area of Agricultural Holdings by Size: 1950, 1960, 1970 Number of Agricultural Holdings. Farm Size and Irrigated Land, Greece, 1961, 1971 Cattle Population of all Ages on Dec. 31, 1961-1972 (in thousand head) Location of the National Herd According to the Altitude Oxen, Bulls, Heifers, Cows and Buffaloes (all ages) on December 31, 1966, by Geographic Region Number of Holdings, Number of Cattle and Number of Cattle per Holding. Greece 1971 Number of Farms Studied and Number of Cows Milked Retail Demand Elasticities for Beef Import Elasticities of Beef Short-Run Price Elasticities of Veal at Retail Level Veal Demand Elasticities at Farm Level Import Elasticities of Veal Elasticities for Milk at the Retail Level Elasticities of Milk at the Farm Level Estimated Elasticities of Supply for Feed-Grain Supply Elasticities for Roughage 'Price Elasticities of Beef X Page 14 15 16 19 24 25 26 36 38 99 111 116 120 124 127 129 152 162 173 9: ~v ‘~ Table 20 21 22 23 24 25 26 27 28 29 3O 31 32 Polynomial Lag Model Parameters and Statistical Results Second Degree Polynomial Constrained Time Lag: 10 Years Polynomial Lag Model Simple and Cumulative (e) Elasticities Second Degree Polynomial Time Lag: * 10 Years Polynomial Lag Model Parameters and Statistical Results Second Degree Polynomial Constrained Time Lag: 10 Years Polynomial Lag Model Simple and Cumulative (e) Elasticities Third Degree Polynomial Time Lag: 10 Years Polynomial Lag Model Parameters and Statistical Results Third Degree Polynomial Constrained Time Lag: 8 Years Polynomial Lag Model Simple and Cumulative (e) Elasticities Third Degree Polynomial Time Lag: 8 Years Polynomial Lag Model Parameters and Statistical Results Second Degree Polynomial Constrained Time Lag: 10 Years Polynomial Lag Model Simple and Cumulative (e) Elasticities Second Degree Polynomial Time Lag: 10 Years Polynomial Lag Model Parameters and Statistical Results Second Degree Model Time Lag: 4 Years Polynomial Lag Model Simple and Cumulative (e) Elasticities Third Degree Polynomial Time Lag: 10 Years Mid-Year Estimates of Pogulation in Greece: 1963-1972 and Number of Tourists Coming into Country from 1969-1972 Per Capita Total Meat Consumption and Beef and Veal Consumption, Western European Countries, 1960 and 1972, in kg Production, Consumption, Imports and Self- Sufficiency Rates for Feed- Grain in Greece: 1965-1972 in Thousand Metric Tons and TDN and in Percentages xi 179 183 200 201 209 210 211 214 220 221 222 229 234 Table 33 34 35 36 Predicted Values of the Endogenous Variables for the Greek Feed-Grain Cattle Economy, Annual Average 1971 and Predicted 1972 Predicted and Actual Values of the Endogenous Variables for the Greek Feed-Grain Cattle Economy for the Years 1970-1972, and Theil's Inequality Coefficient Predicted and Actual Values of the Endogenous Variables for the Greek Feed-Grain Cattle Economy, Annual Average 1972 and Predicted 1972 Predicted and Actual Values of the Endogenous Variables for the Greek Feed-Grain Cattle Economy for the Years 1970-72, and Theil's Inequality Coefficient xii 236 239 256 257 Figure ‘0me 1O 11 12 13 14 15 16 17 LIST OF FIGURES Bull Calves and Heifers Flow Diagram Input-Output Relationships in Livestock Products Production (Beef, Veal, Milk) The Feed-Cattle Economy in Greece and Its Domestic and Foreign Environment Demand, Supply and Price Structure of the Greek Feed-Cattle Economy The Major Economic Relationships in the Feed- Livestock Economy Milk-Feed Price Ratio Distributed Weights BeeffFeed Price Ratio Distributed Weights Cumulative Elasticities: Second Degree Polynomial Milk-Feed Price Ratio Distributed Weights (Third Degree Polynomial) Beef-Feed Price Ratio Distributed Weights (Third Degree Polynomial) Cumulative Price Elasticities: Third Degree Polynomial Beef Price Distributed Weights: Third Degree Polynomial Feed-Grain Price Distributed Weights: Third Degree Polynomial Cumulative Elasticities: Third Degree Polynomial Milk-Feed Price Ratio Distributed Weights Beef-Feed Price Ratio Distributed Weights Cumulative Elasticities: Second Degree Polynomial xiii Page 32 44 48 50 142 203 203 205 212 212 213 216 216 217 223 223 225 Figure 18 19_ 20 21 22 Milk-Feed Price Ratio Distributed Weights: Third Degree Polynomial Beef-Feed Price Ratio Distributed Weights: Third Degree Polynomial Cumulative Elasticities: Third Degree Polynomial Flow Chart for Projecting Quantities and Prices in the Feed-Grain Cattle Economy Formation of Time Lag in Veal, Beef and/or Milk Production xiv Page 228 228 230 260 277 l-Lo . . ~1' . r’. ll; CHAPTER I INTRODUCTION AND STATEMENT OF THE PROBLEM Introduction The feed-grain cattle economy (sub-sector) is a major element in the Greek agricultural economy. In fact, the vast expense asso- ciated with the importation of red meat and feed-grain coupled with a growing domestic and world demand for both products should make livestock, and cattle in particular, a crucial sub-sector in the overall growth of the Greek agricultural sector. This thesis does not intend to examine Greece's comparative advantage in the production of beef, veal, milk, feed-grain and roughage. Instead, this study will contribute to the growing stock of knowledge concerning the dynamics of the Greek feed-grain cattle economy. More specifically, it intends to provide both des- criptive and quantitative information on this sub-sector which would be used for further quantitative, predictive and prescriptive analysis. The Problem The agricultural situation in Greece over the sample period examined here (1951-1972) could be described as one wherein: a) low and unstable farm incomes persisted; b) low product prices have held in the market; c) uncertainty has existed as to the future; 1 0?: i ‘o- .I. d) there has been inadequate production planning leading to chronic mismatching of supply and demand; and. e) there have been huge payments for imports of red meat (veal and beef) and concentrated dairy products (especially evaporated milk). Almost everyone in Greece has experienced an increasing cost of living or has noticed that prices in the country have increased tremendously during the 19705. Price indexes recordd this price in- crease during the last five to six years. One of the expenditure groups that has been of particular importance to Greek consumers over the last twenty-five years is red meat, an item that all Greek consumers want to purchase. Therefore, when the price of meat increases, all consumers - and particularly those within the low income brackets - begin to feel the impact of high prices in the economy. But food prices not only represent a cost of living to con- sumers; they also act to determine farmers' income and allocate re- sources within the economic system. Thus, while consumers desire low food prices, farmers desire high prices for their products, and society desires prices that result in an efficient allocation of resources. As a result of these conflicts, national economic policy is formulated to maintain a stable.price over time. Yet a stable price level is not the only possible goal of a government's economy policy; there could be other goals as well, such as: (a) increase in income, (b) improved distribution of income, (c) full employment, (d) a balance of payments, (e) saving of traditions and values. Inn. Iii- 1f. ' b .I\, o.‘. n '5 ‘1. \‘. . 1 (II The supply of meat is relatively fixed at any point in time due to the technical and biological aspects of production and dis- tribution. It typically takes a relatively long period of time for red meat producers to respond to changes in product prices or input costs. Thus, a policy designed to control these prices or costs would not seem to affect the supply of meat in the shortrun, though in the longer run such a policy could affect this supply. Due to the high inflationary pressure which Greece experienced from 1973 on, the government was forced to reduce its money supply. The result of this reduction was that interest rates went up; this represented an increased cost of production for meat producers in Greece.* Such an increase in production costs, ceteris paribus, should tend to reduce meat supplies in subsequent time periods, further driving upwards meat prices whiCh in turn affects people's standard of living. A government facing such perplexities looks at imports as a solution to a high meat prices problem. And an examination of the import figures reveals that, indeed, a considerable amount of Greek currency is used to pay for meat imports. But Greek farmers and agricultural economists argue that the government should not import cheap meat and feed-grains to subsidize consumers and thus reduce the cost of production of industrial goods at the expense of farm incomes and farmers' welfare. They also argue that the government should use stronger protective policies to promote a self-sufficient * There is evidence, however, that the relative importance of the interest rate is small except for the highly specialized new firms. pAf s'j our og’g ‘F‘A . h \ c 'F‘. A-' u"y. s meat production program to cope with the current world-wide red meat shortage. According to the aforementioned view, further liberiza- tion of trade would eliminate all small farms in Greece. Import policy changes regarding beef and veal may affect domestic production, distribution and consumption of beef and veal through expectations and related uncertainty. There are strong interdependencies prevailing between the beef and veal and milk and feed industries everywhere. Feed supplies, for example, are affected by unpredictable elements such as weather, which influence production decisions to adjust livestock inventories. By the same token, con- ditions of unstable demand for beef and veal at the retail level - due mainly to government intervention - creates problems for beef producers which are transmitted back to the demand for feed. Scope of the Problem and Study Hypotheses Low prices for beef and veal at the farm level cause - in the short run - great fluctuations in the price(s) of these products and, thus, in producers' incomes. This means that farmers slaughter their beef and veal cows in order to meet their current expenses. From the other end of the spectrum a great demand for red meat was observed in Greece over the sample-period examined, a demand which cannot be matched so far by domestic production. Domestic producers could meet this demand for beef and veal, however, if the prices of these products are set at a profitable point at the farm level. For this to be so, two things have to take place: first, either the price of products should increase to high 1.. ii 1" ("I III enough levels or, second, the feed-grain prices should be substantially reduced or some combination of both. But even if these two events happen to take place, for a stable flow of beef and/or veal to appear in the market, coming from the domestic producers only, would take a considerable amount of time. Supply response to price changes is not an automatic process; it takes time. It is the purpose of this research effort empirically to test this proposition. It is the belief of domestic producers that imports of beef and veal should be abandoned altogether since policy on these imports have caused erratic production cycles of these products in Greece. Import policy, they say, has not been consistent over the last two or three decades in Greece; it has not dealt adequately with the long range demand for meat. And this inconsistency causes anxiety and un- certainty for domestic producers and, hence, has serious effects on their planning and decision-making process. In terms of importation of meat, it is significant to ask here whether imports of beef, veal and feed-grain have either been "pushed in" or "pulled in." This thesis will examine this proposition. Feed-grain and forage are the most important cost items in the production of beef, veal and milk. Yet feed-grains have to com- pete for resources with other products which use the same resources in their production process. The same holds true for roughage pro- duction and/or demand. u '4- O‘- '1 1 I'D "Technological change" always has an impact on the produc- tion of livestock products. Since there is no way to retreat from science and technology, there has to be a better understanding of scientific evolution and its impact. This dissertation tests and quantifies the impact (If technological change upon the produc- tion of the livestock products at hand. Rising income is hypothesized to be the most crucial demand shifter for beef, veal and milk in Greece, over the period cited here. It was the aim of this research to quantify the income rela- tionships with respect to the demand for the products considered here. I Finally, some analytical aspects of the supply function were examined in order to better perceive the production process for beef, veal and milk in Greece. Purpose of the Stugy The overall purpose of this study was to test how relevant the price mechanism has been to the development (growth) of the feed- grain cattle economy in Greece over the period 1951-1972. Low prices at the farm level of livestock products, price uncertainty, instability of farm income and future government pol- icies, profitability of crop enterprises and, import-export policies complicate decision making and planning processes for livestock producers. Under these unfavorable conditions, they are skeptical about expanding and/or adjusting their operations despite increasing i demand. In terms of these problems and in view of the goals in achieving self-sufficiency in livestock products and at the same time reducing livestock imports, this study was undertaken in an effort to link the cattle industry to the feed-grain industry and then these two to the rest of the economy, to analyze some past unique characteristics and trends of the cattle-feed-grain economy which may be relevant in the future. Research Objectives General Research Objectives The first major objective of the study was to obtain des- criptive knowledge of the dynamics of the feed-grain-cattle economy 'in Greece in order to understand the forces behind the demand and supply schedules of outputs produced and of inputs used and to pre- scribe actions to be taken in case they are needed. The second major objective was to estimate a system of re- lationships which integrates some of the interrelationships between the consumer, the producer and the importer. Clearly, these segments are interrelated, but the main aim was to ascertain what the relevant factors are which have a serious impact within each subsystem, and not how to drive the whole system (economy) simultaneously. Specific Research Objectives 1. To investigate the beef producing industry in Greece to deter- mine: (a) what changes have taken place during the period between ' ‘1‘ .V 1 '7‘. '1 8 1951 to 1972 and what implications these changes have for the future of beef; (b) what the critical relations and/or links in the beef production and consumption process are which need to be understood by both farmers and policy-decision makers, if any action to to be taken in the future to correct the gap between production and consumption of beef in Greece. 2. To investigate the veal producing industry in Greece in order to see: (a) what changes in the composition of the national herd have taken place over the sample period and what impacts these changes have had on the consumption and production of veal over time; (b) what factors affect the consumption and production of veal in Greece and what policy implications these factors have for tackling the deficit in veal production. 3. To investigate the milk producing industry in Greece to deter- mine: (a) what factors influence production and consumption of milk and what variables are relevant from a policy-design point of view; (b) what structural and locational changes have taken place in milk industry. 4. To investigate the mechanism which links feed-grain production with the cattle economy by first describing an economic model to represent the endogenous mechanism'and then fitting the model using econometric estimation methods to see what factors influence supply and/or demand for feed-grains in Greece. 5. To investigate the impacts of a growing economy and particularly of increasing incomes on the consumption of beef, veal and milk and, further, to see what factors determine the imports of these livestock products and the inputs (feed-grains) needed to produce these products. L (1“ I 0.. - fl I 0-. l’l Data Collection Data were collected from both primary and secondary sources: primary data, from cattle feeders, dairy enterprises and feed-grain growers, and secondary data, from the Ministry of Agriculture, the Agricultural Bank of Greece (A.B.G.), the Central Market Bureaus in , Athens and Thessaloniki and publications both in Greek and in English, such as OECD publications, EEC publications, FAO publications and others. I This study is based mainly upon secondary data, which were collected from several sources as the aforementioned and compared for consistency. Where differences arose, efforts were made to determine reasons for these inconsistencies. In some cases judgments were necessary and these were made on the basis of reasonableness and the author's experience as both an agronomist and an economist. Limitation Of the Study The purpose of the study was to make an inquiry into the pro- duction and the consumption of beef, veal, milk, feed-grain and roughages on one hand, and into the imports of beef, veal and feed- grain on the other. I Based on time-series data, this study was limited by the data available and their accuracy.) Greater availability of the data would have improved the validity of the inferences drawn. Significance of the Study; It is expected that the results of the study will be widely used both in Greece and abroad, though the Agricultural Bank of on: 0". 4“ .u5 \F.‘ ;.. 10 Greece and Greek Ministries of Commerce and Cooperation will be among those government agencies with primary interest in the results of this thesis. Furthermore, the primary people in the business involved in the feed-grain cattle economy of Greece want to use some of the results of this research effort as will farmers and their Cooperative Associations. Outline of the Study After this introductory chapter, Chapter II is devoted to a dis- cussion of the Greek agricultural sector and the major problems faced during the study period and today. Chapter 111 provides a description of beef, veal, milk and feedgrain industries to provide the reader the necessary familiarity with the feed- grain cattle economy of Greece. Chapter IV discusses the methodological approach used to derive the final econometric models estimated in this thesis. In addition, the theoretical economic model underlying the feedgrain-cattle-economy is discussed in detail. Chapter V discusses the development of demand relationships and the factors which influence demand for both feedgrain and livestock products. Furthermore, the empirical results for demand analysis are given in this chapter. . Chapter VI discusses the supply relationships underlined in the feedgrain-cattle-economy and the empirical results of the supply analysis are given, along with a discussion about the use of polynomial long models. ' Chapter VII provides a discussion about trends and projections and the projection capability of the model is examined using Theil's 11 inequality coefficient to check each equation's prediction efficiency. Finally Chapter VIII gives a sumary, conclusions and reconmenda- tions for future research, along with some policy implication drawn from the empirical analysis. '-. .- (‘- ..' ‘t. .1 .A r‘. i ‘ \ i CHAPTER II THE AGRICULTURAL SECTOR IN THE WHOLE ECONOMY The purpose of this chapter‘was: (a) to expose the unfamiliar reader to the features of the Greek agricultural sector and its place within the whole economy, and (b) to provide descriptive knowledge of the industries which comprise the feed-grain cattle economy. Tables referring to this chapter can be found in Appendix A of this work. General Views Writing in 1953, Professor G. Koutsoumaris stressed that there are "two main shortcomings in Greek agriculture; a low labor productivity, and allocative and productive inefficiency in respect to capital (including land). It is on these areas that agricultural development policy should focus."1 Writing at the same time, the 2 in a remarkable study about the farm late Professor C. Evelpidis crisis in the Greek agricultural sector pointed out that the crisis is a permanent one caused by structural problems, the subsistence 1G. Koutsoumaris, "Resource Productivity and Development Policy for Greek Agricultural - An illustrative Study," Journal of Farm Economics, 36 (1954). 2(I Chronia Georgiki Krisis is tin Ellada (Athens: Papazisis, 1953), pp. 3, 18, 42, 43.). "The Permanent Farm Crisis in Greece " (in Greek) (Athens: Papazisis, 1953). 12 l“ t I .Ch. u... p’; U o ’9. V.- 1“. .m- .h. u‘.‘ v~' fa! I.'J A:' '- l 1“ '. an 13 orientation of the agricultural sector, excessive labor employed in agriculture, lack of any modern marketing system and lack of any international trade orientation which explains the inability of Greek agriculture to cope with international developments in that sector. In 1965 Professor S. Triantis maintained that a striking characteristic of Greece's agriculture is that of relatively low resource mobility and hence less economical allocation of productive resources than in more advanced countries.3 And Professor P. Yotopoulos, after studying one of the poorest agricultural regions in * Greece, Epirus, concluded that Greek agriculture is efficient but poor."4 In light of these very general remarks a more definite picture of the Greek agricultural sector will be given in the sub- sequent sections. Excess Labor Devoted to Agriculture The table below indicates the agricultural population and the economically active population engaged in agriculture in Greece. It can be observed that the population employed by the Greek agricultural sector was rather high, 40.7 percent in 1971 and 35.0 in 1975. 3Common Market and Economic Development, Center of Planning and Economic Research Center (Athens, 1965). 4P. Yotopoulos Allocative Efficiency in Economic Development Center of Planning and Economic Research (Athens, 1976), pp. 217-225. *"Efficient in the sense that marginal productivities of the factors employed in agriculture do not differ significantly from their oppor- tunity cost" (Ibid, p. 11). 14 TABLE 1 ACTIVE AGRICULTURAL POPULATION IN GREECE 1951 1961 1964 1967 1970 1971 1975 Active employed civilian Total Population 3,278 3,640 3,555 3,469 3,384 3,327 (in thousands) Of which jfl_agricu1ture 1,864 1,960 1,770 1,600 1,447 1,355 (in thousands) As a percentage (%) 56.9 53.8 49.8 46.1 42.8 40.7 35.0* Sources: (2) KEPE, Athens (1) OECD, Manpower Statistics *(3) Ministry of Agriculture Adopted from: OECD, Agricultural Policy Reports, Agricultural Policy jn_Greece (Paris, 1973), p. 20. Such excess labor results in a labor productivity which compared with other sectors in terms of Gross Domestic Product (GDP) is rather low. This is given in Table 2. For an optimum allocation of resources and under the assumption of perfect knowledge and perfect competition, the excess labor needs to be combined with appropriate amounts of land and capital in order to give the maximum product possible. Since these latter factors are not available, the economic theory easily explains the low labor pro- ductivity in the agricultural sector compared with that in other sectors or in other countries. little land and capital and the result is lower productivity of the Too much labor is combined with too more abundant factor, i.e., labor. . . IO. R '. Q.- L. “I- u ‘n he a; ‘ A t 1' ‘ 1 Vi I 15 TABLE 2 LABOR PRODUCTIVITY. GROSS PRODUCT AT 1963 PRICES. Million Drachmae* 1963 1966 ' 1969 . 1970 Whole economy Gross Domestic Product 120,402 150,661 183,164 197,345 (GDP) Active population 3,583 3,497 3,412 3,384 GDP per active person 33,604 43,083 53,682 58,317 Agriculture Gross Agricultural Product (GAP) 31,472 34,395 35,298 38,346 Active population 1,831 1,655 1,497 1,447 GAP per active person 17,188 20,782 23,579 26,500 Other sectors GDP 88,930 116,266 147,866 158,999 Active population 1,752 1,842 1,915 1,937 GDP per active person 50,759 63,119 77,215 82,085 Sources: (1) OECD, National Accounts (2) KEPE, Athens Adopted from: OECD, Agricultural PolicyReports, Agricultural Policy jg_Greece (Paris, 1973), p. 22. *30 drachmae = $1.0 in 1973 and floating thereafter. Agricultural Land Greece's land area is about 131 thousand square kilometers, or just over 50 thousand square miles, though only a little over one- quarter of this is cultivated. About half the total cultivated area ’5" UV .- I” I it) 1- w 16 consists of farms of less than five hectares while farms of over twenty hectares accounted for only 9.3 percent in 1970. Eighty-eight percent of holdings are farmed directly by their owners, and tenant farming accounts for a very small percentage. Table 3 below gives the distribution and number and area of agricultural holdings by size in 1950, 1960 and 1970. Economic pressure in agriculture and non- existing opportunities for employment in the other sectors of the Greek economy have resulted in no profound changes of the farm structure, i.e., in the number of farmers employed in agriculture and the size of farms. TABLE 3 DISTRIBUTION AND NUMBER AND AREA OF AGRICULTURAL HOLDINGS BY SIZE: 1950, 1960, 1970. Size of Holdings Percentage of Total Percentage of Total (1 hectare = 2.471 No. of Holdings Cultivated Area acres) 1950 1960 1970 1950 1960 1970 Up to 1 hectare 28.0 23.0 21.8 6.0 3.6 3.1 l - 4.9 hectares 57.0 57.8 57.3 43.0 45.1 41.7 5 - 9.9 hectares 11.0 15.1 15.8 22.0 31.1 30.5 10 - 19.9 hectares 3.0 3.4 4.1 10.0 13.6 15.4 20 and over hectares 1.0 .7 1.0 19.0 6.6 ‘ 9.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 Sources: (1) FAO, World Agricultural Structure. StudyNg, 1, "General Introduction: Number and Size of Holdings," (Rome, 1961), pp. 59, 66. (2) For 1960 and 1970: Statistical Yearbook gf_Greece. National Statistical Service of Greece, Annual Series. ‘ T. I. ..d v afl. . d U I”. TU‘ I." 4 oV‘ »\v ..‘4 O I 17 Smallness of farms is an obstacle to the use of modern machinery and techniques in general, and the land consolidation program which exists today in Greece cannot produce any really viable holdings. On the other hand, the Greek laws and traditions of in- heritance law and dowry result in excessive subdivision and fragmenta- tion of agricultural holdings, and this in turn means misallocation of time and other resources as well.5 Due to this fragmentation problem Greece is pursuing a number of policy alternatives today to increase economically viable holdings. Such policies include the establishment of "group farming"6’7 sub- sidized officially by government agencies and the establishment of a "Soil Bank"8 to take care of the land which is left behind mainly 9 by emigrants, the introduction of Societes Anonymes in agri- 10 culture and the enhancement of the cooperative movement. Other 5K. Thompson, Farm Fragmentation in Greece. The Problem and Its Settin . (Athens: Center of Planning and Economic Research, 1963), p. 29. 6S. Mariadis, and V.C. Kalaitzis, "Collective Farming.t:A Solution to the Structural Problem of Small Farm Acreage in Greece" (in Greek) (Geoponika, July-August, 1976). 7V.C. Kalaitzis, "Observations on the New Institution of Group Farming Entegprises," Hellenic Agricultural Economic Review, 9, No. 2 (July, 973 . 8A. Pepelasis, "To Provlima Tou Mikrou Klirou," "The Problem of Small Farm Holdings" (in Greek). Agrotiki Trapeza, Tephchi 10-12, 1976 and by the same author: "Agrotiki Politiki Kai Anaptixi" (Athens: Papazisis, 1976), pp. 146-161. ' 9x. Zolotas, “I Ellas Kai i 5.0.x." Oikonomikos Tachydromos, 1976. IOKalaitzis, 19. O'.; '5 pans-n ~\ 5.. I!) n ‘O U. 18 measures taken by the state to overcome this structural problem in agriculture include investment in irrigation projects: 18.4 percent of cultivated land is irrigated which makes irrigation one of the most essential elements in Greek agriculture. The irrigation pro- jects are financed mainly by the government which provides long-term interest-free loans and which, since the end of 1972, has assumed the responsibility for all the costs.]] Farm Size in Greece Table 4 below reveals that the average size of a farm in Greece was 7.67 acres in 1961 and 8.23 acres in 1971. Thus the size prob- lem today remains actually the same as it was 10 or 20 years ago. This smallness is further deteriorated by a fragmentation problem. Each farm in Greece is, on the average, divided into 7 separate plots and that makes farm business operations even more inefficient. Capital in Greek Agriculture Greek agriculture is characterized as a labor intensive sector12 and, as such, it might feel relatively little need for mechanization. But this does not mean that capital is not required for investment. The gross domestic asset formation was 3,415 million nOECD, A ricultural Policy B_ports, Agricultural Policy jg Greece. (Paris, 973), p. 21. - 12S. Triantis, "Common Market and Economic Development. The EEC and Greece," (Athens: Center of Planning and Economic Research, 1965), p. 48. 19 TABLE 4 NUMBER OF AGRICULTURAL HOLDINGS, FARM SIZE AND IRRIGATED LAND, GREECE 1961,1971. 1961 1971 Number of Holdings 1,140,1631 1,036,600] Total Areas ('000 Stremmata) 36,733 35,863 Total Areas ('000 hectares 3,673 3,586 Average Size (Stremmata) 32.21 34.59 Average Size (hectares) 3.2 0.8 Irrigated Land ('000 Stremmata) 4,890 7,337 Irrigated Land ('000 hectares 0.484 0.734 Source: National Statistical Service of Greece, Statistical Yearbook of Greece (Athens, 1973). 1Excluding 16,009 for 1961 and 10,660 for 1971 holdings with animals only. dollars in 1960 (in constant 1958 prices) and it increased to 6,320 million dollars(in constant 1958 prices)in 1972 which is almost a 100 percent increase over that of 1960. However, if we compare this with other sectors' asset formation, for example, dwellings, we see that the latter's increase was almostthreefold:13 it increased from 5,646 million dollars in 1960 to 15,606 million dollars in 1970 in con- stant 1958 prices. Greece's climate is dry and precipitation is irregular and varies considerably from year to year. This requires the government to spend a considerable amount of money on irrigation which in 1970 13National Accounts of Greece, 1948-1970, No. 21. (Athens, 1972). 20 accounted for about 80 percent of investment in land improvement. Several types of subsidies (input subsidies, price intervention schemes and compensation payments) provide other ways for government to invest money in the agricultural sector, while training and re- search is still another field in which money is invested by the govern- ment with the aim of improving farmers' productive capacity and to increase knowledge about the farm sector's operation. Finally, only in the last few years has the government supported co- operatives in assuming responsibility for commercial and industrial operations previously in the hands of the Agricultural Bank of Greece (A.B.G.) or the State. Despite such investment efforts by government, however, the capital invested in the sector is not sufficient. Agriculture and Foreign Trade While Greece's overall trade balance is heavily in deficit (it reached a record level of more than $800 million in 1969), trade in agricultural products shows a favorable balance. The agricultural trade balance, including raw cotton, moved from a surplus of $68 million in 1955 to one of $170 million in 1967. Yet the favorable balance of trade in agriculture is less than the balance on each of the two main invisible items (emigrants' remittances and maritime transport), but higher than that of tourism. Idhp Ocrfi‘ it vi ':.:.r Dy.» if” uhug Yr" .M .I 2,-- "we (1 .11 AA‘A .21 The Geographical Pattern of Agricultural Trade Imports are coming into Greece notably form three major re- gions: North America, EEC and, to some extent, from EFTA. Approx- imately 1/6 of all Greek agricultural imports consists mainly of cereals used as feeding stuff which come from North America. Beef and veal imports come mainly from EEC countries, although Yugoslavia has been one of the main suppliers of live animals and, to some extent, meat. Eastern Europe also plays a part in agricultural imports, but its exports to Greece are rather irregular and are carried on through bilateral trade agreements. 0n the other side, Greek agricultural exports go mainly to EEC, which at present absorbs roughly one-half of total Greek exports. North America and the EFTA countries absorb much smaller quantities, usually even smaller than that of Eastern Europe. This heavy EEC export orientation of Greek agriculture has been criticized quite a lot by many in Greece, since it is believed that by this heavy reliance on the EEC market Greece loses her bargaining power. These critics support instead the idea that a wider range of export markets is more profitable for Greece since this naturally could even out risks. With violent fluctuations in the prices of export goods and the strong competition by other countries this heavy reliance on the EEC market seems by itself very considerable. Myrdal writing in 1956 pointed out that 22 A country like Greece, trying to earn half its export proceeds by finding markets for its tobacco, is con- tinually forced to accept a number of concessions re- garding its imports, which it would not accept if it had a freerer position since they run counter to its development policy. Often it is compelled to open its boundaries to the import of a number of consump- tion goods while there is idle capacity at home to produce them.‘5 15Gunnar Myrdal, An International Economy. Problems and Prospects (New York: Harper and Row, 1956), p. 256. one Ill ‘21 o ‘n. " s D;'.. 5".” ’1 51 CHAPTER III THE INDUSTRIES INVOLVED 1. The Cattle Industry The objectives of this section are: (a) to present some of the unique features of the cattle economy and (b) to provide the necessary descriptive knowledge which may be used for the quantitative analysis which follows in Chapter V. Size and Composition of the Cattle Industry Table 5 reveals the size of the national herd and its components for the years 1961 to 1972. The national herd consists of: (a) the domestic herd (local unimproved), (b) the local improved herd, i.e., cattle resulting from cross-breedings with domestic cattle and dairy- beef and, finally, (c) the so-called foreign improved herd, i.e., exotic or pure breeds used mainly for milk production. Regional Patterns of the Cattle Industry Most cattle were (and still are) maintained in the plain areas over the sample period and these numbers have increased since 1961. The numbers in Table 6 illustrate that the number of cattle fed in the last two categories of land is decreasing, and cattle population in the level areas is increasing. This means that pasture areas which are utilized by beef and dual purpose breeds are underutilized, While 23 i -§ om>ommzH onmmou om>ommzH 4omazmza 4osasH cmpmcom cm>oeasH Peuog um>osaspcz ”coo; Page» opeemzm mzou .mcmewm; .mppzn .cmxo cowmmm upsnecmomw zo Ammee 43m .comp .Fm mmm2mumo m m4m4). But there was a decrease in the area under oats which dropped from 138,000 hectares in 1954 to 80,000 hectares in 1970. Its production also declined from 150,000 tons in 1954 to 106,000 tons in 1970, while average yield increased from 1083 kg/h in 1954 to 1,325 kg/h in 1970 (see Table c-l)_ Finally, the area under maize decreased from 253,000 hectares in 1954 to 170,000 in 1970, although production almost doubled, in- creasing from 254,000 tons in 1954 to 481,000 tons in 1970. The average yield almost trippled, increasing from 1,005 kg/h in 1954 to 2,829 kg/h in 1970 (see Table C-l), The table in Appendix C thus reveals that a considerable increase in productivity has occurred in wheat, barley and maize production. This is partly due to better varieties used over the time, to better management techniques used, to increased mechanization al- most in every phase of production and to use of better pesticides. 40 It seems that a limiting factor exists, however in that Greece now has almost reached the limit of geographic expansion of feed-grains land. Further expansion on feed-grain acreage will de- pend largely on the major irrigation projects carried out in the agricultural sector of Greece, since water supply is the most impor- tant limiting factor. Another factor which is revealed is that Greece achieved a certain degree Of diversification of farming during the sample period. The acreage once devoted to other principal crops is now under in- dustrial and feed crops. Thus, the rate of growth over the period 1947-49 to 1965-67 was as followszlo Grains 5.0 percent Feed Crops 9.4 percent Industrial Crops 7.0 percent Tree Crops 5.5 percent Grains experienced a rate Of growth equal to 3.7 percent in Thrace in the 1952/54 to 1961/63 period, the highest in the country. Macedonia had a rate Of 3.6percent. Feed crops and hay experienced rates Of growth as high as 24.8 and 28.4 percent, respectively, in the Aegean Islands over the same period, while legumes had their highest 11 rate of growth, 22.0, in Epirus over the same period. In 1966, the largest area under feed-grain was the region of Macedonia. 10Shaw . pp. 54-60, Table 2.6. ”Shaw , p. 62, Table 2.7. 41 The structure of the farms which grow feed-grains remains typical of the Greek farm in general which has an average of 8.3 hectares of arable land and is largely diversified. Large feed-grain operations exist only in the region of Thessaly. Nevertheless, im- portant changes in farm structure are occurring in Greece (though slowly) or can be expected in the years to come. The farm labor move- ment to the cities or abroad and increased labor costs have resulted in a more labor extensive agriculture, with the relative importance Of dairying and, hence, of grassland farming diminished, while that Of crops, especially cereals, has increased. CHAPTER IV METHODOLOGICAL APPROACH Introduction The general and specific objectives of this study have been stated earlier. In a broad sense, time series data were used for demand and supply analysis in a primarily single equation approach. Numbers of livestock and yields of livestock were utilized as separ- ate dependent variables affected by farm prices. It was assumed that most of the supply response was related to the conversion Of some farm products (inputs) into livestock products (outputs) and that a substitution took place with other farm products. The method of estimating each equation was the ordinary least-squares method. The equations given in the models were chosen by considering a priori knowledge of the industries, by using rele- vant economic theory and, finally, by taking into account the reliable data that exist in Greece today. Models which utilize time lag techni- ques were also tried. 42 43 Diagrammatic Presentation of Feed-Cattle Economyzlnputs and Outputs An effort was made to include important features of the economic problems faced by both the producers Of livestock products such as veal. beef and milk and by the producers of feed-grains and roughage. The model depicts the imports economy for livestock and feed-grains products in order to examine the relationships between the domestically produced products and the imported ones. It helps the reader to visualize, in a simple diagrammatic presentation, the association of inputs used to pro- duce the outputs in the feed-cattle economy (see Figure 2). Figure 2 attempts to give a visual picture Of the major inter- actions involved in the greek feed-grain-livestock economy. Four major distinct interactions can be distinguished: a) demand for and supply of inputs which go to the production Of feed supplies. These inputs are also demanded by the entrepreneurs who produce other outputs in other enterprise combinations; b) supply and demand interaction for each commodity produced, i.e., feedstuffs. Feedstuffs are used as inputs to produce livestock products; c) livestock products which are directly or indirectly (processed) consumed by humans and/or animals (milk, for example), thus giving another set of interéction forces of supply and demand; 44 .Axpwz .Fem> .emmmv cowuoauoea muuznoea xuoumm>w4 c? mnwcmzowpepmm psau301p=acH .N meamwm _ 1 Fem> use mmmm we mpeanH mpeewc< m>w4 we mueoasH me> mzou xsweo mcweew been eo mpeoasH \ Acpo meema NV mm>peu memegmaom came, Ac muzchv. sm>o new menu» N .mzoo new .mcmmmm: .meum xpaaam emcee ummm Au mpzncmv Acowuuczm copuusuoeav u mpaauao meuamu Loaeg can; Feeeeee Lone; eeeu . < mpaacm m mesapeo 45 d) The existence of a competitive process for the foreign feed- grains imported to Greece since other importing countries compete with Greek (domestic) livestock producers for access over the same feed grains. The recursive linkage among these major interactions is more easily understood to take place in the livestock feed-stuffs demand and livestock products price (supply) formation procedures. Due to the dual character Of the cattle herd and due to the very diversified type of farming prevailing in Greece, multiple enterprise or resource allo- cation competition is assumed to exist in the production side of the feed-grains-cattle economy. Land, buildings and tools are still considered as fixed factors in a specific livestock enterprise in Greece and cannot be shifted rela- tively easily from one enterprise to another. This observation (assump- tion) leads to specification Of a recurSive type of model. Whenever a time lag seemed to be the major factor explaining a production phenomenon, distributed lag techniques were employed to Obtain "better" results. 3 Figure 2 shows that outputs A are produced from inputs A. In other words, feed-grains are produced when inputs such as land, labor, capital (in the form Of fertilizer, machinery, seed, etc.) are used. These outputs A can then be either treated as such or used as new inputs (inputs 8) to produce new products (outputs B). Outputs B, which include the live animals of calves, steer, heifers and cows can again, either be treated as such and be sold as live animals, or be used as inputs (inputs C) to produce outputs C.which comprise the live- stock products being studied here, i.e., beef, veal and/or milk. 46 Figure 2 also gives the path production of feed-grain and roughage and of beef, veal and/or milk. In this diagram imports Of feed-grain, live animals and (final products of) beef and veal are provided in order to illustrate the second Open alternative (path Of production) which can be fOllowed by the Greek government. This second route can be a complementary path Of securing livestock products for Greek consumers as well and is considered as such in this thesis. From Figure 2 it is clear that domestic (Greek) consumer demand can be met either by following the first route i.e., by pro- ducing beef and veal domestically, or by importing them or by a com- bination of the two alternatives. The inputs used in the production process of beef, veal and milk can be provided to the cattle industry by choosing among the same three routes . Thus, Figure 2 specifies the policy options Open to the domestic producers, to domestic consumers and/or to the government. This topic will be elaborated further later on. y Figure 2 shows, too, that feed-grain and (to a lesser extent) roughage compete for production resources with other crops. Land utilized for feed-grain is also used for grain which goes for human consumption. The same holds true for labor and capital; Thus, since in this case land and capital are the most rare resources in Greece it can safely be said that cattle production competes with human popu- lation for grain. In addition, other crops use the production re- sources which are devoted to grain (and hence, livestock products) production. It helps,then, to think of other crops which should enter into the model structure as competitive or complementary products. 47 Further Diagrammatic Develgpment of the Model For a more explicit diagrammatic presentation of the feed- cattle economy of Greece the diagram in Figure 3 has been drawn. Here, a more detailed picture of the feed-cattle economy is given. More specifically, the new element which has been added in Figure 3 is the government which has its influence on the aforementioned economy through trade regulations, institutional framework and policy designing. This diagrammatic presentation provides thought-stimulating insights and helps one to think of the whole system Of the economy considered here. Thus, beginning at the base of the diagram retail demand and supply for inputs and outputs can be formulated. Then, wholesale demand and supply can be taken as a new subsystem. Pro- ceeding upwards, the farm supply of livestock products can be distin- guished and demand for inputs can be recognized easily. Furthermore, the foreign trade element (imports) of both inputs and products is clearly formulated. The rest of the world demand stands at the very top of the diagram, while the government stands at the very bottom. Needless to say, in this last component of the feed-cattle economy variables such as subsidies, advisory work, resources and development (technology), etc. are included. Thus, this segment plays an important role and seriously affects behavior of the parts involved in the system at hand. In terms Of this paper's organization and estimation procedure, the wholesale supply and demand component was reduced for reasons such as lack of data concerning the structure and behavior of the wholesale 48 % REST OF .THE WORLD DEMAND E .f . -. .. . , . 'T* _ 1 - éi; REST OF THE WORLD SUPPLY FOREIGN SUPPLY FOREIGN SUPPLY FOREIGN SUPPLY - OF OF LIVESTOCK OF FEED-STUFFS ALIVE ANIMALS PRODUCTS 5 INTERNATIONAL TRADE \‘~.\_ (IMPORTS IMPORTS DOMESTIC SUPPLY DOMESTIC SUPPLY OF OF FEED-STUFFS LIVESTOCK PRODUCTS IMPORTS FARM SUPPLY DEMAND FOR 1 1 FEED-STUFFS WHOLESALE SUPPLY AND DEMAND CONSUMERS RETAIL DEMAND AND SUPPLY 1 . 1 ' GOVERNMENT, Institutional Framework Policy-Making Trade Regulations, etc. Figure 3. The Feed-Cattle Economy in Greece and Its Domestic and ' Foreign Environment. 49 industries involved in the feed-grain-cattle economy of Greece. This by no means implies, however, that this component of the system is not important. 0n the contrary, the efficiency of this component and its performance influence both supply and demand for the products being studied here. Relationships between marketing margins and retail and farm prices are explained in a later section of the next chapter. An even more detailed graphic presentation of the relationships which exist in the whole feed-grain economy is prOVided in figure 4. This diagram depicts the endogenous variables of the system as circles (and the exogenous ones as rectangles. Yet this is a kind of tenta- tive division Of these variables in these two categories since, some- times, there is no clear demarcation line to what variable is endog- enous and what variable is exogenous to the system. Beef consumption and production are linked with both current and lagged prices, and any difference between these two schedules may be presented in the form of an identity equation. Here beef produc- tion is defined as the total meat which is ready for consumption at time t, regardless of its source of origin (domestic, imported as frozen or as live animals which are slaughtered within the country etc.). Heavy arrows depict stronger relationships, most Of which have been estimated in the empirical analysis Of this thesis. The setup of this diagram in Figure 4 helps one to visualize both the individual variables used in each equation and to list all the equations which fOrmulate the sybsystems and the whole system in general. maozoom mapumouemom xmchc mg» no musuosuum mafium cam hammzm.ccmson .: whsmwh EMU. ..._.. SO— 0>Hc o vmsTlg coauosnoem HMpcwst_ I. 121. Ncwm 3” .5869 IWrIHrIHUE *1 _ F 1.1 I 1 I U 0 .c 6.6ch . R. o moahmf Boo .ma..o .111... / an 33.... . Mmdpmmau LcAe>< Feluwmme anemflealfi I . _ .Lo mac) I1. , rinmml IKE ,0Huozcoem .wcou .mmm>vhom a 9...... were: 556 .eucemeam amends. EMA HA No wwoflpm mo wuhomrw Heceee Hc>cé cfidflflH fl wowhm HCHCCCU PECCWLQA .rflmpo ecmhv. o woocchemmu_ Hp ;om .m _ can mmu:.£ I 1.14 mamm:.uewu Hmmzm cNHOCOh 51 Livestock Supply and Derived Input Demand Supply response functions for the commodities produced in the feed-cattle-subsector are based on micro-level assumptions con- cerning the technical nature Of feed grain and livestock production and producer's objective functions. Equations (4.1) and (4.2) are the general production function and Objective function assumed for livestock activities, while equations (4.3) and (4.4) are the gen- eral produCtion and objective function assumed for feed grain activities; equations (4.5) and (4.6) are the production function and the objective function for roughage (pasture) activities. Finally (4.7) and (4.8) are for the trade activities. QLPP = f (QFG, QR. QL. 0K) 1 (4.1) RLP = PLPP QLPP ' PFG ' QFG ‘ PR ° QR ‘ PL ' QL ' PK ° QK where: QLPP = Quantity of livestock products produced f = denotes function QFG = Quantity of feed grain used QR = Quantity Of roughage used QL = Quantity of labor used Quantity of capital used :0 O K I 11 LP - Objective function in livestock products production, i.e., Gross Margin = Total Revenue - Total Variable Cost v 11 Price 52 FC = Fixed capital expenses used A = Lagrangian Operator. The subscripts are defined as follows: LPP = Livestock products produced FG = Feed grains index R = Roughage index L = Labor used in the production process K = Capital (in general) used in the production process For the feed-grain production activities the equations are: QFG f (QF. 0L. OK) (4.3) R = P FG FG QFG ' PF ' QF ' PL ' QL ‘ Pk ‘ Qk ' FC 1 + A2[QFG ' f (QF’ 0L9 QK)] (4'4) For the roughage activities the equations are: OR = f (0F. 0L. OK) (4.5) RR = PR 'QR" PF ' QF ‘ PL ° QL ' PK ' QK ‘ FC + ABIQR - f (QF: QLS QK)] (4'6) where: QF: Quantity of fertilizer used in the production of feed grain and roughage Finally for the trade activities the equations are: QLIMP = f (QK’ QL) (4.7) 53 RIMP = PLIMP ° QLIMP ' PK ' QK ' PL ' QL ' FC + where: LIMP = Imports of livestock products (beef and veal) imported. Livestock production in Greece is assumed to be a function of the quantity of feed grain, the quantity of roughage, the quantity of labor used, the quantity of capital used in a general form, mainly the variable capital. In addition, livestock production is a func- tion of the fixed assets existing in the specific livestock category. The fixed assets invested in the livestock industry in Greece are not that specialized, on the average, over the time period examined. But when fixed is defined to include feeder animals, feeding facilities and other specialized capital and/or machinery used to produce the livestock products being considered, then, for the Greek conditions, and compared with the total capital investment in other farm enter- prises, it can be said that fixed assest play a considerable role in these two livestock industries. For the country as a whole the magnitude of fixed assets in the veal and beef industries could by proxy be presented by the breed- ing stock (of each industry) in existence at any given point of time. Strictly speaking, due to the aggregation of inputs used in equation (4.1) it cannot be defined as a production function but more correctly as an aggregate input-output equation. The objective function assumed for livestock producers (veal, beef, and milk producers) is given by equation (4.2). A Single enterprise objective function is assumed 54 based upon the assumption that livestock production activities re- quire in general specialized inputs not easily adoptable to other farm production activities. In this sense the livestock producer is taken to view his revenue as being realized from a single source and does not consider other enterprises. Thus, the existence of fixed assets for the enterprise constrains the producer from maximizing his revenue by its effect upon the marginal productivity of variable in- puts and by the generation Of fixed costs specific to the enterprise. The feed grain production function is thought to depend upon the use of fertilizer, the amount of labor and the amount of capital in form Of combine machines and upon the weather index taken as the rainfall measured in mm of rain. Finally the import production function is taken to depend mainly upon the amount of capital needed to buy the quantity of beef and veal needed to meet the domestic demand and/or the quantity of labor in terms of real labor employed in that industry. The micro- 1evel production functions and Objective functions can easily be transformed to macro-supply response fuhctions and derived demand functions as it was shown diagramatically in the previous section. Thus, all that is needed is the transformation of the micro-level assumptions to macro-level ones. In the language of mathematics this transformation is accomplished by assuming the producer is maximizing his Objective function (R) subject to certain constraints imposed due to the fact that his budget expenses are limited. The mathematical conditions for optimizing such an objective function require that the partial derivatives of the objective functions 55 * with respect to production factors are equal to zero. The Marginal Value Product (MVPxi) Of each production factor must equal its price for the constraints to be satisfied and the profits (revenue) to be maximized. Thus, one equation for each input can be written expressing what was said in the text. aR 30 LP = . LPP _ = aQFG. PLPP aQFG PFG‘ 0 (4-9) BR 30 LP _ . LPP _ = 30E"’ PLPP aOR PR. 0 (4'10) 3R 30 LPP _ . LPP _ = L L OR 30 LPP LPP = K K 3R __LEE.= _ = 3A] QLPP f (oFG . QR . QL. OK) 0 (4.13) The system contains five functions and five unknown quantities, since prices are assumed to be given and known at the micro-level. This system of equations can be rewritten in reduced form making all endogenous variables a function of exogenous prices. When this is done the general form Of the functions will appear as follows: QLPP = f (PLPD PFG 9 PR 9 PL’ PK) (4.14) * Within the context Of static theory. QFG = f (PLP’ PFG; R PL’ PK) (4°15) OR = f (PLP. PFG. R PL, PK) (4.16) QL = f (PLP’ PFG R: PL’ PK) (4°17) QK = f (PLP’ PFG R: PL’ PK) (4°18) Equations (4.9) - (4.18) represent the supply response and derived demand functions theoretically generated by the static micro- level conditions specified. The term QK is taken to denote capital used in a general sense in the productive process. This includes both the variable capital and the fixed capital (fixed assets) and the assumptions concerning the existence of fixed assets (inputs) specialized to the enterprise lead to the appearance of a separate quantity term chosen as a proxy for these fixed assets in the macro supply and derived demand functions. Here the fixed assets are viewed as an input into the livestock and feed grains and roughage producing activities which become variable over a longer time Span and hence the same factors are assumed to generate the demand for inputs. The assumption that a fixed asset is fixed only for one production year changes only the nature of the economic theory involved in specifying the derived demand relations for the productive resources. Over the sample time period what is classified as a fixed asset and sets constraints to profit maximization of a producer becomes variable and is not a constraint to this goal achivement any more. But to understand producer's investment behavior and his decision making 57 each year, the notion of the fixed asset theory seems to be very crucial. Thus, to have a clear picture of the system the quantity of breeding stock demanded in a static sense can be specified as follows: (NBC, NVC, NMC) = f (PLPP’ PFG" PR”’ PL’ PK, PIMPBV’ NBCt_], NVCt_], NMCt_]) (4.19) where: Number of beef and veal cows and number of milk NBC, NVC, NMC cows respectively PLPP = Price of livestock products produced, i.e., price Of beef and veal and price of milk PFG = Price of feed grains PR = Price of roughage PL = Cost of labor PK = Capital cost PIMPBV = Price of imported beef and veal NBCt_], NVCt_ NMCt_1 = Number of beef, veal and milk cows 1. respectively in t-l year. The beef and veal industries are viewed in this study in relation with the feed grain economy and the import economy for the two livestock products, i.e., beef and veal. The derived demand for feed grain and roughage generated in the livestock sector are a major source of demand for these grains as this is shown by the variables which are used in the feed grain demand equation(s). Thus, the 58 interaction Of the livestock sector's (here only the beef and veal industries) derived demand for feed-grain and the supply of feed- grain form the market price generating process for feeds. Supply and Derived Input Demand for Feed-Grain and Roughage The economic theory behind the specification Of feed supply response functions is the same with the theory used in specifying livestock supply response functions. However, a different set of assumptions have to be made for the specification of the feed supply response functions and hence, different empirical specifica- tions must be derived. One to the small farm enterprise in Greece the feed-grain grower cannot achieve a farm income to live upon and he is taken here to have a multiple enterprise Objective function which is max- imized thus giving him the appropriate level Of income. His assets are not so specialized and he is not constrained that much to shift his assets from one enterprise to another. Under the Greek con- ditions quite a few farmers in the plain of Thessaly are specialized as grain producers. Most of them consider these activities as complementary since the only requirements specified by the grains are not that much labor and/or capital. Since Greece has been a net importer of grains, she subsidized them most of the time andin different ways. These conditions and these factors lead to the specification Of the following production relations and associated Objective function always within the general micro-static level. 59 0C1 = f (L, LB, K) (4.20) RC = (PCl 1 SPCl) ' QC1 1 (Pcz I spcz) ° QC2 (PL ° QL) ’ (PLB ' QLB) ' (PK ' QK) + A1 (QLC1 + QLC2 ' GI) + A2‘QFG1 "f (0L1: QLBZ’ 0K2)] .1.- A3‘QF62 ‘ f (QLZ’ QLBZ’ QKZ)] (4'21) where: QC = Quantity Of crop L = Land LB = Labor K = Capital RC = Objective function for a crop producer; the Same as in livestock production SP = Support price or subsidy price A = Lagrangian operator WI = Weather index 1 = Crop enterprise number 1 2 = Crop enterprise number 2 GI Government intervention program(s). It can be seen from (4.20) that given weather conditions feed grains and/or roughage production is a function of land, labor and capital - in all its forms - used in each crop enterprise. 60 All inputs are assumed to be Variable to the industry but not to the firm. The objective function of a typical Greek feed grain and/or roughage producer indicates that his revenue is realized (generated) either by selling his product to the market or by selling his product to the government agencies responsible for the grain storage program. The constraints of his Objective function are set from the fact that his land input devoted to the grains production is limited (given). The physical input-output relationships existing for each activity and the technology included are given as well. To determine the conditions necessary to maximize the constrained Objective function of a typical producer and in order to set up this set of functions in the form of supply response and derived demand functions one proceeds mathematically as follows: QC] = f (Pcl. SPC]. Pcz. SPCZ, PL. PLB. PK) (4.22) ch = f (PC1. SPc]. Pcz. SPcz. PL. PLB, PK) (4.23) OC1 = f (P SP P SP P P P ) (4 24) L 01’ C1’ C2’ C2’ L’ LB’ K ° ch = f (P SP P SP ' P P P ) (4 25) L C1’ C1’ 02’ c2: L’ LB’ K ° c1 _ QLB ‘ f (PCl’ SPCZ’ PCZ’ SPCZ’ PL’ PLB’ pK) (4:25) QCZ = f (P SP P SP P P P ) (4 27) LB C1’ C1’ C2’ C2’ L’ LB’ K ° C1 _ ‘ C2 QK = f (PC1’ SPCl’ Pcz’ spcz’ PL’ PLB’ PK) (4°29) 61 If the equations (4.22) - (4.29) in this section are compared with the ones in the section dealing with the empirical equations adopted to explain what was intended to be done, it will be seen that supply of crops is generally thought of as consisting of units (head of cattle, acres, etc.) times yield (pOunds, kilograms, etc.). The abstract theoretical relations presented here includes no such dis- tinction. The question is a simple one of the nature of the resource(s) used; how many resources are needed to generate that much (say Q) quantity of output. The decision to be made by farmers is to choose alternative policies which lead to the same result but with different sets Of costs. Such a decision can be: Should a farmer increase acreage or yields in order to arrive at the production of the same output Q? These two simple alternatives dictate different response to different producers and this finally depends on the marginal value product of each factor considered. Given the Greek conditions under which land is a scarce factor the dynamic response pattern may differ from that of an American farmer since resources associated with altering yield are more variable in nature than those associated with land use in the case of a Greek farmer. The opposite could be true for his American colleague. Policy constraints may be acreage or yield specific in regard to the type Of inputs or activities constrained as is the case with tobacco in Greece during recent years. In this study yields and unit 62 functions are used along with a single supply response function.1 The primary reasons for estimating yield and unit functions separately is the fact that it is believed that the dynamics of the two relations are different and different variables come into play in each relation- ship. This, however, has been done for the livestock products and for feed grains and roughage as well. 1For a theoretical discussion on this topic see: Gordon Gemmill, "The World Sugar Economy: An Econometric Analysis of Production and Policies." Unpublished Ph.D. thesis, M.S.U., 1975. 63 The Econometric Model- The General Linear Model 2 The General Linear Model can be regarded as follows: k . _= , .‘I- , O: ,2,..., 4°] YJ 1.518%” UJ J l n ( ) where: Y.: the dependent (endogenouS) variable or predictant; X.: The independent (exogenous) variable or predictor(s); 8.: Unknown parameter that depicts the influence of inde- pendent variable(s) upon Yj; U.: random disturbance term on the jth observation repre- senting the influence of the predetermined variables Xi left out of the equation or errors in the measure- ment of the dependent variable Y. The endogenous variables Xi can be fixed or random variates. It was generally assumed that these variables are of the first case (type Of endogenous variables) since such an assumption Simplifies the analysis. However, when the exogenous variables are assumed to be random variates, they can be assumed to have a probability distri- bution independent of the disturbances. It is further possible to consider errors of measurement in the independent variables, errors which complicate the analysis considerably. Thus, it was always assumed that the independent variables are measured without error. 2 The present exposition heavily draws from H. Theil's work, Principles Of Econometrics (New York: John Wiley and Son Inc., 1971). 64 Many assumtions could be made about the first and second moments of the disturbances. However, for the purpose of this analysis here it will be assumed that they have expected values of zero (normal- ity), constant variances (homoscedasticity) and zero covariances (non- autoregression). These assumptions imply that the expected value of Yj equals 1 pixij J = l,2,...,n "MK" E(Yj) = i i.e., the variance of the dependent variable is constant for all Observations (this is called homoscedasticity), and the disturbance terms are pairwise uncorrelated for all observations. The model becomes unduly complicated if any one of the XS is linearly related to any other XS in the model. Furthermore, no statistical problems are involved when the number of Observations is equal to the number Of regression coefficients in the model. Thus, the following assumptions are made which rule out these possibilities: a) no linear relationships exist among the observations on the indepen- dent variables, the X5, and b) the number of regression coefficients is less than the number Of Observations. Because matrix notation aids in conciseness and brevity, it is adopted here for this short discussion about the linear regression model employed in this research effort. Letting 65 1" H r" 1' Y1 1.x]2,...,x1k1 1' Y2 1,X22,...,X2k 2 Y = ’ x = ’ = LYnJ d’an”"’xnk_) Lke (nxl) (nxk) (kx1) 11,1 U2 and U = U L n‘ (nxl) then the general linear regression model can be written succinctly ' as follows: Y = xe + U (4-2) 66 §iggle Equations vs. Simultaneous Equationggystem A first economic system which was thought to describe the feed-cattle economy of Greece consisted of nine equations for the beef segment with nine dependent variables, of which four were in an identity form. The veal segment employed the same equations and the same number of both endogenous variables to be estimated and exogenous variables appearing in an equal number of identities. The milk segment employed eight equations, of which three were in the form of an identity. Imports of fluid cow milk are not in- cluded in this model since there are almost none. 7 The feed-grains segment also included nine equations, Of which four were in the form of identities, while the pastures (roughage) segment of the model included eight equations, of which four were in the form of identities. Thus, 19 endogenous variables remained to be estimated with 22 degrees Of freedom. The overall system was overidentified since in a complete model the number of structural equations must be equal to the number of endogenous variables. And it was expected that this would be the case here since, according to economic theory, to manage to construct an exactly-identified system for such a large part of the 67 agricultural sector would necessitate the insertion Of variables which contribute very little or at all to the explanation Of the vari- ation Of each independent variable. Apparently, these variables are not backed by the economic theory itself. ' The structural form of the equations which was set up to model the feed-cattle economy of the Greek agricultural sector called for both "recursive" and fsimultaneous" relations to be specified and estimated. A recursive relation is defined as a structural function containing at least one lagged endogenous variable (and thus exogenous to the period in question) as an independent variable but no current period endogenous variables as independent variables. On the contrary, a simultaneous relation is defined as a function containing at least one current period endogenous variable as an independent variable. This study did not enter into the historical debate about the nature Of the "economic world." In other words, it did not address itself to the question:.are economic phenomena taking place inauiinterdependent way (in a Haavelmosian approach) or in a recursive (Causal-chain Woldian approach) way? There are excellent references on this issue and the reader should consult them if interested.3 ’ The statistical estimation of the parameters of simultaneous equations is confined to identifiable equations, but this is not the serious problem it may seem since structural equations are usually 3Karl Brunner (ed), Problems and Issues in Current Econometric Practice (Columbus: The Ohio State University, College of Admin- istrative Science, 1972) 68 overidentified; and the reason for this is that in a system like the one under examination the number of predetermined variables tends to increase with the number of equations, whereas the number Of the endogenous unknown parameters to be estimated in any particular equation is rather small. Two-stage least squares technique produces an instrumental variable for independent endogenous variables in the function. And, ideally speaking, an instrumental variable is nearly independent of the function's error term but highly correlated with the independent variable for which it is an instrument. TO the extent that the instrumental variable achieves its desired properties, the bias of the estimate of the parameter for the endogenous independent variable is reduced. But this reduction of bias occurs at the expense of some efficiency. The instrumental variable never perfectly matches the variation of the endogenous independent variable it represents; and, due to that reason, it is Obvious that some infOrmation is lost. In problem solving research the decision whether to use simultaneous equations estimators, such as the ones coming from the use Of ZSLS, must take into account whether the elim- ination of bias achieved is worth the efficiency lost. In this study, following Karl Fox's approach, single equations were tried, since it was believed that an equation which expresses the price of the commodity (or, what is the same for that matter, the quantity) as a linear function of its supply (price) and consumer income would contain only one endogenous variable as a function Of the other two predetermined variables. This is also a true "strUctural" demand equation, and, if it is assumed that the disturbances (residuals) 69 from this equation are random and normally distributed, then this equation fitted by the method Of least-squares with price as the dependent variable is identical to the Maximum Likelihood Estimate (MLE) indicated by the commission approach. Thus, the single equation approach was fully justified in this case (Koopmans, 1945).4 Model Design and Identification The econometric modeling approach aims at specifying the model according to the existing body of knowledge of economic theory at the time Of model specification. The estimation of the model's parameters came from a rather general knowledge of the population from which the study sample was drawn and, hence, was based upon statistical inference techniques which used samples of data describing the population for which geheralizations are made. This leads to the econometric modeling approach which depends on time series data and various forms of regres- sion analysis. It is of course implicitly assumed in all these kinds of studies that relationships and parameters estimated from these time series data can adequately be described by the relationships modeled. The task of a researcher who uses time-series data like those used in this study is to know the structure and behavior Of the sector or sub-sector that is under investigation and thus be able to formulate a model which more O, the product is called superior or normal. e) If Ey < O, the product is considered to be an inferior good. f) If E > 0, the goods compared are substitutes. g) If E < O, the products compared are complements. Under special, but not unusual circumstances, the three elasticities (own-price elasticity, income elasticity and cross-pro- duct elasticity) will sum up to zero for a given product. In other This suggests that products which have many or close substitutes also have more price elastic demands, and those products that are highly superior when income increases, even if they have few substitutes, may be price elastic. 79 Sometimes the concept of total price elasticity is used. Whereas own-price elasticity is a measurement of changes in quantity demanded when price changes, other things remaining unchanged (ceteris paribus), total price elasticity is a measure of quantity change in relation to price change, allowing other factors to change (mutatis mutandis). The total elasticity for a given product will be less elastic than the more standard ceteris paribus elasticity. Demand Estimation With the previous chapters and the static theory of demand ex- plained in the theory of demand section as background, one first big step has been completed toward the demand estimation. As stated by Waugh: The first step in any statistical analysis should be to set up some sort of theoretical model describing how the markets for a commodity work. The model generally starts with a listing of factors that are believed to affect the supply, demand, and price of the commodity. Diagrams are Often eelpful in por- traying various interrelationships. The model may be complex, with several equations representing a number of supply-demand equations and constituting what is called a simultaneous equations system, or it may be rather simple, with one equationrepresenting demand. The second approach was followed in this research effort while the first approach remained for use in a further investigation of the feed-cattle economy. The model finally ...should be put into a form that can be fitted by statistical techniques to determine if it is 1Frederik Waugh, Demand and Price Analysis, U.S.D.A., Technical Bulletin 1316, NOV.71964, pp. 6-7} 80 consistent with the observed data. To set up a good model for measuring the demand for any commodity, the researcher must have an intimate understanding of the markets for that particular commodity. The routine fitting Of the same model to cotton, beef cattle, and canned peas is poor research method.2 Once the variables that determine demand for a given product and the mathematical form of the relationship among the variables have been selected, the researcher is in a position to use Observed data to fit the relationship. Of the two types Of data, the cross-section data and the time- series data, the second type of data was used in this thesis since it was more easily available. The quantity data for annual consumption were computed by dividing the total annual domestic disappearance (consumption) of the products examined by the population (both humans and/or animal population). The resulting figure represented the average annual consumption for an average consumer (human and/or animal). The prices used were annual weighted prices published in various sources, and the method utilized was multiple regression analysis which was briefly discussed in the chapter on methodology. In the section on theory of demand the variables which enter into the demand function were provided. In this section the a priori knowledge of the variables which enter into the demand function in the empirical analysis will be given for each product separately. 2S.P. George and G.A. King, "Consumer Demand for Food Commodities in the United States with Projections for 1980," Giannini Foundation Monograph 26, California Agric. Exp. Station, March 1971. Pp. 1-2. 81 Factors Associated with Demand for Feed-Grain and Roughage Price Factor Certainly, it would be expected that a change in price will result in a change in the quantity of feed-grain demanded. The impact of feed-grain prices on the demand for it is different under the various uses, availability Of substitutes, level of per animal con- sumption and the kind and size of livestock population existing in Greece over the sample period. In Greece it is difficult to determine the relationship be- tween changes in prices of feed-grain and demand, due to the lack Of complete and reliable data either of a time-series or cross-sectional nature for consumption per animal. But it is logical to expect that the proportional change in the quantity of feed-grain and roughage consumed by cattle and, for that matter, used by cattle growers will be less than a proportional change in price since these feed-stuffs are the most prevalent stuffs in cattle feeding. This means that the demand for feed-grain and roughage is expected to be inelastic with respect to the change in price. Cattle Population Cattle population is an important factor in determining the general level of demand for feed-grain and roughage. Thus, with an increase in cattle population, it is in general expected that the de- mand for feedstuffs would increase. The Price Of Related Goods Cattle population is usually fed with feed-grain and fodder. plants; and roughage should be considered as a related good to feed- 82 grain and vice-versa. Pasture condition and quality should have been other factors if there were data available on them since, if pasture conditions and quality were improved, then it should be expected that cattle would consume less feed-grain. Cattle Feeding Practices and Management Cattle feeding practices and management also cause demand for feedstuffs to change from time to time. These practices involve not only the feedstuffs being considered here, but also other feedstuffs as well which may be substituted for those under consideration. Since cattle population was not fed with other feedstuffs over the sample period considered (but only late in the 19605), it can safely be con- cluded that there was no other product that could replace feed-grain and roughage in the cattle feeding practices in Greece for that period. Farm Income Changes in farm income Often lead to major changes in demand for feedstuffs. The degree of impact will depend largely on the existing level of farm income and income elasticity of demand for feed- grain and roughage. The Availability of Feed-Grain and Roughage The quantity of feed-grain and roughage available to the cattle growers always has had a positive impact on their quantity de- manded. The imports of these feedstuffs and the positive effect on the growth of the livestock industry which the P.L. 480 program has had, is a sign that, up to a certain extent, the larger the quantity available, the greater the demand. 83 The Prices of Livestock Products Prices of livestock products is an important factor in deter- mining the level Of demand for feedstuffs since the demand for these stuffs is an (immediate) derived demand because it is derived from primary or consumer demand for livestock products. Generally speaking, it is expected that the higher the prices of livestock products are, the higher the quantity demanded for feed-grain and roughage will be. Empirical Estimation of Demand The Variables Used In this section the empirical results are obtained by solving the various equations of the feed-grain and roughage model by means of OLS. Here all the variables used in the entire study are presented in order to have tham all together in one place for easy reference. Endggegous Variables QBD : Quantity Of beef demanded at the farm level divided by Greek population, in kg/head. BCSL : Number of beef cows slaughtered, in thousand head. YBC : Average yield per animal unit, i.e., per beef cow, in kilo- grams. QBIMP: Quantity Of beef imported divided by Greek population, in kg/head. QBS : Quantity of beef supplied which is the result of quantity of beef produced plus the quantity of beef imported, in thousand tons. The quantity of beef produced is the result of BCSL x YBC. Identity. QVD VCSL : YVC QVIMP: QVS QMKD : NCM YCM QMKS : QFGD : QFGP : QFGIMP: QFGS : QRD QRS FPFS : 84 Quantity of veal demanded at the farm level divided by Greek population, in kg/head. Number of veal-cows slaughtered, in thousand heads. Average yield per animal unit, i.e., per veal cow, in kilo- grams. Quantity of veal imported divided by Greek population, in kg/head. Quantity of veal supplied = quantity Of veal produced plus quantity of veal imported, in thousand tons. Identity. Quantity Of milk demanded at the farm level divided by Greek population, in kg/head. Number Of cows milked, in thousand head. Average yield per animal unit, i.e., per milk cow, in kilo- grams. Quantity Of milk supplied, in thousand tons. Identity. Quantity of feed-grain demanded at farm level divided by the animal population,'h1kg of TDN. Quantity of feed-grain produced, in thousand tons of TDN. Quantity Of feed-grain imported, in thousand tons of TDN. Quantity of feed-grain supplied = QFGP + QFGIMP, in thousand tons TDN. Identity. Quantity of roughage demanded divided by the cattle population in Greece, in kg of TDN. Quantity of roughage supplied which is equivalent to quantity of roughage produced since no imports of roughage are assumed here, in thousand tons of TON. Farm price of feed stuff deflated by the CPI, in drs/kg. 85 Exogenous Variables GNP T FP(L+M) QBC BCSLt_1 (FPB/FPFG)t-1: 0V V RPV RP(L+M) Gross National Product divided by the Greek population and deflated by the CPI, in drs/head. Time variable. Farm price of lamb and mutton deflated by the CPI, in drS/kg. Quantity Of beef consumed at the retail level divided by the Greek population, in kg/head. Beef cows slaughtered in t-l period, in thousand head. Ratio of farm price Of beef over farm price of feed- grain deflated by CPI in period t-l. Dummy variable for subsidy of 2 drs/kg liveweight paid for animals weighing more than 250 kgs liveweight for the years 1963-1968. ‘ Retail price Of beef deflated by the CPI, in drs/kg. Quantity of veal consumed at the retail level divided by the Greek population, in kg/head. Number of total cows in t-l period, in thousand head. Farm price of veal in t-l period deflated by the CPI, in drs/kg. . Dummy variable for subsidy given in the years 1965- 1972 in order to promote increase of veal production, in drs/kg. Retail price of veal deflated by the CPI, in drs/kg. Retail price of lamb and mutton deflated by the CPI, in drs/kg. QMC FPCH NCM FPMKt_1 DV AUF 86 Quantity of milk consumed at retail level divided by the CPI, in kg/head. ' Farm price of cheese deflated by the CPI, in drs/kg. Number of cows milked in t-l period, in thousand head. Farm price of milk in t-l period deflated by the CPI, in drs/kg. Dummy variable for milk subsidy given to the farmers during the years 1965-1972 to promote increase in milk production, in drs/kg. Animal units fed. Here total carcass meat production of all categories Of meat is taken and divided by 100 kgs of carcass weight; the result is taken as animal units; in thousand units. Farm price of feed-grain in t-l period and deflated by the CPI, in drs/kg. Wholesale price of corn in USA, in cents/kg. Animal units fed in t-l period, in thousand units. Farm price of roughage in t-l period deflated by the CPI, in drs/kg. Dummy variable for subsidy given to the alfalfa grow- ers during the years 1965-1972, in drs/stremma. Beef cows slaughtered in t-i period, in thousand head. Veal cows slaughtered in t-i period, in thousand head. Number of milk cows in t-i period, in thousand head. Farm price of beef in t-i period deflated by the CPI, in drs/kg. (FPMK/FPFG) (FPMK/FPFS) FPMKt_i (FPB/FPFG) (FPMK/FPFG)t-i: (FPB/FPFG)t-i : FPB/FPR CPI POPN FPFSt_1 87 Farm price of milk over the farm price of feed-grain; this variable proxies the profitability of the milk enterprise and is generated by dividing the two vari- ables. Farm price Of milk over the farm price Of feed-stuffs; the last variable includes both feed-grain and roughage as well. Farm price Of milk lagged i periods and deflated by the CPI. Farm price of beef divided by the farm price of feed- grain; this proxy variable is taken to give the pro- fitability of the beef production enterprise. The same variable as above lagged i periods. The same variable as above lagged i periods. Farm price of beef divided by the farm price of the roughage variable to proxy the cost relation of the beef production enterprise with respect to roughage. Consumer Price Index. Population of Greece in thousand head. Farm price of feedstuff divided by the CPI, lagged ~i periods, in drs/kg. 88 The Demand for Feed-Grain* and Roughage ** From the factors listed in the section before the previous one which seem to be associated with the demand of feed-grain and roughage, the most important one is the factor animal population. The two equations fitted to explain the variation in the quantity of feed- grain demand and roughage demand, respectively, were as follows. (FGI) OFGDt = 611.2921 + 0.5599AUFt (200.2921) (.0912) 'R2 = 0.80 6* = 1.34 (i) (5-9)*** (R1) QRDt = 68.7566 + 0.5606AUFt (76.0652) (.0346) ‘R2 = 0.96 ‘ (5-10) Equation (5-9) is the most promising one among many others tried. It seems to support the a priori knowledge since 80 percent of the variation Of the demand for feed grain is explained by the number of animal units fed (AUFt). * Feed-grain refers to feeds which belong to the category of concentrates and include the various grains (wheat, corn, oats, barley, rye, etc.) and the high-grade by-products, such as wheat bran, cotton seed meal, linseed meal, corn gluten feed, sugar beet pulp, etc. They can be either low in protein or rich in protein. ** Roughage refers to feeds that are high in fiber and therefore low in total digestible nutrients. Such feeds as hay, corn fodder, alfalfa, glovers, straw and silage belong to this class of feedstuffs. In all equations in this study the standard errors of the coefficients are given in parentheses, and d*: Durbin-Watson Statistic: i: inconclusive. na: negative serial correlation. 2a: Zero serial correlation and 'R2: coefficient of determination corrected for the degrees of freedom. 89 The variable carries a positive regression coefficient, mean- ing that the greater the number of animal units fed, the higher the quantity of feed-grain demanded will be. Also, the variable is statistically significant, thus giving more faith to reliance on this relationship. The model shows that for a unit change in the animal units.fed a 0.5 unit of change appears to take place in the quantity of feed-grain demanded. The same holds true for roughage. Here, again, the most im- portant factor seems to be the number of animal units fed. The vari- able AUF is positively correlated to the QRD, meaning that the greater , the number of animal units fed with roughage, the larger the quantity of roughage demanded will be. The variable is statistically signifi- cant, and the explanation power of the equation is 0.96 which means that 96 percent of the variation of the roughage Observed is explained by the number of animal units fed. There is a 96 percent probability - out of 100 - that the two variables move together. The quantitative relation between the two variables is such that for a one unit change in the variable AUF a 0.5 unit change corresponds to the variable QRD. Imports of Feed-Grain It was thought that after an examination of the demand for feed-grain in the previous section the imports Of feed-grain should be given in a subsequent (successive) section to provide a more com- plete picture of the demand for feed-grain. Feed-grain imports expressed in TDN showed an upward movement during the sample period, starting from 83,000 tons of TDN in 1951, 90 reaching a peak Of 770,000 tons of TDN in 1970 and 1971 and then de- clining to a level of 128,000 tons of TDN in 1972. A brief comment is in order here. Prior to World War II, the world feed-grain economy experi- enced increasing world surpluses, declining trade and falling world prices. Since World War II, signifiCant developments have taken place in the world feed-grain economy. Thus, in 1953-1954 the United States Congress enacted P.L. 480 in order to provide for the distribution of U.S. wheat on a concessional or noncash basis. The world's two largest feed-grain exporters, the United States and Canada, experi- enced their largest accumulation of feed-grain stocks in history, and Canada sold sizable quantities Of wheat to Communist China and the U.S.S.R., and, for the first time in history, Canadian farmers were paid to take land out of feed-grain production. The European Economic Community (EEC) was formed, and, because of increased price supports, caused cereal production to increase substantially. Up to that point, there was a rather economically favored world environment for Greece's imports and the only problem Greece had to face was that of her balance of’payments. Since 1972, however, conditions have begun to change in the world market. During the last two years the U.S.S.R. imported large quantities of wheat from the U.S. A drought in South East Asia and in Central Latin America caused some more severe problems in feeding not only the animal population but the human population as well. The question has been raised in recent years as to whether this feed- grain world deficit is a permanent one and how long will it take to be over. 91 Today's world picture is highlighted by the extreme swings in exports, commodity prices and wide areas Of starvation occurring over the last few years. These problems were emphasized in the Rome Food Conference supported by the United Nations.34 In 1973 world grain reserves fell to the lowest level since 1953; the Yom Kippur War initiated a five-fold increase in Oil prices, affecting fertilizer prices and production and transportation costs, and induced world-wide inflation which reached the level of 33 percent in Greece in that year, the highest rate of inflation the world over. To these dire circumstances add the uncertainty of knowing whether the U.S.S.R. and Communist China would enter the market and buy siz- able quantities of grains as they did in 1960 and the picture becomes even more gloomy. At this point it is necessary to mention this fact as necessary background information to an understanding of Greece's open scenarios from now on. 1 Greece has three general options Open to her. The first is to buy feed-grains from the world feed-grain market, thus keeping her status with EEC as it stands today. The second option is to pursue a self-sufficiency policy by subsidizing her feed-grain production and the third scenario is to join the EEC as a full member and follow EEC's Common Agricultural Policy (CAP). 3United Nations Economic and Social Council, World Food Conference, Notes by the Secretary General, November 22, 1974, "Hunger and Diplomacy: A perspective on the U.S. Role at the World Food Con- ference," Submitted to the Subcommittee on Foreign Agriculture, Committee on Agriculture and Forestry, United States Senate. (Washington, D.C.: U.S. Government Printing Office, 1975). 92 The equations explaining the variations of feed-grain coming into Greece were the following: QFGIMP = 30.7171 - 0.1748OQFGP + 0.2679AUF t (00.9451) (0.089) 1“ (0.0590) t 'R2 = 0.54 6* = 2.11 (za) (5-11) 'QFGIMPt = 60.4664 - 3 561.2153PCORN + 0.1681AUF (651.5595)(15 455.5335) ”5A (.0701) t" 82 = 0.43 6* = 1.76 (za) (5-12) From these two equations it can be seen that the number of animal units fed this year and the year previous constitute a major factor in determining feed-grain imports into Greece. The variable appears to be statistically significant in both equations and bears the right positive sign, meaning that the greater the number of animal units fed, the greater the quantity of imported feed-grain. The variable quantity of feed-grain domestically produced bears the right sign in the first equation meaning that the lesser the quantity Of feed-grains produced in year t-l, the greater the quantity Of imported feed-grain. The variable seems to be statis- ically significant, and its estimator bears significant economic weight. Both models give a rather low R2, meaning that some other economically relevant factors have been left out. The import elasticity of demand for feed-grain with respect to animal units fed is found to be +0.3 in equation (5-11) and +0.2 in equation (5-12). 93 This, again, shows that the import elasticity Of demand for feed- grain with respect to AUF is very low, verifying the fact that feed- grain had "been pulled" into Greece because it was needed to feed and keep the cattle herd going. Furthermore, this means that feed-grain produced in Greece is not enough to even feed the cattle herd over the time considered. The low import elasticity of demand for feed-grain lends support to structural changes in the feed-grain industry to increase yield. Factors Associated with Demand for Beef, Veal and Milk Factors associated with the demand for these livestock pro- ducts are discussed in this section in order to give a concise and clear account of what factors to expect in the empirical analysis of demand for these products. Product Price Certainly, it would be expected that a change in price will result in a change in consumption of these livestock products. The impact Of livestock prices on the demand for them is different under the various uses, level of incomes, availability of substitutes and level of per capita consumption. It is logical to expect that as the price of these livestock products increases, the quantity demanded will decrease. Consumer Personal Disposable Income Changes in consumer personal disposable income led to major changes in demand for the livestock products at hand over the sample 94 period examined. The degree of impact depends largely on the existing level of per capita income and income elasticity of demand. Generally speaking, a positive relationship between income and quantity of beef, veal and milk is expected to hold as income rises. Population Population is an important factor in determining the general level of demand. With an increase in population, it is expected that the demand for beef, veal and milk would increase. The population increase which Greece experienced over the period 1951-1971 was in absolute numbers equal to 1,135,840; i.e., 56,792 persons per year. In percentage terms this means that the population increase amounted to 14.88 percent over this twenty year period which corresponds to an annual increase Of 0.74 percent.4 The population parameter enters indirectly into the empirical functions fitted, since consumption Of livestock products is given in annual per capita form. The Price of Related Goods Other than beef and veal, people in Greece eat fish, vege- tables, fruits, lamb and mutton, poultry meat, pork and other dairy products in addition to fresh milk. It is generally expected that as the prices of (other) related products go In), the quantity Of live- stock products at hand will increase. This can be seen in the 4N.S.S.G., Statistical Yearbook of Greece (Athens, 1971), p. 18. 95 relationship called cross-demand elasticity and/or under the labels of complements and substitute products. The Range of Foods Available to Consumers Certainly, if Greek people have a greater variety of foods to eat, they will probably consume less livestock products of this kind and vice-versa. Consumer Tastes and Preferences Consumer tastes and preferences also cause demand to change from time to time. These preferences involve not only the food pro- ducts being considered, but also other products which may be sub- stituted for those under consideration. And since Greek people con- sume products other than beef, veal and milk, it can be concluded that there are other products that can replace beef, veal and milk in their diet. In Greece, it is difficult to determine the relationship be- tween changes in prices and demand because of the lack of complete and reliable data in the form of time-series data on consumption. This is why the models here were kept simple; but, nevertheless, they include the factors that seem to play a crucial role in explaining the variations in demand for the livestock products considered in this analysis. Both retail and farm demand relationships have empirically been tested here. 96 The Demand for Beef Retail Level A considerable amount of research has been devoted to estima- tion of demand relationships for food products in the developed countries and in the United States in particular. One of the more recent and comprehensive studies of demand was carried out by P.S. George and G.A. King at the University of California.5 They estimated price elasticities, all cross-demand elasticities and income elas- ticities for 49 foods. Retail Demand for Beef The equation tried for the estimation of the retail demand of beef was as follows: 080 = 6.00492 - 20.5973RPB - 0.3270RP(L+M) + 0.197SGNP (1.5218) (10.3756) (.0477) (.0157) -2 R = 0.90 d* = 1.94 (2a) (5-13) Equation (5-l3) gives the retail demand for beef in Greece in terms of its own price (retail price of beef), the retail price of lamb and mutton and the per capita Gross National Product which is taken to proxy the personal disposable income variable. The equation seems to be quite promising since it gives the right signs for the regression coefficient and a rather satisfactory coefficient of 5P.S. George and G.A. King, "Consumer Demand for Food Commodities in the United States with Projections for 1980," Giannini Foundation Monograph 26, California Agricultural Experiment Station, March 197l. 97 determination. In terms of its own price, the retail price variable seems to be statistically significant since the magnitude of the estimated re- gression coefficient is almost two times its standard error. This then means that there is a fairly high probability (about .95) that the true parameter is not zero. The sign in front of the estimated regression coefficeint of retail price of beef is what one should expect from economic theory. Since this is a demand relationship the quantity of a product demanded and its price are inversely associated. The minus sign verifies the downward sloping demand curve for a livestock product. The corrected coefficient of multiple determination, R2, was found to be equal to 0.90. All this means is that there is a high degree of linear association between the variables included in the model and the quantity of beef demanded (QBD) for consumption. In other words, 90 percent of the variation in the per capita quantity of beef demanded for consumption in Greece is associated with the variables appearing in the empirical estimation. The variable retail price of lamb and mutton seems to play an important role in explaining the demand variation for beef. The per capita consumption of lamb and mutton moved in a parallel way and at higher levels of consumption than that of beef over the entire period of the sample (see Table B-4in the Appendix). 98 The sign of the regression coefficient is Consistent with the logic of the model and of the economic theory according to which a negative sign was expected since beef and lamb and mutton (taken to- gether) are complementary products in the diet of Greek consumers. The magnitude of the estimated regression coefficeint.seems to be reasonable since a unit change in the retail price of beef is asso- ciated with a 0.3 unit change in its quantity demanded per year and per capita, which seems to be logical. The variable RP(L+M) is statistically significant since the magnitude of the estimated regression coefficient is well above two times its standard error which makes using this variable in the model more reliable in explaining the variation in the demand for beef in Greece over the‘period l95l-l972. _ The other striking point about the model (5-l3) is that no autocorrelation appears to hold following the standard Durbin-Natson criterion for autocorrelation. The GNP variable appears to play its role in the explanation of the variation of demand for beef in Greece for the time period 'examined. GNP almost doubled at the end of the sample period. This variable is positively associated with the dependent variable QBD, meaning that as income (or, for that‘matter, the GNP) increased, the quantity of beef demanded increased, too. The magnitude of the coefficient seems to be reasonable since it means that whenever the GNP changes by l,000 drachmae per year, the per capita quantity of beef demanded changes by 0.198 kg per year. Retail Elasticities for Beef The three demand elasticities for beef at the retail level are shown in TablelO below. TABLE 10 * RETAIL DEMAND ELASTICITIES FOR BEEF Dependent Variable 4 Independent Variables RPB RP(L+M) GNP QBD -l.4l -0.02 + 5.90 In Table 10 the three elasticities of demand for beef are given. The own-price elasticity was found to be -l.4l and reveals that beef demand was price elastic in Greece over the sample period. The cross-price elasticity of demand for beef with respect to lamb and mutton (taken together) is negative, indicating that the two pro- ducts are complementary commodities. The income elasticity of demand for beef at the retail level was found to be +5.90, revealing that the beef is considered a normal good in Greece. Empirical estimates of price elasticities of demand vary, depending upon the price and quantity series used in the analysis. In addition, they are also influenced by the kind of shift variables included in the equation. Furthermore, they are also affected by the kind of shift variables included in the regression equation, i.e., the factors that are held constant. In this respect Norking's dis- cussion of what happens to demand price elasticities when quantities rather than prices are used as measures of substitutions in a *All elasticities have been calculated at the mean value. 100 regression analysis is relevant. He states that "if prices of other meats are held constant, the elasticity of demand is somewhat greater than if supplies are held constant."6 The Relationship Between Farm and Retail Demandfifor Begf_ The farm price and the retail price of beef are connected by the marketing margins. If retail and farm prices have the same per- centage relationship to each other, the elasticity of demand at the two levels will be identical. A one percent change in price will re- sult in equal percentage quantity changes at the two levels. But if thennrketingnmrgins tend to be constant in drachma terms and not in) percentage terms, then the elasticity at the retail level will be greater than at the farm level. Unfortunately, there are no data available for marketing margins in this case in order to observe their behavior and their relationship to both farm and retail prices. Still, it is expected that margins may gradually have widened due to pro- vision of better marketing services provided over time, including use of better equipment such as vehicle refrigerators for transportation, freezing facilities at the retail level, etc. The relationship between farm and retail demand in the mathe- matical language of economics is based upon Harlow's work on hogs (1962).7 A formula which combines the above statements is derived in order to compare the farm and the retail price elasticity. 6Nlmer J. Working, Demand for Meat (Chicago: University Press, 1954), p. 69. 7A. A. Harlow, Factors Affecting the Price and Supplygof Hog_, USDA Technical Bulletin, No. 1.71962. 101 The elasticity of demand at the retail level is indicated by the formula: e g dQ§Q_. RPB r dRPB 0B0 where RFD is retail price of beef and 080 is quantity of-beef con- sumed at the retail level. ‘By the same token, the elasticity at the farm level is indicated by the formula: e = d 80 . fig f dFPB 0B0 where FPB is the farm price of beef. From calculus it is known that ngD = ngo . dRPB d PB dRPB dFPB 8y substituting this latter expression in the equation for farm level elasticity and by multiplying by ggg, the following two relationships are derived: OY‘ e .8 4102.539. f r dFPB RDB The relation gggg, which is a ratio of changes between retail and farm prices of beef, depends upon the behavior of marketing margins. a) If retail and farm prices always move in a parallel way bearing the same percentage relationship, RPB = a.FPB, dRPF _ RPB HEPB" a - ?fi§3 then, ef = e , i.e., the two and r elasticities are equal. 102 b) If margins, however, are constant in drachma terms, RPB = FPS + b and the ratio gggg = l. Hence, the ef taken from _ . FPB . the above given relationship becomes ef - er fifigu Since PB farm prices are usually lower than retail prices, RPB < l, and, hence, it is proven that the farm price elasticity is less than the retail elasticity. These findings are similar to those of Brandow8 who states that farm prices are lower than retail prices by the amount of market- ing margins, and, hence, a one percent change in the farm price has less effect on the volume moving through the marketing system into con- sumption than does a one percent change in the retail price. This has serious policy implications since the results are different if farm price is raised and given to the farmer or retail price is lowered and given to the consumer. Generally speaking, farm-level demand for domestic agricultural products is less elastic than the retail demand, another reason why small changes in the production level of farm produCts cause large changes in farm prices. Waugh concludes also that with percentage price spreads, both "price flexibilities", i.e., that of retail price with respect to quantity and income held constant and that of retail price with respect to income with quantity held constant, are the same at retail and at farm level. Prices are more elastic (more flexible) 8G. Brandow, Interrelations Among Demands for Farm Products and Implications for Control of Market Supply. Penn. State Univ. Ag. Exp. Sta. Bul. 680, 1961. 103 at the farm level than at the retail, both with respect to quantity - and income, if the price spread is a constant number of drachmas and its divisions.9 The Equations Fitted The equations fitted to explain the variation in the quantity of beef demanded at the farm level were as follows: QBD = 6.2992 - 30.9965FPB + 0°]996C0Mt-1 - 0.0276GNP (1.9040) (12.8666) (.0419) (.0193) R? = 0.70 6* = 1.96 (za) (5-14) QBD = 0.3582 - 9.8771FPB + 0.0473GNP (2.1009)(17.6515) (.0162) R? = 0.36 6* = 1.84 (za) (5-15) QBD = 5.6978 - 8.0644FPB + 0.0321GNP - 18.7108FP(L+M) (2.8543)(12.2457) (.0277) (16.7250) R? = 0.76 6* = 2.44 (za) (5-l6) The first of the above equations is given as a function of three in- dependent variables: the farm price of beef (FPB), the consumption of all other meats in the previous year (COMt_]) and the GNP variable. The general observation related to the first equation is that the income variable takes the wrong sign since it is expected that the general upward movement of GNP should result in an upward trend of demand 9F. Waugh, Demand Prjge Analysis: Some Examples from Agriculture, USDA, ERS, Technical Bulletin, 1964. 104 for beef at the farm level, given the fact that beef has been found to be a normal good at the retail level. The variable is not, of course, statistically significant, and one may drop it from the model. Nor is the economic weight which the variable carries that signifi- cant. This finding leads one to think that beef consumption in Greece has reached such levels that further consumption should occur only if the product takes other forms of appearance (i.e., ready meals, hamburgers, etc.). The variable COM enters with a positive sign, which means t-l that, as the consumption of all other meats increases, the quantity of beef demanded at the farm level is increased, too. This is somewhat puzzling, but since nothing is known about the behavior of margins in the retail-wholesale business of beef, nothing more can be said about it. The variable appears to be statistically significant. For a one unit change in the per capita consumption of all other meats, which occurred the prior year, the quantity of beef demanded in this year and at the farm level is increasing by only 0.2 unit per capita per year. In other words, if the consumption of all other meats in- creases by l.0 kg per head per year, the per capita farm level de- mand increases by 200 gr. per head per year.. The price of its own, i.e., the farm price of beef, carries a minus sign which means that as the farm price of beef goes up, the quantity of beef demanded at the farm level is expected to go down. This is in accord with the ceteris paribus condition and with economic theory. The variable is statistically significant, and its economic weight seems to be rather great since for a one unit change in FPB, 105 a 31 unit change in the quantity of beef demanded is expected. This is an overestimation of the estimated regression coefficient of the FPB variable, and it is believed that is so due to 0LS method adopted for estimation. The income variable appears to have a wrong sign and it not statistically significant. It is better to drop it from the equation. The second equation (5-15) fitted to explain the variation of the demand at the farm level is a function of the FPB and of the GNP. In such a formulation the GNP variable takes the positive sign and is statistically significant, while the FPB variable keeps it minus sign, I but becomes statistically non-significant. The magnitude of the estimated regression coefficient is still large. Apparently, the R2 is significantly reduced from 0.70 in the first model to 0.36 in this second formulation. Finally, the third formulation (5-l6) seems to be the most promising one since the signs of all the independent variables are in accord with economic theory, though almost all of them could not be considered as statistically significant. Analytically, under the (5-l6) formulation the FPB variable carries a minus sign, meaning that whenever the farm price of beef is increasing the quantity demanded at the farm level is expected to decrease. The variable is not statistically significant in this kind of formulation, but it still carries a large estimated coefficient. The GNP variable carries a positive regression coefficient, meaning that as the per capita income increases, the quantity of beef demanded at the farm level is expected to increase. 106 The farm price of lamb and mutton variable FP(L+M) seems to be in line with what was found at the retail level. The variable carries a minus sign, meaning that beef and lamb and mutton are complements in the diet of Greeks. As the farm price of lamb and mutton increases, the quantity of beef demanded at the farm level is expected to decrease. That lamb and mutton are seasonally consumed means that within the time limits of lamb and mutton consumption the quantity of beef consumed is really reduced. During the period between Christmas and Easter time the consumption of lamb and mutton reaches its high- est peak. At this time the quantity of beef consumed is reduced which is reflected at the farm level. This is the meaning of the minus sign of the estimated regression coefficient of the FP(L+M) variable. Over the other time period the two goods are substitutes. The estimated regression coefficient seems to have been over- estimated since for one unit of change in the farm price of lamb and mutton l8 units of change correspond to the quantity of beef demanded at the farm level, which seems to be a rather great change. Elasticities of Demand for Beef at the Farm Level The elasticity of demand for beef at the farm level was calculated from the (5-l5) model and was found to be equal to EPF = -0.506, which verifies the theory exposed at the beginning of this section which says that the farm-level demand is less elastic * than the retail one. *The EPB calculated from the other two models was found to be -l.58 and -0.4l3, respectively, from (5-l4) and (5-l6). 107 The cross-product elasticity of beef demand at the farm level with respect to the farm price of lamb and mutton at the same level was found to be ECBF.= -1.0918, which is obviously less than zero, meaning that the two products are considered to be substitutes. ‘ Thus, in concluding this section it can be observed that the two cross-product elasticities (at the retail and at the farm level) . support the view that beef and lamb and mutton could be considered as beingsubstitute» products. For this to be further verified, how- ever, more data on seasonal consumption patterns are needed. Imports of Beef Imports of meat in Greece are subject to: (l) an import duty ranging from 15 to 18 percent of the imports value and (2) a "floor price" system introduced in 1964, according to which imports are banned altogether if price of meat at home is at, or lower than the floor price. ‘In the world market meat prices are lower than the domestic ones. And some prices (in Greece) are too high due to the protection of the industry expressed in the form of the import levy and to the preferences of Greek consumers for "fresh," locally pro- duced meat. The reasons for the protection of the industry are to save. foreign currency and to support domestic meat production through its further development, thus maintaining a reasonable level of income for Greek meat producers. Another point is worth mentioning here. Greece imports a lot of meat and sets her imports free of levy deliberately in seasons of high consumption of meat which coincide with certain religious feasts (Christmas and Easter time). But this 108 does not mean that meat imports are not realized during other times of the year. It simply means that at these religious feast times imports of meat reach their peaks. The beef import regression equation is a part of the beef sub-system model. The following equations were tried: QBIMP = -1.6972 + 3.5514FPB + 0.7801GNP t (.7634) (4.9375) t (.0758) t 82 = .87 6* = .47 (5-17) QBIMPt = -1.0550 - 10.3551FPBt + 0.621SQBC + 0.2281GNPt (.5578) (4.7101) (.0645) (.0052) _2 'k . R = .94 d = 1.23 (1) (5-18) QBIMP = 1.0265 - 13.9404FPB + 0.4141QBC + 0.1200T t (.9879) (6.3611) 1 (.1211) (.0345) .2 * .* A R = .92 d = 1.49 (1) (5-19) From the equations above it can be observed that in the first one the variable FPBt is statistically significant but does not bear the right sign, although seems to have a rather fair economic weight. In the other two equations this variable bears the right sign as one would expect from a priori reasoning of economic theory, and in both equations it is not statistically significant but economically very important. FPBt and 08Ct are serially correlated. Indeed, the partial correlation coefficients are particularly high, and the d* statistic gives an indeterminant Durbin-Watson test for serial *In equation (5-19) the 6* statistic with the value 1.49 rejects the null hypothesis only at 1% level of significance. 109 correlation. When serial correlation is not present, as in equation (5-17), the sign is not corrected, but its statistical significance becomes questionable. The variable QBCt seems to verify the a priori knowledge of the beef import industry and its seasonality in character since this variable is statistically significant, bears the right sign and is economically important. The GNPt variable is also significant, bears the right sign and has important economic significance. A time trend used in equation (5-19) in place of the GNP variable did not improve t the equation, although it is statistically significant. The best equation seems to be (5-18) for it carries lower standard errors for each of its coefficients and higher explanatory power (R'2 = .94), and the signs of the variables used are in accordance with the reasoning of economic theory. Imports are mainly influenced by the retail prices of meat (beef here) prevailing in the domestic market; thus, the retail price of beef should have been used instead. Such a formulation was tried by using the variables RPBt and QBPt_1 (quantity of beef produced in the t-l period). = -4.2863 + 21.2586RBP + 0.068908P QBIMP t t-1 (1.2093) (5.5559) (.0360) t .2 * . R = .52 d = 1.16 (1) (5-20) In this formulation the retail price of beef in year t seems to play a role on the quantity of beef imported since, other things being equal, a unit change in the RPBt causes 21.26 units of change 110 in QBIMPt. The standard error of the regular regression coefficient of RPBt is rather small, and the variable seems to be statistically significant and to carry the right sign and has important economic) meaning. The Durbin-Watson statistic d* with a value of 1.16 becomes inconclusive at significant levels 5 percent and 2.5 percent, and it just coincides with the value of the upper limit du = 1.16 at l per- ' cent level of significance, i.e., it is on the border of accepting or not accepting the H0 hypothesis of non—existence of serial correla- tion. The second variable QBP seems at a first glance to play a t-l role in imports since it carries the right sign expected from a priori reasoning, but its standard error of the regular regression coefficient is not even half of the value of the coefficient and its economic meaning seems not to be that important since a one unit change in QBPt_] corresponds to a 0.06 unit change in the dependent variable QBIMPt. The overall explanatory power of this equation is rather poor (R2 = .52). Thus, from the empirical analysis presented in this section it seems rather safe to conclude that imports have been substantial and have rather heavily contributed to Greece's balance of payments problem. On the other hand, imports have been a source of un- certainty to domestic producers. It seems that the most important factors affecting beef imports are the domestic farm price for beef, which indirectly in- fluences both the supply of beef and the demand for beef, the in- creasing per capita income and the level of meat consumption already reached through habitual and income effects. Given the fact that 111 per capita consumption of beef was 16.9 kilograms per year in 1972 and that of EEC-9 was 24.3 kg, it is reasonable to expect that beef imports will continue to be a problem to be tackled in the future. Import Elasticities of Beef Tablell reveals that retail price elasticity of imported beef is rather high (+3.08), which means that whenever the domestic retail price of beef changes by one percent the quantity of beef imported changes by almost 3 percent. TABLE11 IMPORT ELASTICITIES OF BEEF Equation Dependent Explanatory Variables Variable FPBt GNPt RPBt QBPt_1 5-17 QBIMP - 2.21 - - 5-18 QBIMP -l.l3 +5.19 - - 5-19 QBIMP -l.52 - - - 5-20 QBIMP -2.31 - +3.08 +0.69 Average -l.44 +3.75 +3.08 +0.69 It is also extremely important to know how demand responds to changes in income. The responsiveness of demand to changes in income is termed income elasticity of demand and is defined as the percentage change in quantity of beef imported (demanded) over the percentage change in income (consumer's income). For most goods, increases in income lead to increases in demand, and income elasticity will be positive. Obviously, here the income elasticity of demand is positive, 112 ranging from +8.21 to +5.14 in the two equations (5-17) and (5-18). The reason for knowing the income elasticity of demand is that in- dustries with low income elasticities will find the demands for their products expanding only slowly, while industries with high income elasticities find the demand for their products expanding rapidly. This may be taken to mean here that with further income increases, other things being equal, beef imports will be still coming into the country. As far as the relationship between the percentage change in the quantity of beef imported and the percentage change of retail price of beef is concerned, it can be observed that, under the ceteris paribus condition, as the domestic retail price of beef rises by one unit the quantity of beef imported rises by 3 units. The retail price of beef in the domestic market is higher than in the world market, and the difference is leveled off by the import levy. Never- theless, the price import elasticity of demand for beef shows that, even with higher retail prices of beef, beef will still be demanded due to the strong income effect. Another striking result of the empirical analysis is that the price elasticity of beef imports with respect to farm price of beef which, in equation (5417), comes out to be positive and rather small and is not statistically significant, should be used with caution. The explanation lies in the fact that whenever the farm price in- creases by a small percentage farmers form positive price expecta- tions and hold beef cows from being slaughtered. The short-run re- sult would be a shortage of beef supplies in the market and, hence, an increase in the quantities of beef imported. 113 The Demand for Veal Retail Demand for Veal Veal wasconsumed at a rate of 4.96 kg. per capita per year in Greece, almost double the quantity compared with beef consumption. In fact, the EEC-9 average per capita consumption of veal was 2.5 kgs per year while that in Greece was 8.9 kg. in 1973. There is, of course, a difference in the definitions of the product "veal," but, nevertheless, habitual reasons explain why consumption of veal is higher in Greece. The higher consumption of veal in Greece is due: 1) to low per capita income over the first decade of the sample period, 2) to the freshness and flavor which constitute an important factor in veal purchases, 3) to better nutritional aspects and, finally, 4) to a "demonstration effect" which is well developed in the consumption pattern of the Greek people. These are the explanations for the in- come elasticity of demand for veal at the retail level which was found to be +1.86. It seems that, on the average, the typical con- sumer has already spent enough of his budget on veal purchases so that, inevitably, the income elasticity of demand for veal will be rather high. The Equations Fitted Two equations were tried to explain the variations in the quantity of veal demanded at the retail level. The two equations were as follows: 114 (va) QVO = 0.7236 - 1.8239RPV + O.6553RP(L+M) (5.9696)(14.9915) (.2414) + 0.5981GNP - 0.063OCPI (.0393) (.0492) .2 * _ R = 0.96 d = 1.32 (1) (5-21) (sz) QVD = -6.5543 + 4.5107RPV + 0.3689RP(L+M) + 0.063ZGNP (l.8505)(l4.3973) (.0934) (.0399) '82 = 0.96 6* = 1.32 (za) (5-22) Equation (5-21) reveals that the quantity of veal demanded is influenced by the retail price prevailing at the retail level. The retail price variable, of course, is not statistically signifi- cant. The variable bears the right sign since the dearer the veal becomes at the retail level, the less it is demanded by Greek con- sumers. The retail price of lamb and mutton bears a positive regres- sion coefficient which is what one should expect from economic reality in Greece. The dearer the veal becomes, the more lamb and mutton is demanded and vice versa. The variable seems to be statistically significant and to carry some economic validity since for one unit of change in the RP(L+M) a 0.6 unit change in ovo is caused which illustrates the competitiveness of the two kinds of meats. Lamb and mutton are more seasonally consumed, and the empirical evidence is more valid when the issue is looked at from a seasonal consumption point of view. 115 The GNP variable carries a positive sign which means that as the GNP increases, the quantity of veal demanded at the retail' level will also increase. The GNP variable is almost statistically significant in every equation tried, and in the first equation (5-21) bears a rather fair economic significance since a one unit change in GNP causes a 0.6 unit change in the dependent variable. The in- clusion of the GNP variable together with the CPI variable may be the cause for the indeterminancy of d* statistic for serial autocorrela- tion. There is, of course, a problem here, and that is, when one talks of the CPI he at the same time talks about retail veal price which is included as one of the items used to construct the con- sumer price index in Greece. The association between RPV and CPI is high and that is another reason for the d* statistic to be indeter- minant. The overall explanatory power of the (5-21) model is rather high, but the indeterminancy of the d* statistic for serial auto- correlation and the fact that the variable RPV is not statistically significant should be kept in mind when using this model. The (5-22) model is almost the same as the (5-21) model, the‘ only difference being that the CPI variable is excluded from the second one. This exclusion makes the variable RPV carry a wrong sign, although its economic significance increases quite a bit. When the indeterminancy of the d* statistic disappears, the RP(L+M) is almost not affected and the GNP variable loses some of its economic significance. The overall explanatory power of the equation remains' exactly the same. 116 Elasticities of Demand for Veal at Retail Level Equations (5-21) and (5-22) are used to calculate the short- run retail price elasticities which are shown in Table 12 below. TABLE 12 SHORT-RUN PRICE ELASTICITIES OF VEAL AT RETAIL LEVEL Dependent Independent Variables Variable RPV RP(L+M)__ GNP CPI QVDt -0.09 +0.03 +1.18 -1.69 QVDt -0.22 +0.01 +1.18 - Average -0.15 +0.02 +1.18 -l.69 Table 10 shows the short-run price elasticities of veal at the retail level with respect to certain variables. It can be ob- served that the direct own-price elasticity of demand for veal is ‘0 The dif- . -0.15, while that estimated by Papaioannou is -0.61. ference may be due to (l) the fact that Papaioannou uses the retail price of salted fish as one of the explanatory variables, while the RP(L+M) was utilized in this study and (2) different time-series were used in the two studies. The cross elasticity with respect to retail price of lamb and mutton did not exceed the direct own-price elasticity of demand . for veal which means that the model specifications are the correct 10M. Papaioannou, "An Analysis of the Supply and Demand Condition in the Animal Breeding Industry in Greece," Unpublished Ph.D. Thesis, Oxford University, 1970. 117 Ones. In Papaioannoufs study the own-price elasticity of demand for veal exceeds that of (the) cross elasticity only in TSLS formulation. The income elasticity in this study was found to be equal to +1.18, while it is 2.71 in Papaioannou's study. But the elasticity found here is plausible since the income elasticity for the USA is 0.60.11 The demand elasticity for veal with respect to CPI was found to be -l.69 which also seems to be reasonable considering that Greeks spend a lot on veal consumption. Farm Demand for Veal Substantial research has been done in agricultural economics on questions related to price differences between the farm level and the retail level of farm products everywhere in the world. Nonethe- less unanswered questions remain. The difference between the price received by producers and that paid by consumers is a marketing margin. Both producers and 'consumers are greatly concerned about the size of marketing margins, their changes and their incidence of changes for both these categories of people. This section will not, however, deal with marketing mar- gin behavior in the veal market in Greece because there were no data available to the author about these margins. Only the two elasticities will be discussed here. The Equations Fitted In order to compute the farm demand for veal two equations were tried. They were as follows: nBrandow, p. 19. 118 QVD = -12.6407 - 4.1808FPV + 0.1601GNP + 15.3368FP(L+M) (2.5992)(25.8933) (.0503) (41.6885) '82 = .83 6* = .88 (na)(il%) (5-23) QVD = -6.8720 - 12.7200FPV + 0.0731GNP + 15.0941FP(L+M) (2.2874) (18.7886) (.0418) (30.0717) + 0.3838T (.0915) 82 = .91 6* = 1.43(i) (5-24) The two equations fitted here are the same, the only difference between them being that a time trend was allowed to enter in the second equation. But this entrance of the time variable increased the R2 by only 8 points and left almost everything else unaltered, although a slight improvement in the standard error of estimates appeared in the second model. The own-price variable, i.e., the farm price of veal (FPV), appears to have the right sign since a negative estimated regression coefficient in front of it means that quantity of veal demanded at the farm level and the price of veal are inversely related. The vari- able appears not to be statistically significant, and this could proba- bly be attributed to the assumption made about the random error, i.e., that random errors are independent 0f one another may not be true. The income variable and the DVD at the farm level are positively related which is what one should expect from economic theory and based on certain assumptions about the behavior of marketing margins. The variable is statistically significant in both models. 119 The farm price of lamb and mutton is positively related, meaning that the two products are complements since an increase in the price of lamb and mutton is associated with an increase in the quantity of veal demanded at the farm level. This complementarity is also supported by the fact that Ec was found less than zero, i.e., EC = 0.56 < 0. There is a consistency in the findings of the empirical analysis as far as the complementarity of lamb and mutton (taken together) and beef and veal is concerned. Lamb and mutton are complementary to both beef and veal. The variable (FP(L+M)) is not, of course, statistically significant, but it does contribute to the explanation of the variation in the quantity of veal demanded at the farm level. The time element is statistically significant and positively associated with the quantity of veal demanded at the farm level which is really what happened over the sample period examined. The average cross elasticity of veal with respect to the farm price of lamb and mutton seems to be sufficiently good since it agrees with other analyses (see Table 13 below). Papaioannou, using salted fish as a substitute for veal, has found a cross-elasticity of 0.70 with OLS and 0.56 with 151.5.12 This elasticity was found to be 0.56 here or fell within the range of 0.56-0.58, using OLS and seems to be reliable. The sign and the magnitude of its coefficient is reasonable even when OLS is used. Its positive sign with respect to per capita veal consumption indicates that the dearer the lamb and mutton become, the greater the quantity of veal demanded and vice versa. .12Papaioannou, p. 156. 120 TABLE 13 VEAL DEMAND ELASTICITIES AT FARM LEVEL Dependent Variable Explanatory Variables FPV GNP FP(L+M) QVD -0.16 +3.01 +0.56 QVD -0.49 +1.37 +0.56 Average -0.32 +2.19 +0.56 Imports of Veal Because veal like beef, is regulated by the same import rules to which reference is made in the section dealing with imports of beef, these rules are not repeated here. What is not considered in that section is some theoretical reasoning as to how imports of both live cows and/or calves or imports of meat (beef and veal) influence farmers in formulating their price expectations and their foresight for:the market. The government usually sets domestic prices at no regular times and decides about the special import policy of meats one or two months before religious feasts when it estimates that domestic production will not cover consumption needs. This policy, along with institu- tional rigidities inherent in the whole system of import policy, creates uncertainties with regard to the duration of the new domestic price level for veal and the price of the imported veal. To this add the duration of imports and the perception of the two veal prices in farmers' minds. 121 Once the decision is made by farmers and once they have formed their price expectations it must be put forward for approval and support, if any, by the state. All this needs time to be executed, and the length of time depends on the policy followed by the govern- ment and its agencies and institutions. Due to this difference in price and/or production perception by farmers and by the government, inconsistencies in individual and collective aims come into the picture, inconsistencies which are multiplied by domestic and international market uncertainties, weather uncertainties for feed-grain, etc.and make the whole situation very complex. The equations tried for an explanation of veal imports were: QVIMPt = -3.0645 + 2.4357RPvt + 8.11790v0t + 0.00056NP. (l.0546)(10.1287) (.2983) (.0005) Re = 0.88 6* = 0.70 (za) (5-25) QVIMPt = 1.1756 + 0.32700v0t - 4.4270vat - 0.004SGNPt (.8334) (.0616) (4.8585) (.0146) _2 * . R = 0.87 d = 2.03 (1) (5-26) QVIMPt = -2.6059 - 4.2182vat + 0.05136NPt (.6846) (7.6960) . (.0161) _2 1k R = 0.66 d = 0.94 (na) (5-27) Equation (5-25) tries to explain the imports of veal in terms of retail price of veal (RPVt), quantity of veal demanded and per capita gross national product. This equation reveals that this 122 year's retail priCe of veal, although not statistically signi- ficant, carries a relevant economic weight since it shows thatii’the retail price of veal in the domestic market increases by one unit the quantity of veal imported increases by 2.4 units. This could be taken to mean that, although Greeks prefer to consume locally produced veal because of its freshness and taste, they are rather sensitive to A price changes on the other hand, and, if the price of domestically pro- duced veal increases by one real drachma, consumers turn to buying imported veal by 2.4 kg. more per head and per year. According to Papaioannou, Retail price of veal is used to account for any ban im- posed on imports owing to the low price in the national market, as well as for any case in which imports are set free of any duty for the necessary period of time in order to bring the increased retail price back to the 3.3323221:liviis.12801212331321???“ by ”mm" As was expected, the sign of the (RPV) coefficient is positive agrees with the previous study carried by Papaioannou and emphasizes the positive causal effect of retail price of veal on imports of veal which is in line with the a priori knowledge of the veal import industry. The presence of the negative autocorrelation contributes, of course, to an upward bias in the estimated coefficient of price and, consequently,.in the estimated elasticity of veal imports with respect to retail price of veal in the domestic market. The variable quantity of veal demanded per head and per year seems to be very positively associated with the imports of veal. It 13Papaioannou, p. 156. 123 carries a large positive regression coefficient, and the use of OLS technique greatly contributes to this problem. The variable bears the sign that one should expect from knowledge of the consumption patterns of veal by Greek consumers. As veal becomes dearer and dearer in Greece, and as domestic veal production does not meet con- sumers' demand, Greeks turn towards meeting their needs in veal by buying imported veal. The variable seems to be statistically signi- ficant. The income variable, although it bears the right sign, seems not to be statistically significant. The positive coefficient of in- come indicates that imports of veal will increase as income increases. The only explanation which can be given for the negative sign of the estimated regression coefficeint of the income variable is the inde- terminancy of the standard test (Durbin-Watson) for serial autocorrela- tion. There is obviously a symmetrical upward movement in the time- series data used for the variables included in the model which may contribute to the appearance of the minus sign for the income variable. Import Elasticities for Veal Table 14 reveals that the short-run price elasticity of im- ported veal with respect to domestic retail price is 0.41, while that with respect to FPV is 0.56. The positive and negative elasticity coefficients with respect to price variables each has its own explana- tion. And the explanations behind these variables differ from each other. As the retail price of veal goes up, it is obvious that con- sumers tend to substitute for domestic veal either other domestic sub- stitutes or foreign veal coming into national market whose price is 124 lower, while, when farm price of veal goes up, farmers slaughter their calves, thus causing a short-run decrease in the imports of veal. TABLE 14 IMPORT ELASTICITIES OF VEAL Dependent Variable Explanatory Variables RPV QVD FPV GNP QVIMP +0.41 +27.26 - +0.03 QVIMP - + 1.10 -O.57 - QVIMP ‘ - - -0.55 +0.17 Average +0.41 +14.18 -0.56 0.10 The import elasticity with respect to quantity of veal de- manded is very large and should probably be taken to mean that if per capita veal consumption is going to increase, then veal imports will dramatically increase. More reasonable elasticity seems to be that calculated from the (5-26) model. The veal import elasticity with respect to GNP seems to be very low. One reason for this seems to be the inclusion of the variable QVP in the same model, but it still re- mains low even when this variable is excluded, as happens in model (5-27). The best model seems to be the first one which gives positive elasticities with respect to three variables included since one should expect such a sign from economic reasoning and from the data. One comment is in order here. The veal import elasticity with respect to income is low because, on the average, Greek consumers spend enough of a proportion of their food budget on veal consumption. 125 The retail price elasticity reveals that any increase in domestic veal price will mean increase in imports. Thus, from a policy point of view it seems that if the government wants to reduce veal imports the best policy to follow is to increase farm price of veal. Of course, this is a short-run view and should be taken only as such. But an increase in the FPVt could probably mean an increase in the RPVt which, in turn, could mean an increase in veal imports. This empirical finding explains why the government steps into the market and regulates the retail prices and the imports. But this is only a short run solution to the problem, a solution which favors price regulations and veal imports control. From a long run perspective the increased price of veal should be expected to contribute posi- tively in farmers' minds as far as price expectations are concerned, and it is also expected, other things being equal, that farmers should positively respond and try to produce the quantity of veal demanded by Greek consumers. The Demand for Milk Retail Demand for Milk ‘Fluid milk is indispensable for human beings, especially during the earliest period of growth, and through the entire human lifespan as well. Milk is easily digested and assimilated, and the nutrients are applied in forms that are particularly adapted to the underdeveloped digestive systems of young human babies at birth. Also, milk is high in minerals, on the dry basis, and it is especially rich in calcium and phosphorus, the two minerals needed in largest - amounts by growing humans. This nutrient value of milk has been 126 recognized by Greek people especially during war periods and that is why Greek farmers have taken care to have one or two milking cows in their homes. Apparently, it is a cheap food, too, if the cows are grazing outside in the fields. The equations fitted to explain the variation of the quantity of milk demanded at the retail level over the sample period considered in this thesis were as follows: QMKDt = 79.7307 - 2 708.7250RPMKt + 26.8157RPCHt + 0.22380NPt (31.9533) (1 012.2449) (10.4292) (.1175) .2 * . R = 0.87 d = 1.07 (1) (5-28) QMKDt = 73.5651 - 2 490.3619RPMKt + 0.3025GNPt (36.2642) (1 148.0322) (.1291) __2 * R = 0.84 d = 0.41 (za) (5-29) In the retail side, the retail price of milk is of great importance since it is statistically significant, bears the sign ex- pected from economic reasoning and carries significant economic weight. .The estimated regression coefficient is rather overestimated since the two variables RPMK and RPCHS are strongly influenced by the GNP variable. Nonetheless, all the variables in this model seem to be statistically significant and they do have to contribute to the explanation of the variation of the quantity of milk demanded at the retail level. Cheese, which is a substitute for milk as a source from which one can get the nutrients embodied in milk, was found to be statistically significant,and the magnitude of its estimated 127 regression coefficient seems to be plausible. The problem of the in- determinancy about the serial correlation may contribute somewhat to both the large regression coefficient and non-statistical1y-significant variable in the (5-28 model. Milk and cheese are positively associated in the retail level. ‘ When the variable RPCHS is omitted from the model, things do not change much. The only noticeable difference is in Re. R2 is re- duced by three units in the (5-29 model, and the problem of auto- correlated disturbances disappears. Elasticities of Milk at the Retail Level The own-price elasticity, Ep, of demand for milk at the retail level was found to be -4.21, which means that as the retail price of milk goes up by one percent, quantity of milk demanded is expected to go down by 4.21 percent and vice-versa (see Table 15). This elas— ticity was found to be equal to -1.31 in Papaioannou's research effort. TABLE 15 ELASTICITIES FOR MILK AT THE RETAIL LEVEL Dependent Variable Explanatory Variables RPMK . GNP RPCHS QMKD -4.39 +0.45 +0.16 -4.03 +0.60 - Average -4.21 +0.52 +0.16 The two models fitted for the retail demand of milk gaVe in- come elasticities equal to +0.45 and +0.60 which provide evidence 128 that fluid milk is considered as an inferior good in the retail sector in Greece (Ey < 0). And the retail cross-product elasticity of milk with respect to cheese was found to be +0.16 (Ec < 0), providing evidence that fluid milk and cheese are complement goods. Farm Demand for Milk For the calculation of the farm demand for milk at the farm level the following equations were tried: QMKDt = 52.0933 - 3 050.9752FPMKt + 0.1807GNPt + 539.4646FPCHSt (16.6418) (524.9463) (.0811) (62.5893) -2 'k R = 0.97 d = 1.38 (1) (5-30) QMKDt = 73.5651 - 2 490.0000FPMKt + 0.30250NPt (36.2642) (1 148.0322) (.1291) _2 * R = 0.84 d = 0.41 (sa) _ (5-31) QMKD = -14.1406 - 1 464.6039FPMK + 2.5546T t t (17.1844) (657.1072) (.2509) _2 * R = 0.94 g d = 0.74 (sa) (5-32) From equations (5-30) to (5-32) it can be seen that within the time t year span farm price of milk plays an important role in explaining the variations in the quantity of milk demanded at the farm level. This variable is statistically significant in all models used and bears the sign expected from economic theory since the minus sign means that as the farm price increases, the quantity of milk demanded will decrease and vice-versa. The regression coefficients carry significant economic weight, although seems that the OLS used give upward biased estimators. 129 The income variable appears statistically significant and has the right sign. The time variable tried in one of the equations lends empirical evidence to support the data which show an increased per capita consumption of milk over the sample period. The farm price of cheese, again, has an impact on the quantity of milk demanded at the farm level. The two products seem to be sub- stitutes at the farm level since they are positively associated, which means that as the farm price of cheese goes up the quantity of milk demanded is expected to increase. Elasticities of Milk at the Farm Level .The own-price elasticity of milk at the farm level was found to be -l.05, or,.EpM > 1 (see Table 16). Hence, milk has an almost elastic demand at the farm level as was found in the first two models. TABLE16 ELASTICITIES 0F MILK AT THE FARM LEVEL Dependent Variable Explanatory Variables FPMK GNP FPCHS QMKD -l.38 +0.36 +1.66 QMKD -l.12 ; +0.60 - QMKD -0.66 - - Average -1.05 +0.48 _+l.66 The income elasticity at the farm level was found to be +0.48, meaning that as the income, i.e., GNP here, changes by 1 percent, the percentage change in quantity of fluid milk demanded at the farm level 130 is almost 0.5. The cross-elasticity of demand for milk with respect to farm price of cheese was found to be positive and equal to +1.66, which suggests that milk and cheese are substitute commodities in the farm as well as in the retail level. Since no imports of fluid milk had been realized there was not any equation fitted for that matter. CHAPTER VI SUPPLY Introduction Knowledge of supply response for agricultural products helps both in understanding the farmers' decision making process and in designing public policy (farm policy) related to the agricultural sector. The first of these two elements, the farmers' decision making process, is closely associated with the mechanism of supply response. On the other hand, public programs and farm policies related to supply regulation, stabilization measures, price support, etc. implicitly assume a certain level of supply response to changes in prices. Among the studies of various agricultural products, supply response for cattle and feed-grain should be of interest to both pub- lic policy makers and private policy makers (farmers and/or business- men involved) for at least two reasons: a) public farm policy was directed towards diversification over the first twenty years of the sample period in Greece to meet the uncertainty which is involved in agricultural production; only lately has it been directed towards more specialization in certain regions in an attempt to alleviate farmers' income; b) such a study contributes significantly towards the economic well-being of all the numerous participants involved. 131 132 1 the objectives of research on supply According to Nerlove, response include (1) improving the understanding of the mechanism of supply response, (2) improving the ability to forecast supply changes and (3) improving prescription of better solutions to problems related to agricultural supply. This study is focusing mainly on the first two objectives. The Role of Prices in Supply Response Farmers' decisions about their production depend on the re- lative costs and prices of all their inputs and outputs; they are also influenced by their overall income, by their liquidity situation, and by their expectations as to future price developments. The extent to which price changes affect the supply of agricultural products is thus often difficult to determine. Many other factors influence production; such factors in the short-run could be weather and diseases, and in the long-term, developments in technology, changes in cost of produc- tion and other factors. Unless price changes are large, such other factors may override their effects. Opportunities for work off the farm may also play a role. Moreover, the reaction to a price change in the short-term, i.e., in a period during which most production resources are fixed, is not necessarily the same as the reaction to one in the long-term, when machinery, buildings, etc. can be replaced or altered. The purpose for adjustment also varies between farms. The farm with a limited 1M. Nerlove. “Estimates of the Elasticities of supply of Selected Agricultural Commodities," Journal of Farm Economics, Vol. 28, 1956, pp. 496-509. 133 area is typically suitable to enterprises yielding a high return per unit of land, such as dairying, vegetables, etc. The larger farmer, however, may have greater flexibility in choosing his line of produc- tion - subject, of course, to natural constraints - but is liable to have at any given moment a heavy investment in capital equipment which is designed for certain enterprises and which cannot be quickly replaced. In the very long term, the possibility of changing farm sizes depends on the aggregate land and capital availability. In terms of production economics, an analysis of supply re- sponse to price changes often gives inconclusive results, particularly when a long-run response is sought. Cases are even noticed wherein a reduction in price causes a farmer toincrease his output in the short- term in order to maintain his income, if he has no profitable alter- native to such an action, or if he is short of feedstuffs. Further, the effects of rising or falling prices are almost never symmetrical: an increase in price, for example, may stimulate an expansion of produc- tion, involving investment in equipment, while a subsequent decline in price may not have much effect because the equipment remains in use. These considerations may seem to imply that price policy can have only a limited role in guiding production. It must be added that price policy has other aims besides that of demand and supply adjust- ment, and, indeed, the preoccupation with farm incomes restricts the scope for reducing prices in order to discourage output. There are, in fact, rather few cases in which support prices for major commodities have been significantly reduced. There are nevertheless some clear. cases wherein a prolonged price reduction resulting from market forces has cut back not only production, but also productive resources. .134, For several agricultural products short-term adjustments, mainly through the market, can be seen to occur, although usually with a time lag which may in some cases set up a cyclical pattern (Cobweb theorem case). Price policy therefore, cannot disregard its effects on supply. It should be realized, too, that when governments step in to provide price guarantees or any other price support scheme, they are bound to exert a sustaining influence on production. The level of returns to agriculture must - in the long-run - influence the volume of resources used within the sector and, hence, (through it) the volume of production. A price change for a specific product can definitely have a substantial effect if alternative lines of production are available or if alternative investment opportunities in other products are offered. Such substitution in production between various crops and animal feeding enterprises could thus take place, especially in a somewhat longer term. The determination of price relationships is therefore an important instrument in guiding production. This is why this study concentrated on supply relationships in cattle. Finally, it should not be forgotten that price changes also have an effect on consumption, an effect whiCh should be looked at, since for the products studied here demand is elastic in response to price, and also since changes in producer prices often have a limited effect on consumer prices, especially in cases where only a small share of the final value of the product consists of its farm value. Nevertheless, the effects, though limited in magnitude, are bound to be in the direction of improving the supply-demand balance. The 135 absolute price level of feedstuffs, for example, has an important impact on the price policy for beef and/or veal produced and consumed. Changes in relative prices of concentrates and of roughage influence the composition of compound feeds; in addition, the general level of feedstuff prices influences the extent to which livestock are grazed on pasture or fed on concentrated feedstuffs and can even govern the level of supply of livestock products. These various considerations underline the need for price policy both to take account of supply adjustment problems and to in- dicate the scope for effective action in this respect. A price policy can in practice be freely used to influence supply. The major problems of Greek meat production developments should be centered around supply response, i.e., the relationships be- tween output quantities and resource use and prices. Agricultural supply models may help in understanding these problems in that, through appropriate quantitative relationships, they may provide estimates of changes in output, acreage and yield per acre associated with changes in input use and in prices, which may be useful informa- tion in the hands of policy makers. Formulation of the Models Market supply relationships for agriculture as well as for other sectors of an economy are ceteris paribus relationships; that is these are the quantities that will be offered at the alternative prices ' with all other factors affecting supply held constant. Such factors can be prices of inputs, prices of other products that can be produced with the same resources, technology and the number of firms. 136 Thus, the supply relationship, like the demand relationship discussed in Chapter V, can be expressed by an implicit functional form such as in equation (6-1): 0S = f(Pi, P Te, E, N, c) j’ PO’ where: 05: quantity supplied f : function symbol P.: price of the commodity at hand P.: price of inputs P0: price of other products Te: technology E : expectations N : number of firms C : capacity of the firm (6-1) A priori knowledge of economic theory dictates that the follow- ing outcomes should be expected from this implicit formulation of a supply relationship. (6-2) These relationships in (6-2) will serve as a guide - based on knowledge .of economic theory - to understand and evaluate results from empirical analysis. The hypotheses which underline the supply response of farmers to changes in price are those of static economic theory where perfect 137 knowledge and foresight prevails in the perfectly competitive markets of both inputs and/or outputs. The individual farmer (and, by the aggregation procedure, all farmers) is usually free to vary the levels of both cost (cost of inputs) and output, and his ultimate aim is the maximization of profit. The total revenue of a farmer who sells his output in a perfectly competi- tive market is expressed in the number of units he sells multiplied by the fixed unit price he receives. His profit, then, is the difference between his total revenue and his total cost. The costs the same farmer is facing are represented as the units of inputs he buys in the input market times the unit price of the input. The empirical analysis of supply deals with the same products as the demand analysis did in Chapter V. These products are: feed- grain, roughage, beef, veal and milk. However, the supply analysis refers only to the farm level. Domestic production is calculated by multiplying the number of animal units (or land units) by the average yield per animal (or per land unit). Thus, if A stands for the number of acres planted, N for animal units (slaughtered and/or milked) and T for average yield per land and/or per animal unit, the production (domestic) is represented by the following identity: — O E A x T (6-3) 0' E N x 7 The identity (6-3) suggests that the quantity of the product produced could have been taken as a -dependent variable to explain the supply response in feed-grain production (direct estimation). 138 The procedure followed in this thesis (i.e., to break down the production equation into two steps) seems to be more realistic from the analytical purpose point of view. The direct estimation of the feed-grain supply response does not reveal the whole set of variables which contribute to the explanation of the variation in feed-grain production. On the contrary, the indirect method (a procedure which gives two separate equations with the area or“the aninel unit as one dependent variable and the average yield as the other dependent vari- able) provides the whole set of variables which contribute to the variation of supply response for feed-grain and/or for livestock pro- ducts. The same holds true for each one of the products being studied here, and the same approach has been followed throughout the entire analysis of supply response in this thesis. Thus, in an abstract way the identity (6-3) comes from the following two equations: A = fl(At_], Pt_], K], Ut) (6-4) Y = f2(Pt_‘|9 At, K2) K3, t], Vt) (6'5) and/or N = fl(Nt_], Pt_], K], Ut) (6-4') and I Y = f2(Pt_], Nt, K2, K3, t, vt) (6-5 ) 139 where: A: area in acres (stremmas) N: animal units, i.e., number of animals At-l: area in acres in t-l year price of the product studied in t-l year U : random disturbances Y: average yield per acre (land unit) or per animal unit. K],K2,K3: weather conditions in critical production periods. since 0 E A x V. O' 5 N x 7, then 0 = fl(At_], Pt_], K], Ut)'f2(Pt_], At, K2, K3, t, vt) and/or (6-6) 0 = fl(Nt_], Pt_], K], Ut)-f2(Pt_1, Nt, K2, K3, t, Vt) It can be observed that the weather variable enters into both equations, i.e., in equation (6-4) which gives the area planted and in equation (6-5) which gives the average yield per unit of land. The weather itself is a function of many other variablesz’3 but the analysis here is kept simple since otherwise things begin to get complicated and the data requirements (and availability) increase enormously. Direct vs. Indirect Method of Estimation As far as which method is better to use for the supply response of an agricultural product, it should be mentioned here that the 2L.H. Shaw, “The Effect of Weather on Agricultural Output: A Look at Methodology," Journal of Farm Economics (1964): 218-30. 38. Oury, A Production Model for Wheat and Feed Grains in France (1946- 1961). (Amsterdam: North Holland Publishing Company, 1966). 140 indirect method has been widely utilized in the empirical analysis con- ducted by agricultural economists. The direct method has two dis- advantages. First, the area planted is not entirely independent from the weather conditions prevailing during the sowing period. Second, the partial elasticity of the area planted with respect to the price of the product is not equal to the elasticity of the quantity with respect to price. These two elasticities could be equal if EAP (elasticity of area planted with respect to product price) were equal to zero. But such a hypothesis can hardly be supported. On the other hand, it is plausible to assume that the average yield depends - to a certain extent - upon the product price changes. From identity (6-3) it can be shown that Eop = EAP + Elp (6‘7) and E0P = ENP + EYP (6'7') and that E0P > EAP (6-8) and EOP > ENP (6-8') where: EDP: elasticity of (production) quantity produced with respect to price EAP: elasticity of area planted with respect to price ETP: elasticity of average yield with respect to price EDP: elasticity of quantity of a livestock product produced with respect to price of that product 141 E elasticity of the number of animals with respect to NP' price of the product produced from these animals. On the other hand, due to non-homogeneity of the production land (and/or the animal units), it could be the case that changes in the area planted (and/or in the number of animals kept on farms), assuming production resources per unit of land (and/or per animal) are given, result in changes in the average yield per unit of land (and/or per animal). This means that (6-8) could be written as: I V E E 0P AP A (5'9) 1 .Z E0P < ENP Empirical Estimation of Feed-Grain Supply In the category feed-grain the items of corn, barley, oats and rye are included. To attain a "common denominator," these four items have been expressed in Total Digestible Nutrients (TDN) weight accord- ing to appropriate conversion factors for each one of them. Foote, Klein and Clough,4 writing about the Demand and Structure for Corn and Total Feed Concentrates for the U.S.A. in 1952, presented a simple diagram showing the major economic relationships in the feed- livestock economy (see Figure 5 below). This simple diagram depicts nothing else but the factors which enter into the implicit supply re- lationship in (6-1). Thus, letters were added to the diagram to present 4R. Foote, J. Klein and M. Clough, The Demand and Price Structure for Corn and Total Feed Concentrates, USDA Techn. Bull., No. 1061, October 1952. . 142 A A—A Feed Prices,Pj Supplies 62 I _Feed, Q8» J ‘ ‘s' > ._ - - . ' , _—' L Animal Units ( Prices of Heat Ania-1 ch,AUF male and Livestock Products,Pi,Po JR "' "' """""'\ ' Technological 1 pfilfilflefle. Production of Livestock and Livestock Pro- Consumer ducts,Qo Incomes,Y Figure 5. The Major Economic Relationships in the Feed-Livestock Economy. (Arrows show direction of influence. 143 the factors which can be found in (6-1) in order to help the reader to have both a visual diagrammatic presentation and a kind of functional form in his mind at the same time. In any given year, supplies of feed-grain in Greece are cal- culated mainly from: a) the number of stremmas (acres) planted (and harvested) for feed-grain; b) yields per stremma (acre); c) stocks, which vary considerably from year to year, tend- ing to be large when production is relatively large, (and the storage program supported by the state is in opera- tion) and low when production is small (and the storage program is not in operation); d) imports which are always coming into the country since Greece belongs to a feed-grain deficit group of countries; 8) weather conditions, since year-to-year changes in yield depend mainly upon weather and level of fertilizer used plus some cultural changes. Cultural practices used in feed-grain production are intro- duced faster when prices are high, and, once they have been intro- duced, they tend to remain in use. «Current prices of feed-grain have little or no effect upon the total acreage used for feed—grain in Greece. The formulation of the supply equation for feed-grain de- veloped into two separate equations. The first one deals with the number of stremmas planted, and the second one, with the average 144 yield per stremma (l stremma = .24 acres). This break-down was made because the key variables are both the number of stremmas planted and the average yield per stremma. The final outcome of multiplying these two variables is the domestic supply of feed-grain (domestic produc- tion). The total supply of feed-grain existing at any point of time in Greece is expressed by the following identity: Total Supply of Feed Grain - Domestic Production + Beginning Inventory (Stocks) + Imports With this information as background, the following equations were tried in estimating the feed-grain domestically produced in Greece: NSTRFGPL = 3023.1811 + 0.7746NSTRFGPL (2542.0021) (.1576) + 27 181.1706FPFG t" (56 383.6908) t“ - 56 779.7982FPRt“1 + 5 080.0690FPFERT (106 419.5469) (14 709.9458) + 616.39030v - 3.7225K (199.4423) 9 (3.08118) R? = 0.84 6* = 1.93 (6-10) NSTRFGPL = 2598.9794 + 0.5935NSTRFGPL (2185.1469) (.1615) - 14 492.7857FPFG 1" (48 488.7291) t“ + 21 639.8038FPFERT + 300.9037DV - 6.2719K + 0.511AVF (15 175.3846) (225.9058) 9 (2.9397) (.2417) 'R2 = 0.91 6* = 2.26 ' (6-11) 145 The two models (equations) fitted to explain the variation in the numberiyfstremmas planted did not give satisfactory results in the sense that some of the variables included do not carry the sign specified by economic theory. The deficiency with these types of models is that they employ many variables, all of which greatly con- tribute to the explanation of the variation, though two of them - in each model - carry a wrong sign. The weather variable, K, taken as the rainfall in mm during the months October and November when the sowing of winter feed-grain is taking place, is statistically significant only in one of the equa- tions fitted, and its estimated regression coefficient seems to be fair. Its sign is inversely related to the number of stremmas (acres) planted. This is taken to mean that as the rainfall during that period increases, the number of stremmas planted is expected to be fewer, a situation verified by the weather conditions in Greece. The number of stremmas planted in the previous year appears to be a statistically significant variable since it picks up the (habitual historical) pattern of feed-grain production. The sign of the estimated regreSsion coefficient is positively related to the dependent variable, and its magnitude is rather fair. The variable farm price of feed-grain prevailing the previous year carries a wrong sign and is not statistically significant. This is due to model specification since another model, which is expressed below, yields a right sign and significance to the variable. Number of stremmas planted and the price of roughage prevail- ing one year earlier are inversely related, which means that as the 146 price of roughage goes up, the expected stremas to be devoted to feed- grain are expected to be fewer. Such a relationship is valid since both of these inputs compete for the same resources, mainly land. The other input variable, the farm price of fertilizer, appears with a wrong sign in both equations, is not statistically significant, and has a very large estimated regression coefficient. Price of inputs such as fertilizer of all kinds was kept at rather low levels during the sample period since the fertilizer industry has been heavily sub- sidized by direct and indirect government support. Thus, the cost of fertilizer does not seem to be a constraint to the number of stremmas devoted to feed-grain production, at least over the sample period examined here. The subsidy variable used in the form of a dummy variable seems to be statistically significant with a right sign and with a consistent and significant magnitude of the estimated regression coefficient. This subsidy variable refers to the subsidy given to feed-grain growers in 1963-64 and beyond as a unit payment for the amount of output pro- duced. Indeed, Greece reached - for the first time in her long history the level of self-sufficiency in grain for human consumption only late hithe 19505.1 Due to such an attainment along with an inability to compete in the world feed-grain market, farm policy has been oriented towards the increase of feed-grain production. Another, simpler formulation which seems to be more promising includes only two explanatory variables in the lagged form of t-l. The lagged variables are FPFG and NSTRFGPLt_1. Such a formulation t-l provided the following equation: 147 NSTRFGPL = -5062.2144 + 110.5514FPFGt_] + 0..8451INSTRFGPLt_1 (1043.1211) (34.2270) (.1824) _.2 , 'k R = 0.91 d = 1.74 (za) (6-12) In equation (6-12) each variable included is statistically significant and bears the sign expected from economic theory; also, each one carries rather important economic significance in explaining the variation in the area planted with feed-grain. The second part (element) fo the supply equation was examined by trying the following equation: TFG= -107. 8157 + 14. 3567T + O. 5633K + 16256.173FPFGt (93.1408) (2. 0012) (.4538) (9104.1) 1 _2 * R = 0.84 d = 1.60 (6-13) Equation (6-13) explains the variation in the average per stremma yield in feed-grain production in terms of a time trend, weather conditions during the October-November period and the price of feed- grain the previous year. The model gives rather satisfactory results in the sense that all but one variable (the weather variable) are nearly statistically significant, and all of the variables included bear signs expected from economic theory. The time variable is used as one of the explanatory variables to pick up whatever "technological change" has taken place in feed- grain production. Such change could involve better seeds used, the utilization of pesticides, weed control, the use of fertilizers, and $0 011. 148 To tackle the problem of variation in the dependent variable in terms of a regular movement over time, two alternatives of action are open.5 The first is to find the relationship between the de- pendent variable and the time and subtract this component of the varia- tion before going on to consider the relationship between the dependent variable and the independent variables. The second alternative, to include time as one of the independent variables, has been followed here so that the correct number of degrees of freedom could be allowed for in carrying out tests of significance. A more simple formulation here should be to use only the T variable and the farm price of feed- grain variable lagged one year. As far as the weather variable is concerned, better results could have been obtained if the April-May rainfall could have been used instead of the October-November rainfall. In fact, the April- May rainfall plays a more crucial role in determining yield in winter feed-grain production. Such data, however, were not available to the author at the time of the analysis. In what follows the quantity of feed-grain domestically pro- duced is used as the dependent variables to test how much the empirical analysis lends support to the economic theory expressed in (6-9). QFGS = 611.2921 + 0.5599AUF' (200.2010) (.0912) 'R2 = 0.80 (6-14) 5James Thomas, Notes on the Theory of Multiple Regression (Athens: Center of Economic Research, 1976), p. 122. 149 QFGS = -1 214.2439 + 151 034.9085FPF6Pt_1 + 0.5746AUF + 66.77620v (232.4241) (49 514.6699) (.1443) (236.7308) 9 _ * R2 = 0.81 d = 1.39 (i) (6-15) QFGS = -1 296.8261 + 0.619AUF + 152 090.4849FpF6Pt_1 (622.3847) (.0956) (49 146.1364) “R2 = 0.70 6* = 1.34 (i) (6-16) QFGS = 1203.4458 + 153 150.2436FpFept_1 + 46.7720K + 0.6201A0Ft_1 (82.9034) (80.4260) (15.2231) (.0440) + 25.004 1 (8.1224) ....2 * R = 0.96 d = 1.66 (6-17) The striking finding here arises from equation (6-14) in which the variable animal units fed explains 80 percent of the variation in the variable quantity of feed-grain domestically produced in Greece. Maybe the quantity of feed-grain affects AUF. The explanatory vari- able is statistically significant and bears the right regression co- efficient sign, but its economic weight seems to be underestimated. The magnitude of the estimator of this variable seems to be very con- sistent on all three equations used. The price variable lagged one year appears to be statistically significant in both equations tried and has a positive regression co- efficient, which means that the higher the last year's price of feed- grain were, the more feed-grain farmers produce. The magnitude of the regression coefficient is large which verifies the situation of the real world as far as the feed-grain industry in Greece is concerned. Feed-grain is mainly growing on lands where other crops cannot be 150 competitive. Feedgrain production is also rather "easy" activity and, following mechanization, is carried out even by people living in the towns and/or cities and, even further, by people living abroad as emigrants. These people lease their lands to others or they them- selves sometimes take care of this farm activity. Feed-grain has - from time to time - been subsidized by the state, especially later in the sample period when the P.L. 480 program was not at work. The dummy variable used for the years when the sub- sidy was in effect shows a positive causal effect with thebe the case in economic theory and in practice under the conditions prevailing in this industry in Greece. Thus, from policy point of view it is safe to conclude that the livestock numbers existing and fed in any year t, the last year's price of feed-grain and a subsidy variable are the most relevant variables in the feed-cattle economy influencing domestic production of feed-grain. The weather variable and the time variable used as_a sub- stitute for the technological change taking place in the industry should increase ‘Re. Such a model resulted in equation (6-17). 151 From the last equation fitted (6-17) it can be concluded that the previous models give rather satisfactory results since they explain almost 70-80 percent of the variation in the quantity of feed-grain domestically produced without taking into account any explicit con- sideration of the weather and the terminology. The weather variable, at least in a short-run context, undoubtedly plays an important role in the average yield per land unit of feed-grain and, hence, in the quantity of feed-grain produced. Elasticities of Supply for Feed-Grain It should have been noticed by now that the statistical estimates are rather satisfactory up to this point as far as both statistical and economic significance is concerned. With regards to the signs of the estimated parameters, economic theory helps to con- clude what estimator carries a wrong or right sign. And, as far as 'size of the estimated coefficients of the parameters is concerned, due to the lack of a priori knowledge the results should be based on a) the degree of significance of the parameters, b) their consistency in the different formulations (equations) tried in cases where there was more than one alternative model and c) the findings in other studies both within the same country and/or abroad. The calculation of different elasticities is of help to the interested researcher and to the policy maker. Thus, in Table 17 below these estimated elasticities are given. Table 17 indicates that the parameter of the farm price of feed-grain lagged one year carries a wrong sign and that it is not statistically significant in the formulation which uses as a Va llor Ver; 152 TABLE 17 F ESTIMATED ELASTICITIES OF SUPPLY FOR FEED-GRAIN Equation Dependent Expanatogy Variables var'ab'e NSTRFGPLt_] pr6t_1 FPR t_1 AUF 6-10 NSTRFGPL +0.77 20.08 -0.08 -, 6-ll NSTRFGPL +0.59 -0.04 - +0.70 6-12 NSTRFGPL +2.03 -0 000 . - - 6-13 TFGS - ‘ +1.24 - - 6-14 FGS - +0.21 - +0.76 6-15 FGS - +1.96 - +0.78 6-16 FGS - +1.98 - +0.85 6-17 FGS - +1.99 - +0.85 dependent variable the number of land units planted with feed-grain. It is suspected that this non-significance is due to model specifica- tion. The elasticity of number of land units planted under feed-grain with respect to the farm price of feed-grain prevailing the previous year is very small. Indeed, the data reveal that there were no great differences in the number of acres planted from year to year. In- stead, the increase in feed-grain production resulted from a more in- tensive farming method adopted by feed-grain farmers. Indeed, when this elasticity is calculated from equations which use as a dependent variable the quantity of feed-grain, it is higher, more plausible and more consistent. Generally, it can be observed that EAP < EOP’ which verifies what was said about this relationship on p.140- 153 This elasticity relationship can be viewed as an indica- tion that the market prices serve as a stimulus in farmers' decision making behavior for them to produce a certain quantity of feed-grain on their farms. The same quantity, of course, can be produced not only by changing the number of acres (strenmas) planted with feed- grain, but also by changing the average yield of land under feed- grain. In some parts of Greece this last alternative is the only alternative open to farmers who are facing the severe land constraint problem. The elasticity of both number of acres planted with feed- grain and quantity of feed-grain produced with respect to animal units fed seems to be consistent in both types of models. A positive sign and such a magnitude verify the fact that animals existing on farms need to be fed at a maintenance level regardless of economic justifica- tion. A It has been found elsewhere that for a dairy cow which yields 1000-1500 kg of milk per year the land which is required for develop- ment under barley-feed is 0.59 stremmas.6 Since the average yield per milk over the sample period was 1,004 kg per year, the results found here seem plausible. The price r0ughage elasticity seems very low and should be used with great caution. 6Kitsopanidis, p. 8. 154 Empirical Estimation of Roughage Supp1y_ Under the classification roughage the main fodder plants for hay are included. These plants are the following: barley, oats, vetch, peas, alfalfa, grass cut for hay, bitter vetch, vetchling (lathyrus), lentil, maize, sorghum and marigolds. As in the case of feed-grain, these fodder plants have been converted into Total Digestible Nutrient weight in order to arrive at a common denominator. Hay and pasture production heavily depends on weather condi- tions during the critical sowing period and in the spring period when rainfall is rather a scarce good in Greece. The price and level of fertilizer used are among the variables which contribute to the ex- planation of the variation in roughage production. Soil conditions and types of soil are also contributory factors. The same kind of supply functions as those used in deriving the supply functions for feed-grain was employed here. From an economic point of view, variables expressing com- petitive products should be included as well in the formulation of the roughage model. Thus, wheat production should be among these com- petitive parameters since wheat production competes for the same resources as roughage production. A price ratio of wheat price to (roughage price should have taken care of that problem, but this has not been tried in this thesis. I Furthermore, a problem is raised here as far as the model formulation is concerned when alfalfa is included in deriving the quantity of roughage produced. As it is known, alfalfa (and other perennial plants) is on the ground more than one year, and it is 155 (usually) planted every five years. Thus, a time Lag element enters into account when the number of acres (stremmas) planted under alfalfa is used as a dependent variable. {The area planted with alfalfa in year t is the result of decisions made in t-i time periods prior to t period. The t-i (i = 1,...,5) decisions refer to the prices of alfalfa held over these t-i periods or, better, over the t-5 period when the actual area of alfalfa was planted. This time lag problem could have been overlooked, the re- sults of the empirical analysis would have then been faulty. But through use of the parameter QRPt_] the cumulative effect of the t-i time periods are hopefully captured. Equation (6-18) does not, of course, capture the whole cumulative influence on the number of acres planted under alfalfa in year t, but it does capture the change in the number of acres. To explain the variation in the quantity of roughage produced, the following equations were tried: NSTRRPL = 88.5619 + 0.9428N5TRRPLt_1 + 66 347.1417FPRt_1 (.1178) (86 426.8533) - 61 328.5113FPF6Pt_1 + 38.13580v + 2.2163K (66 589.0535) (136.7057) 9 (2.7604) ‘R2 = 0.92 6* = 2.41 (6-18) NSTRRPL = -2458.0452 + 1229.4023FPR + 1.1434NSTRRPL _ (1040.5214) (423.8201) t" (.2542) 1" * R2 = 0.98 d = 2.89 (6-19) 156 YRSTR = 242.4423 + 0.347OQRP + 0.7057QFERT + 15.02971 (31.6997) (.0574) (.5044) (6.4438) R2 = 0.94 6* = 1.62 (6-20) QRS = 3335.4018 + 81895.3537FPR - 35 636.326OFPFERT (1279.2312) (23971.0578) t-] (6 015.8899) - 352.7713DVr - 2.3029K (124.1469) (2.13) R2 = 0.88 6* = 1.24 (6-21) QRS = 1201.4080 + 73 520.1313FPRt_] + 0°8792QRSt-1 (402.6275) (17 133.9801) (.0928) R2 = 0.98 6* = 1.86 (6-22) Equations (6-18) to (6-22) illustrates the models fitted to explain the variation in the quantity of roughage supplied in Greece. The first two models, (6-18) and (6-19) use the variable number of acres (stremmas) planted under roughage, while the equation (6-20) gives the average yield per acre (stremma). Finally, equations (6-21) and (6-22) use as their dependent variable the quantity of roughage produced. Thus, a comparison can be made again as to which method gives more satisfactory results. The first model uses many parameters which may contribute to an explanation of the variation in the number of acres planted under roughage in Greece. The variable number of acres (stremmas) planted in t-l year bears the right sign and seems to be statistically significant, and its estimated regression coefficient is rather fair. This variable picks up the habitual pattern of farmers who are involved in the. 157 roughage production business, and a coefficient of a magnitude of 0.94 means that this habitual pattern will be followed by the farmers in the future if other things do not change. The own-product price, i.e., the farm price of roughage received by the farmers one year earlier, bears the right sign in both models and carries great economic significance, and statistical significance in the (6-19) model, though not in the (6-18) model. This has to do with model specification and the variables included in the model. In the (6—18) model the weather variable is included among the explanatory variables of the model and, along with the other two main variables, the farm price of a competitive product (such as feed-grain) and a subsidy variable (DVr) given to the alfalfa growers during the period 1965-1972, picks up some percentage of the explanation of the variation in the dependent variable in that it may cause the variable farm price of roughage in t-l year to behave differently in the two models in terms of its significance. But in both models this variable carries the right sign and fair economic significance. A The price of feed-grain in model (6-18) carries a right sign in the sense that as feed-grain becomes dearer, a greater number of acres of roughage is expected to be planted by the farmers. This parameter is not statistically significant in such a formulation, and the size of its estimated regression coefficient has been rather over- estimated. The subsidy variable used in the form of a dummy variable carries a right sign in the sense that whenever the subsidy is given, farmers are expected to increase their area under roughage production, 158 although the magnitude of the estimated regression coefficient is rather low and the parameter seems not to be statistically significant in this formulation. The insignificance of the subsidy variable in- dicates that since the demand for roughage is a derived demand, the final outcome on supply of roughage at its long-run equilibrium point should depend upon the price(s) of the livestock products produced, in the production of which roughage is used. Model (6-19) uses the number of acres (stremmas) planted under roughage as a dependent variable which is explained in terms of the price of roughage prevailing and in terms of acres planted under roughage in t-l year. There is, of course, a high degree of similar upward trend in the parameters used in this formulation, and the re- sults should be used with great caution. Both variables used in the (6-19) model carry a right positive sign and are statistically significant, and the magnitudes of their estimated coefficients carry a considerable economic weight. The corrected coefficient of multiple determination is high enough (0.98), but this is due greatly to high intercorrelation between the parameters NSTRRPLt and NSTRRPL used in the model. t-l Equation (6-20) explains the variation in the average yield per acre (per stremma) of roughage production. The variable QRP, i.e., quantity of roughage produced in year t, is taken as a proxy variable for the weather parameter. The variable is closely correlated with the dependent variable which may contribute to the high value of R2 (R-2 = 0.94) Thevariable quantity of roughage produced is positively re- lated to the TRSTR variable which means that as the quantity of roughage 159 produced increases, the average per acre (stremma) yield of roughage is also expected to increase, although this last increase is equal to 0.34 of the increase of the QRP. This positive relationship should be taken to mean that the aggregate increase in roughage production (in the future) - other things being constant - is expected to come not only from an intensive farming system of roughage production, but also from an increase in the number of acres (stremmas) devoted to roughage production. Related to this problem of roughage production is the use of fertilizer and other technological improvements which are embodied in the time trend variable T which is included in the equation. The fertilizer parameter indeed contributes to the explanation of the variation in the average yield per acre (stremma) of roughage produc- tion. The fertilizer variable carries a positive sign which means that the higher the level of the quantity of fertilizer per acre (stremma) is, the higher the average yield of roughage is expected to be. In another formulation where the October-November rainfall was used the fertilizer variable either came out with a wrong sign or it was not statistically significant. The time variable in the same model (6-20) appears to be statistically significant and carries a right sign, and its regression coefficient weight seems to be valid. From the data used it is revealed that, indeed, the average per acre (stremma) yield in roughage production almost doubled, but so did the number of acres (stremma) under roughage production. However, the total quantity of roughage produced during the sample period went up almost five times. 160 This fivefold increase in total production of roughage is in part due to more fertilizer used, more irrigated areas used, better quality of land and/or seeds used, etc., factors embodied in the parameter of technology T. The significance of the time variable further means that investment projects which are related to roughage production - even if they are small-scale in nature - have to seriously be con- sidered and promoted all over the country. The models (6-20) and (6-22) use as a dependent variable the quantity of roughage produced. Here, again, the farm price of roughage prevailing in t-l year is positively related to the quantity of roughage produced, which means that the higher the price of roughage is, the more the quantity of roughage to be produced is expected to be. The price parameter is statistically significant, and the magnitude of its estimated regression coefficient seems to be valid. The significance of this parameter (both the economic and the statistical) does mean that farmers who grow roughage in Greece do respond to market price changes in Greece. The fertilizer variable used in the form of the price of fertilizer appears to be statistically and economically significant, and it seems to carry a right sign. The magnitude of the estimated regression coefficient seems to be rather high, but, in general, due to the fact that the quantity of roughage production depends heavily on the°leve1 of fertilizer used, the empirical analysis apparently supports that a priori knowledge. The dummy variable used to represent the subsidy variable, although statistically significant, carries a wrong sign which leads to some reservations as far as the model specification is concerned. 161 The weather variable appears not to be significant as it was used here (rainfall during October-November period). Rainfall plays a more important role for roughage production during spring and summer time. Finally, the last model (6-22) uses lagged variables, and it appears to give significant results. But the problem of correlation has to be taken into account when this model is utilized. Both vari- ables, the FPR and the QRSt_], are statistically significant, bear t-l the right sign (being positively correlated to the QRSt) and carry significant economic weight. The corrected coefficient of multiple regression, R2 = 0.98, is high enough, but it seems that the serial correlation problem contributes somewhat to that. In closing this section it can be observed that, in general terms, the econometric models do contribute to quantifying the supply relationships in the supply schedule for roughage in Greece. Elasticities of Supply for Roughage Producers of roughage do not respond to price changes in an unknown mechanical (or imaginary) way. Prices play a central role in guiding production. This is not, of course, taken to mean that pro- duction decisions of farmers are governed solely by prices. Government programs, the limits set by climate and soil and the availability of equipment (capital in its broad sense) obviously exert a strong in- fluence over what farmers plant each year. Since the Greek producer (farmer) is a profit maximizer, the view which is taken here is that prices, especially relative prices, influence his behavior. 162 Knowing the elasticities of supply - in a quantitative aspect - one has an idea of what to expect if certain parameters change while. others are kept constant. Thus, with such a view the elasticities of supply for roughage can be calculated and are given below in Table 18. TABLE 18 SUPPLY ELASTICITIES FOR ROUGHAGE Equation Dependent Independent Variables var'ab'e FPRt_1 FPFGRt_] 6-18 NSTRRPL +0.25 -O.48 6-19 NSTRRPL +0.005 - 6-21 QRS +0.31 - 6-22 QRS +0.27 - Table 18 above reveals that the supply elasticities of roughage calculated from the direct—approach based models is more consistent and greater than the elasticity calculated from models based on the models which take into account the indirect method, although in this particular case here one model (6-18) of the two that use the NSTRRPL as a dependent variable gave as high an elasticity as 0.25. The supply elasticities of roughage and of feed-grain calculated from both models offer clear evidence that the elasticities which were calculated from models which used the quantity variable (as the de- pendent variable) rather than the number of acres (stremmas) planted variable are better than those coming from the second type of model,. in the sense that they are more consistent and look more plausible. 163 Empirical Estimation of Beef Supply Response- In this section the results for beef supply which were ob- tained from the empirical analysis are presented. This empirical analysis seems to be of great significance to all segments involved in the production of livestock products, especially to the policy makers. In the demand analysis it was shown that beef is considered price and an income elastic good in Greece. This should be taken to mean that expansion of beef production shbuld be welcomed in Greece. Thus, from a farm policy point of view it should be very important to know the supply elasticity for this livestock product, given the fact that the encouragement of beef production in Greece is the main aim of Greece's farm policy. It was asserted earlier in this thesis that Greek farmers are assumed to be profit maximizers. This is especially true for beef and veal producers who for the most part produce for the market despite their habitual inertia, the structural constraints and the economic uncertainty which prevailed over the sample period. Thus, the ex- pected price seems to be the determining factor in guiding the risk 'taken in beef production. Lack of more data constrained model specification, but in general terms both the "direct" andgthe "indirect" methods were used and distributed lags were tried to test the actual production process. To explain the variation in the number of cattle going to slaughter and the quantity of beef supplied the following models were tried: 164 BCSL = 26.3955 + 0.8999BCSL + 0.2113(FPB/FPFS) _ (27.5236) (.1513) 1‘1 (.8051) t 1 R2 = 0.70 6* = 1.95 (6-23) 78C = 69.9434 + 16.98886ovb + 5.74291 (6.7350) (8.7328) (.5046) R2 = 0.87 6* = 1.01 (za) (6—24) _ FPBt_] + FPBt_2 085 = 14.8231 + 12.7128 ( 2 ) + 2.4817A0Et_1 (2.4252) 5.6274 (.7922) 'R2 = 0.96 6* = 1.64 (6-25) OBS = 20.4089 + 1.1278085 - 0.3182FPFS (6.3018) (.1412) 1‘1 (.0215) 1'1 R2 = 0.98 6* = 1.67 (6-26) The first of the above equations (6-23) deals with the number of cows slaughtered and the second one (6-24) with the average yield, in carcass weight, of beef per head of cattle. This break down was made because the key variables are both the number of beef cows on hand and the average yield per cow. This type of formulation, as in the case of feed-grain and roughage production, gives more insights and presents the whole set of variafiles which contribute to the ex- planation of beef supply. Some equations expressing the direct method were also tried in order to obtain both models' reaction. Furthermore, Almon's distributed lags were tried in order to explain the variation in the number of beef cattle slaughtered over time. By using this last method a more 165 dynamic picture of the beef production industry can be obtained, and it is hoped that this should contribute towards a better understanding of the beef production process and, hence, of the impacts of policy measures taken towards increasing beef production. The introduction of the variable "cattle population" either as the number of beef cows slaughtered in t-l year or as the number of cows milked and/or number of calves slaughtered in the prior year, seems to be verified by a priori knowledge that there exists a certain relationship between the quantity of beef meat produced and the number of animals existing on the farms or the number of animals slaughtered the previous year. This relationship can be disturbed at a certain point of time either by factors affecting the meat production directly, such as animal diseases and weather, or by factors which influence meat production through the price mechanism which affects production and number of animals existing by making or not making the business profitable. The variables quantity of feedstuffs and price of feedstuffs were tried in a couple of alternative equations, but the results were (rather poor. These two variables are related to each other, but they 'present different courses with regards to the number of animals slaughtered. The variable quantity of feedstuffs produced repre- sents the existing conditions in the industry for feeding the existing cattle on farms, while the second variable, price of feedstuffs, in- dicates the cost side of the business. The cost structure of a firm or of an industry plays an important role in deciding that firm's or industry's competitive position, and, hence, that cost structure affects the producers' (farmers') decision making process. 166 A priori knowledge does not provide an idea as to what sign to expect for these two variables from the empirical analysis. At the beginning it seems rather obvious to expect a positive sign for the first variable, i.e., the quantity of feedstuffs, since as more feedstuffs are produced, more cattle are expected to be fed, and, hence, more cattle are expected to be slaughtered. 0n the other hand, a negative sign should be expected for the variable price of the feed- stuffs since the dearer feedstuffs are, the lesser quantities of them will be demanded and fed to cattle, and, hence, the fewer cattle will be slaughtered. However, the opposite could be expected as well, in the case where feedstuffs are scarce and their prices are high. In reference to Greece, it can be observed that it generally rains only in the winter months, from November to February, when low temperatures are unfavorable to the growth of grass. From May to September, on the other hand, there is little or no rain at all. This uneven distribution of rainfall is one of the main reasons for the premature slaughter of calves because of the inadequate summer fodder supplies and the price falls which occur in early summer when the farmers are obliged to sell. Feed cost is considered as a variable cost and the equilibrium output of the industry in the short-run is based on this cost. This does not mean, of course, that the level of the fixed cost of the in- dustry does not have any significance in the analysis of the short- run maximization of profits. Each firm, and, hence, the industry as a whole, faces discontinued production and a loss equal to its fixed cost. The firm or, for that matter, the industry, will produce at 167 a loss in the short-run if its loss is less than the level of its fixed cost, i.e., if industry's (firm's) revenue exceeds total vari- able cost. The weighted prices of the last two years for beef were ‘ taken in the form of their simple average in one of the "direct" equations since it was believed that these prices are the most recent ones in farmers' memory when they compare the beef price at year t when they slaughter or sell beef cattle. The variable animal units in t-l period is, of course, related to the quantity of beef supplied at year t, but it is of great im- portance to know that the level of production of other meats plays a considerable role in the level of beef production. Thus, from a policy point of view the results here could be used as an indicator for the program of substitution of beef meat for other meats and, especially, poultry meat. In general terms, models (6-23) to (6-26) satisfactorily ex- plain the variation in the number of cattle slaughtered and/or average carcass yield per cattle head and/or the direct beef meat supply since over 79 percent of these three variations are explained by the corre- sponding models. The parameter beef cattle slaughtered in t-l year seems to contribute to the explanation of the variation of the cattle going to slaughter (slaughtered). Indeed, there is a direCt relationship be- tween the two variables since the number of beef going to slaughter this year was the same as that the year before, the only difference be- tween the two variables being the calves left for fattening from t-l year. 168 The lagged beef-feed price ratio seems to be positively re- lated to BCSLt, although it does not appear to be statistically significant, and its economic meaning is not important enough. The data reveal that the ratio was in favor of the beef producers since it was increased by almost 75 percent over the sample period. The fact that it is not statistically significant added to the fact that its economic weight is not great should be kept in mind when this model is used. The explanatory power of the model is 70 percent and, con- sidering the model's simplicity, that figure is considered to be high. Seventy percent of the variation in the number of cattle slaughtered is attributed to the parameters BCSLt_1 and (FPB/FPFS)t_]. The model (6-24) explains the average yield in carcass beef produced by a cattle head. One of the two variables included in the model is a dummy variable (DVb) used to take into account a subsidy given to beef producers to raise animals weighing more than 250 kg liveweight for the years 1963-1968. The variable seems to be almost statistically significant and carries a positive sign which means that the higher the subsidy is the higher the average weight of beef cattle is expected to become. Given the administrative difficulties involved in applying a subsidy scheme, it seems that when the subsidy is ad- ministered in a "good way," it does contribute to the increase of beef meat per animal head. The second variable used in the (6-24) model is a time variable which picks up the improvements which had taken place in the national herd. Indeed, the major developments in the cattle herd were: 169 (a) increased calf drop percentage, (b) decreased death losses, (c) im- provements in the quality aspects of the national herd by adoption of the artificial insemination technique and (d) increased number of cattle fed in barns in a more intensive way. All these developments resulted in an increase in the average per animal dressed weight of beef production in Greece. Indeed, the time variable appears to be statistically significant and carries a positive regression coefficient which means that the more improvements adopted for use by the farmers, the greater the average meat yield per animal is expected to be; this is verified by the data observed over the time span examined here. The economic significance of the time variable is high enough since for one unit of change in the time variable, there are almost six units of change in the dependent variable 780. 1 Models (6-23) and (6-24) combined explain the variation in the supply of beef and illustrate in a more analytical way the parameters which influence beef meat production in Greece. Models (6-25) and (6-26) explain the variation of the quantity of beef produced in terms of: (a) the farm prices of beef received by farmers in the last two years (these prices are weighted prices and a simple arithmetic average mean of these two prices has been used here), (b) the number of animal units fed on farms in t-l year, (c) the quantity of beef produced in t-l year and (d) the farm price of feed- stuffs in t-l year. The two models reveal high corrected multiple re- gression coefficients (R2) and parameters which are all statistically significant, carry a right sign and bear significant economic meaning. .170 The size of the herd seems to have quite a considerable role to play in specifying the amount of beef produced.. In addition, feed cost and product prices do play a serious role in beef production for reasons explained above in this section. Beef producers produce particularly for the market and are more sensitive to changes in price of input and price of product taking place in the market. The most serious expense in the short-run span for beef producers is feed expenses, and the producers reaction is precipitated by changes in feed cost. From the other side the change in product prices (Farm Price of Beef) is the only price which dictates slaughtering behavior. A general observation verified by the empirical analysis is that the higher the product prices, the higher the production of beef will be, and the higher the price of feedstuffs, the lesser the quantity of beef will be produced by the farmers. Studies in other countries have shown that beef cow operations are a low profit business and can be viable in places where there is a large amount of underutilized roughage which canbe used by beef cows at a very low cost. Trimble7 talks about the unprofitability of an in- vestment in a beef cow in the United States when both fixed and vari- able costs are included, and he recognizes that costs are involved in input and product markets. A substantial change in the price of both of them has a great impact on farmers' expectations and, hence, on their 7Richard L. Trimble: "An Economic Analysis of the Effect of Monetary Policy on the Beef Industry," Ph.D. Dissertation, Michigan State Univer- sity, East Lansing, Michigan, 1973. 171 decision making process regarding investment. The costs may be due to: a) biological uncertainties which refer to physical and/or biological aspects of production that affect costs and investments; b) unforseen uncertainties related to unexpected diseases, droughts, floods, etc., and finally, c) uncertainties related to the whole economic environment of a country and/or international trade -- in the case of an open economy like that of Greece's. Such un- certainties may be the inflation rate, the rate of un- employment, the monetary and/or fiscal policy of the government, the level of interest rate, etc., all of which make the cattle feeding business an unusually high-risk enterprise which may partly explain why Greek farmers have not invested enough money in it. Farmers look at two things: The first is a "feeding margin" defined to be the difference between the feed cost per kilogram gained and the price received from the grain fed to the cattle. This margin does not involve the land and other fixed factors used to support the cow. The most relevant decision made by farmers seems to be the addition of a cow to an existing family herd, or the substitution of a beef cow herd for an enterprise that uses the same fixed resources. Trimble's data point out that investment in a beef cow will generate revenue sufficient to cover all fixed and variable costs and provide a return on invested capital equal to the firm's cost of capital only if relatively high calf prices and low cost of capital exist. 172 The second thing examined by farmers is a "price margin" which is defined to be the difference between purchase cost and selling prices per hundred weight or per animal unit in case of live animal transactions. A decrease in market price which cannot be predicted in advance by farmers can be disastrous for those who buy improved type of cattle and then must bear the burden of a negative price margin. To make things clear two points need elaboration here. first, a negative price margin does not necessarily indicate a loss, as it may be more than compensated for if liveweight gains are a high pro- portion of final sale weight and the feeding margin is favorable. Segpnd, positive price margins may not mean that the farmer makes a profit if they are offset by a poor feeding margin. However, neither of these margins reflects investment costs (which also affect net re- turns of cattle feeding) since the Greek farmer does not keep very good records and accounts. Price Elasticities of Beef Beef-feed price coefficients in the lagged formulation of the models appear to be positive, while feed cost coefficients in these equations appear to be negative. Supply response analysis is particularly needed currently in Greece because of the severe problems which production of meat con— tinues to face in adjusting supplies to market demands and in sub- stituting foreign imported meats for domestic ones. In addition, the balance of payments problem which Greece is facing dictates possession of such knowledge. Furthermore, such an analysis will give-some positive direction to and shed some light on the confusion 173 existing today in Greece with respect to the causes of these dif- ficulties. Supply elasticities presented in Table 19 below were found to be positive for the beef-feed price and negative for the feed cost parameter. In the (6-23) model, for lack of a better indicator, the TABLE 19 PRICE ELASTICITIES 0F BEEF Equation Dependent Independent Variables Variable (FPB/FPFS)t_] (FPBt_1+FPBt_2)/2 FPFSt“1 6-23 BCSL +0.02 - - 6-24 085 — 1 +0.48 - 6-25 085 - p - -0.62 production response is presented by the number of beef cattle slaughtered. It seems that, although the sign is according to what was expected from economic theory, the size of the elasticity seems to be rather low since, for a 1 percent change in the beef-feed. ratio there is a 0.02 percent change in the number of beef cattle going to slaughter. The low short run supply elasticity of beef cattle slaughtered with respect to beef-feed price ratio in the previous year may have to do with decisions made in more than one period and with expectations concerning present price that were formed during several previous years. The response of beef production to price changes in beef taken as a simple arithmetic average of t-l and t~2 prices of beef was 174 found to be equal to +0.48. This means that 48 percent of the quantity of beef supply response is positively associated with the average weighted prices received by farmers in the last two years. A -O.62 feed cost elasticity of beef supply was found here and calculated from model (6-25). This is taken to mean that for a 100 percent change in the feed cost upwards a 62 percent downward response should be expected in the production of beef. The rest of the supply response seems that it is contributed over to other time periods. Reuttinger observes that "most agricultural economists who have their hands into estimating beef supply functions have observed zero or negative elasticities of beef output with respect to beef price (and similarly, zero or positive elasticity of beef output with respect to feed price."8 He presents supporting empirical evidence that this is due to two reasons. First, cows and heifers have a dual function in that they can be slaughtered to produce meat or they can be retained to build up inventories. Second, there is an aggregation error because the components of beef supply, which are steers, cows and heifers, are not examined separately. It seems that each component has its own supply response. It is thus observed here that past prices do have to contribute to the formulation of price supply response in farmers' behavior as far as both the number of cattle to slaughter and/ or to produce beef is concerned. 85. Reuttinger, "Short-Run Beef Supply Response," American Journal of Agricultural Economics 48 (November 1966): p. 909. 175 The low product price elasticity of supply (+0.48) provides evidence that beef production in Greece has to overcome some "technical problems" (structural and/or better improved animals to be used) if it is to compete with other industries within the country. The fact that the price elasticity of supply of beef with respect to farm price of feedstuffs was found to be greater than the own-price elasticity of supply should probably be taken to mean that the supply of feed- stuffs is the most crucial parameter in determining beef supply (pro- duction) in Greece. This is related to the above mentioned problems of structure of the beef industry in Greece and to technological de- velopments taking place within the industry as well. 176 Empirical Estimation of Veal Supply Response As in the case of beef, the supply of veal is analyzed here in terms of two equations. The first equation examines the factors affecting the number of veal calves slaughtered, and the second equation considers the factors that mainly influence the average carcass yield of finished veal per calf.. Veal Calves Slaughtered (VCSLt) Equation Each of the meat animal species has a somewhat different supply elasticity. In the case of veal, it seems that the number of cows milked on farms on December 31 (in year t-l) and the veal-feed price ratio in year t-l mostly affect farmers' decision to slaughter their calves. Veal calves slaughtered includes slaughtering of calves born and raised on farms and slaughtering of calves which have been im- ported into the country, fed for a period of time and then slaughtered. Veal, as distinct from beef, is produced from calves up to 24 months of age and younger. Male and female veal calves tended to be marketed in a rather stable fashion over the sample period in Greece, although some dif- ferences were noticed during the period's last years. Veal pro- duction was a complementary farm activity to the whole farm business, a fact which complicates the decision-making process of Greek farmers. In the short-run, veal producers may speed up or hold off the slaughtering of veal-calves in response to changes in prices of veal and/or beef. In the longer run period of time, changes in veal 177 production may also occur as a result of variation in price ratios. In an annual marketing period, however, it is not obvious that significant response in marketings of veal can be produced. Hence, the specification of the equations which follow is influenced by these considerations. Veal Calves Slaughtered VCSLt = -136.7922 + 0.8916NCMt__1 + 0.6 97(FPV/FPFG)t_1 (65.644) (.204) (.2 6) R2 = 0.87 6* = 1.89 (6-27) Dressed Carcass Weight per Calf ivc = 25.3818 + 43.35690Va + 5.9330 T t (7.785) (10.771) (.864) -2 'k R = 0.97 d = 1.78 (6-28) Supply of Finished Veal QVSt = 15.2263 + 0.7537QVSt_] - 0.6824FPFSt_] (5.179) (1.742) (.179) -2 * R = 0.96 d = 1.87 (6-29) The results for the veal supply-submodel, in general, were satisfactory for all equations as measured by significant regression coefficients, high coefficients of determination and consistent signs of various parameters. None of the equations in final form showed a. clear presence of serial correlation. 178 The slaughter function was estimated as a relatively simple relationship. It incorporates cows milked - excluding calves - on farms on December 31, in year t—l, and the veal-feed price ratio. The former variable was highly significant statistically, positively correlated with the VCSLt and had a sign expected from a priori knowledge.1 The existence of high inventories of animals in pre- ceding time periods had a positive effect on slaughtered number of calves in the current time period. The veal-feed price ratio yielded a positive coefficient of +0.65, indicating that farmers positively respond to any veal-feed price ratio. In other words, whenever veal prices increased by 1 unit, veal cows going to slaughter tended to increase by 0.65 unit. The weight to which veal calves were fed before slaughter was highly determined by factors such as the quality improvement of the national herd - embodied in the time variable employed - and a subsidy program variable which subsidized farmers who grew feed- stuffs on their farms to feed veal calves. Both of these variables yielded significant regression coefficients, consistent signs for their parameters and a high degree of determination. The model giving the supply of finished veal was estimated as a rather simple relationship of the quantity of finished veal in the preceding year and the cost of feeding the calves in the pre- ceding year. The former variable was assumed to represent the habitual elements of practices by farmers who were engaged in veal production. The latter variable represented the cost structure and position of the veal production industry. Both variables yielded 179 statistically significant results with consistent signs and signif- icant economic weights. Price Elasticities for Veal Estimated supply elasticities with respect to both various price variables and non-price variables are outlined in Table 20. Supply response appeared to be rather inelastic at the two levels at which it was measured, namely, inspected slaughter and at the finished (market) level. A 1 percent change in veal-feed price ratio was associated with only +0.03 percent change in the number of veal cows slaughtered, and a 1 percent change in the price of feedstuffs was associated with only -O.65 percent change in the quantity of finished veal supplied (see Table 20). TABLE 20 PRICE AND OTHER ELASTICITIES 0F VEAL Dependent Explanatory Variables Equation Variable NCMt_1 (FP¥G1t 1 QVSt-l FPFSt_] 6-27 VCSLt +0.87 +0.03 - - 6-29 QVSt - - +0.68 -O.65 Empirical Estimation of Milk Supply The same procedure for beef and veal production was followed here for milk production. Total supply was disaggregated first in the two familiar components, i.e., the number of cows milked and the average fluid milk per cow. .J—' 180 Wilson and Thomson15 found that the supply of milk was largely predetermined with respect to the price of milk in year t and was more elastic, but still inelastic, with respect to the average price of 16 milk in years t-l, t-2 and t-3. Rojko, on the other hand, implicitly assumes that annual milk supplies are predetermined. The equations fitted are as follows: Number of Cows Milked NCMt = -27 1957 + 0.9627N6Mt_1 + 2155.5521FPMKt_1 5 (104.104) (.1049) (3039.587) R2 = 0.96 (6-30) NCMt = 72.01664 + 0.7057NCMt_1 + 1.0094QMCt (33.065) (.177) (.856) -2 * . R = 0.92 d = l.l7(1) (6-31) Average Fluid Milk per Cow inct = 25.050 + 43.1020 0vr + 5.9647 T (6.662) (10.112) (.767) R2 = 0.95 6* = 1.78 (6-32) 15R. Wilson and R.G. Thomson, "Demand, Supply, and Price Relationships for the Dairy Sector, Post-World War II Period," Journal of Farm Economics (May 1967): 360-371. 16A.S. Rojko, The Demand and Price Structure for Dairprroducts, USDA Technical Bulletin 1168, 1957. 181 Total Supply of Milk QMKSt = 35.5269 + 0.6348 0M1<5t_1 - 12 968.879 FPF5t_1 (7.834) (.197) (.0327) R2 = 0.98 6* = 1.94 (6-33) The results for the milk submodel, in general, were satis- factory for all equations as measured by significant regression co- efficients, high coefficients of determination and consistent signs for most of the parameters. None of the equations showed a clear presence of serial correlation, except the two first equations for NCMt which gave a d* value very close to the upper limit of the d values of the tables. The inventory variable, the numbercfl’cows milked in the pre- ceding year (NCMt_]), plays a crucial role in explaining the varia- tion in the number of cows milked in the current year. The own-price of milk in the preceding year was not, strictly speaking, statistically significant, but the weight of its parameter was considerable. This seems to be in accord with what Wilson and Thomson found, results which were stated in the beginning of this section. In a second formulation to explaining the variation in the NCMt variable the per capita quantity of milk consumed was used, but the results did not change considerably. R2 was reduced by 0.04 units, and the d* statistic gave the same result about the presence of serial cor- relation. Average yield of milk per cow was well explained in a rather simple model by using a time variable to represent all the 182 "technological" developments in the dairy industry, and a subsidy variable given to farmers who grew roughage on their farms for feed- ing their own milk cows. The total supply of milk model utilized the quantity of milk supplied in the preceding year (t-l) and the price of feedstuffs. The first variable, the QMKS gives the situation existing in the t-l’ dairy industry in terms of the quantity of milk produced which takes into account every kind of improvement adopted by the dairy in- dustry that resulted in an increase in the quantity of milk. The second variable, FPFS gives the cost structure of the dairy in- t-l’ dustry. In Table 21 estimated elasticities are given. A direct elas- ticity of +0.12 was estimated which indicates that a 1 percent in- crease in farm price of milk would increase the number of cows milked by approximately 0.12 percent. The elasticity with respect to the milked cows inventory was, as expected, high at the level of +0.94. llmeelasticity with respect to per capita quantity of milk consumed was estimated to be +0.12, thus giving the same value as the first model. The elasticities of total quantity of milk supplied with respect to quantity of milk supplied in the preceding year and the farm price of feedstuffs were equal to +0.03 and -O.3l, respectively. These elasticities mean that for a 1 percent increase in the quantity of milk supplied in the preceding year an increase of +0.63 percent in the quantity supplied in the current year is expected. In other words, whatever improvements took place in the dairy industry in 183 preceding years are expected to exert a positive influence on the quantity of milk supplied this year. The supply elasticity of total quantity of fluid milk from cow with respect to farm price of feed-grain supplies was found to be -O.3l which means that a 1 percent increase in the price of feed- stuffs was associated with a -O.31 percent increase in quantity of milk supplied. TABLE 21 MILK SUPPLY ELASTICITIES . Dependent E Explanatory Variables Equat1°" var1ab1e I NCMt_1 FPMKt_1 QMCt QMKst_1 FPF5t_1 6-30 NCMt ‘ +0.94 +0.12 - - - 6-31 NCMt +0.69 - +0.12 - - 6-33 QMKSt i, - - - +0.63 -O.3l Time Lag Consideration General Discussion In economics, as in other studies of human behavior, it is sometimes necessary to take into account the fact that what happens to- day largely depends on what happened yesterday. Thus, for a particular period the quantity of beef produced in Greece by Greek beef pro- ducers (farmers) is determined partly by farm prices and incomes (per- sonal disposable income) in the current period and partly by the number of total cows which existed and/or farm prices of beef which pre- vailed in previous years. Even the consumption demand for beef may be partially determined by patterns of consumption in the previous periods. Indeed, this relationship between past and present forms the basis for what is called the "habit formation" hypothesis on 184 consumption, or Duesenberry's second hypothesis, which states that "pre- sent consumption is not influenced merely by present levels of absolute and relative income, but also by levels of consumption attained in previous periods."9 When these relationships between present and past values of economic variables are expressed as the estimating equations for econometric research, the observations that appear in the equations will include both current values and values for one Or more previous periods. The latter values are described as lagged values of the variables. A model in which lagged values of the dependent variable appear as regressors is termed an autoregressive model. And a model in which a dependent variable is explained by the current value and a series of past values of an independent variable is also essentially an autoregressive model. But this treatment of economic variables means that the relaxation of static assumptions (relaxation of the static theory) is introduced and consideration of dynamic analysis is brought in. To study supply adjustments through time, however, the factors affecting both the speed and magnitude of the adjustment process are considered in a dynamic analysis. In agricultural supply response analysis the reaction to a change in a causal factor is speed over a number of time periods. Thus, time is introduced explicitly in several ways. The lapse of time between cause and effect is referred to as a leg_and may be of fixed duration or distributed equally or unequally over time. 9H.W. Branson, Macroeconomic Theory_and Policy. (New York: Harper and Row Publishers, 1972), p. 188. 185 Nerlove (1956) seems to believe that research on changes in agri- cultural supply aims at understanding the mechanism of supply re- sponse, the ability to forecast supply changes and, finally, the competence to prescribe solutions to problems related to agricultural supply. But in the same paper Nerlove stresses that "farmers react, not to last year's price, but rather to the price they expect, and this expected price depends only to a limited extent on what last year's price was."10 It seems, then, that although price plays a key role in the farmer's decision making process, other factors apparently play an important role, too. Among such other factors, in the case of a country like Greece, especially in the early years of the sample period, farmers have been concerned mainly with labor's employment problem, their aim being how to maximize employment. That explains why Greek farmers have stayed in the tobacco business, a labor intensive enterprise, although tobacco prices have remained low for a long time, and why, when employment opportunities abroad become available, through better information and government encouragement, farmers migrated in mass numbers, Hirschman's11 "Exit" took place. But, again, according to Nerlove10 ...it seems reasonable to assume that the price ex- pected to prevail at some future date depends in some way on what prices have been in the past. Price expectations are, of course, shaped by a multitude 10M. Nerlove."Estimates of the Elasticities of Supply of Selected Agricultural Commodities," J.F.E. 28 (1956): 496-509. nO.A. Hirschman, Exit, Voice and Loyalty (Boston: Harvard Univer— sity Press, 1970). 186 of influences, so that a presentation of expected price as a function of past price may merely be a convenient way to summarize the effects of these many and diverse influences. Koyck1z(l954) and Nerlove (1958)13 have accepted three general reasons for the existence of distributed lags: (a) technological reasons, (b) institutional reasons, and (c) subjective or psycholog- ical reasons. The technological reasons refer to the very realistic fact that production of any physical good requires time. For example, beef production requires two years or more in time and within this time framework it is natural that many delays occur. Delays could occur, for example, in the gestation period, in feeding, and in maintenance biological requirements. Slaughter and processing delays also exist from the time the decision is made to produce the final product. Institutional reasons refer to such a relationship as that existing between income and taxes to be paid, customs put on imported products, etc., which serve as an economic stimulus. But farmers seem to be unable to react immediately to their production plans within the short-run time span when prices change; the result of this kind of delayed reaction is that supply response function becomes irre- versible. The third reason for the existence of distributed lags is subjective or psychological. This category includes a person, a 12M.L. Koyck, Distributed Lags and Investment Analysis (Amsterdam: North Holland PfiblTshing Company, 1954). 13M. Nerlove, The Dynamics of Supply: Estimation of Farmer's Response to Price (Baltimore: John Hopkins University Press, 1958). 187 business enterprise, or another economic subject and describes their reaction in relation to the given circumstances. The attempt, for example, to maximize a utility function and/or profit, given the economic and organizational constraints facing a farmer, will result 'in certain behavioral pattern(s) on the part of the farmer. The view of a price change as a temporary or a permanent change by a farmer constitutes another example in that his perception has a decisive effect upon his farm management decisions. J. Trapp (1976) writes that each force dictates its own for- mulation of a distributed lag relation. And he clearly states that "the econometric techniques developed to estimate lagged output re- sponse over time do not distinguish between these three forces. -The distributed lag structure estimated is an aggregate measure incapable of determining the precise causes of the lagged response."14 Distributed Lag Models A particular autoregressive model used in econometrics is what is called the distributed lag model. Supply response for cattle and/or calves as a tool in studying livestock-feed-grain economy has been expressed by various techniques and/or models. Kulshreshtha (1975)15 ’states that "two types of problems were encountered; either (1) the signs of the price variables were inconsistent, or (2) the sign was 14James Trapp, "An Econometric Simulation Model of the United States Agricultural Sector" (Ph.D. Dissertation, M.S.U., 1976), p. 64. 15N.S. Kulshreshtha, "Canadian Cattle and Calves Industry: A Dis- tributed Lag Analysis," Dept. of Ag. Econ., Univ. of Saskatchewan, Saskatoon, Canada, March 1975, p. 2. ‘— 188 consistent, but the level of significance was very low." He then suggests that one could hypothesize that these problems may arise due to time involved in the production process which may imply model mis- specification. And, if that is the problem, then the researcher could identify the appropriate time lag and estimate the response. Although the problem seems simple at first glance, difficulties are encountered since "the lag between supply and price involves two separate lags: 1) between a change in actual price and expected price and (2) between expected price and the adjustment to this expected price. The re- sponse to an actual price change, then, involves an expectation lag and an adjustment lag, and may be considered as a function of time."16 The pioneer work in this field has been done by Fisher (1937),17 Alt (1953)18 and Hicks (1953). Hicks believes that if past prices are completely dominant then price-expectations can be treated as data, too, along with past prices which are considered simply as data with respect to the current situation. The change in the current price is taken as temporary and does not influence price-expectations. But from the moment when past prices stop being completely dominant then some influence of current prices on expectations must be allowed. 16J.N. Ferris, "Dynamics of the Hog Market with Emphasis on Dis- tributed Lags in Supply Response" (Ph.D. Thesis, M.S.U., 1960), p. 27. 17Irving Fisher, "Note on a Short Cut Method for Calculating Dis- tributed Lags," Bulletin de L'Institut International de Statistique, Vol. 29 (La Haye, 1937). 18L.F. Alt, "Distributed Lags," Econometrica 10 (1942): 113-128. R.J. Hicks, Value and Cagital, Second Edition (Oxford: Oxford Univer- STty Pr9559 5 PP. - 05. 189 Hicks goes on and defines the elasticity of expectations by referring to a particular person's expectations of the price of a commodity as the ratio of the proportional rise in expected future prices of that commodity to the proportional rise in its current price. The elasticity of a person's price expectation takes the value 0 for the case of the given expectations and otherwise the value 1, which means that a change in current prices will change expected prices in the same direction and in the same proportion. An intermediate case of an elasticity of expectations less than 1 and greater than 0 is also recognized by Hicks. The elasticity of expectations is greater than unity when a change in current prices makes people feel that they perceive the change as a permanent one and thus expand their production. On the contrary, the elasticity of ex- pectations will be negative if people perceive the price change as a permanent one and thus decrease their production. Fitting the Polynomial Lag Model The general form of an econometric distributed lag model is written as follows: Qt = a 1 B1P1 1 B2Pt-1 * BBPt-Z +--°+ BnPt-e 1 Et (5'27) where: Qt: quantity supplied in the period t Pt’ Pt-n: price per unit of supply in period t and/or in period t-n 2: time lag of a finite value 0,81,82,...,8n: parameters of the supply function, which here take the name of reaction coefficients at: disturbance term of the equation. 190 The regression equation (6-27) is called a distributed lag model because the influence of the explanatory variable P on E(Qt) is distributed over a number of lagged values of P which are taken over the periods 2. This number of periods, the number 2, can be either finite or infinite. The assumptions concerning the behavior of the variable P and the disturbance term at hold. In order to avoid ex- plosive values of E(Q), it is assumed that the as have a finite sum, i.e., The average lag is defined as the weighted average of all the lags involved, with weights being the relative size of the respective B coefficients. is. i 0 1 Average lag=-————-—-— 8. i=0 1 M3111“): Theoretically, the above model can be estimated by the least squares method or by some other method which leads to estimates with some desirable properties under the usual assumptions about the disturbance term ct. In practice, several difficulties are likely to arise: (a) Theory does not generally indicate the length of the run, and, if 2 is large, we may not have enough observations to estimate * all the parameters. Because of this difficulty the statistical 9: This is a problem of degrees of freedom. 191 significance of the 85 are analyzed in order to find out where to stop adding lagged variables. (b) But, even if we have enough degrees of freedom (enough observations), the lagged values of the P5 are likely to be highly correlated from period to period, thus leading to a high degree of multicollinearity which affects the standard errors of the estimated coefficients. These difficulties have, in practice, led to the imposition of a priori restrictions on both the number of the regression para- meters and the form of the reaction coefficient patterns. The re- strictions reduce the number of parameters, thus saving degrees of freedom and eliminating the need for a number of highly correlated independent variables. In practice, these restrictions have been of two kinds - "one resulting from the requirements that the 85 should be declining in a geometric progression, and the other from the re- quirement that the 85 should first be increasing and then decreasing.”9 The most frequently used assumption about the nature of lag structure is that it should decline in a geometric progression. The geometric lag specification is expressed by two models, the "adaptive expectation model“ and the "partial adjustment" or "habit persistent model". The difference between these two models relies upon the dif- ferent set of assumptions each one makes about the behavior of the 19Jan Kmenta, Elements of Econometrics (New York: The Macmillan Company, 1971), pp. 473-474. 192 disturbance term of the regression equation by which each one of them is expressed. The above specification in the form of the two models has the advantage that the estimation of the regression parameters is easy, but the problem of serial correlation still remains. The serial correlation biases the parameter estimates if such a correlation exists in the regression equation. Lagged variables, it is repeated here, are one way of taking into account the length of time in the ad- justment process of economic behavior and perhaps the most efficient way for rendering them dynamic. This is why they have become in- creasingly popular in applied econometric research. J. Johnston seems to prefer the simple polynomial procedure, 20 while Dhrymes prefers Almon's technique since it is easier to in- corporate zero restrictions on the lag coefficients. Degree of the Polynomial and Length of Runs If Almon's technique is accepted, it is easy to obtain estimators for the n parameters, bo""’bn’ by simply obtaining estimators for the three parameters ao,a1, and a2. It can further be shown that, in cases where there are more b5 than as, the estimators of the b5 obtained by using Almon's technique have smaller variances than the direct estimators of the b5 that would be obtained by applying the multiple regression technique directly.21 Generalizations 20Phoebus Dhrymes, Distributed Lags. Problems of Estimation and Formulation. (San Francisco: Holden Day, Inc., 1971). 21E.E. Kelejian and W.E. Oates. Introduction to Econometrics. Principles and Applications (New York: Harper and Row, Publishers, 1974). 193 and variations of this technique are easily applicable. For example, inclusion of additional variables in no way affects the analysis. Furthermore, in practice, Almon's technique is applied to each of the several lagged independent variables in the same equation. A priori knowledge of theory and of the industry might dictate that either or both bo or bn equals zero. One way to incorporate such knowledge into the model is by simply dropping FPBt and/or FPB from t-n the basic model and proceeding as before. In practice, the informa- tion that either bo = O and/or bn = O, or both, is transformed, by using the basic assumptions, into one or more restrictions on the as and then the resulting equation is estimated. A second way to incorporate a priori knowledge in Almon's method is for the researcher to know both the length of the lag structure and the degree of polynomial which gives the general pattern of bs. But, in reality, none of them is known. To overcome that obstacle the researcher usually chooses a degree for the polynomial, (for instance, 0) that is high enough to include any reasonable pattern of the bs. In most cases a third or fourth degree polynomial is sufficient to deal with the data at hand. To test for the curve (which the data form gives), a flexibility to the data is usually given in such a way as to fit the best curve with no restriction on the data. This is done by not specifying any restriction on the b5 so that no a priori curve is accepted. By doing this, the researcher has the adVantage of not being restricted on the kind of curve he has to choose from in ad- vance since, if a wrong one happens to be selected, there is always a way to check for it before he is making his final decision. 194 Suppose, now, that the reasonable lag which is believed to be consistent with the relationship at hand is L*. Once D and L* have been selected, estimation of the relationship under consideration for L = D,D+l,...,L* is possible. The assumption made here is that lags (L ;=D) are usually chosen on the basis that they are greater than or equal to the degree of the polynomial (0), since it is assumed that the length of the lag, L, is at least as long as the degree of the polynomial, D. This means that there are as many bs as there are as, since, if L < D, then the number of independent variables has to in- crease; but the purpose of ad0pting Almon's technique is to estimate the number of parameters in the model. Another point is worth mentioning here. If the various equa- tions fitted are to be compared using the highest value of R2, i.e., the coefficient of determination corrected for the number of degrees of freedom, then all the regression equations corresponding to the values of L should be estimated with the same data. This, in other words, means that the first L* observations are lost and only the remaining T-K* observations are used (T denotes the number of ob- servations used in the sample). That value of L is used with maximizes the R2, and/or minimizes the standard error of estimate. Recent examinations of polynomial lag models have indicated that statistical criteria for selecting the polynomial degree and lag length which will lead to unbiased distributed lag estimates 22 are presently not available. Due to this lack of information the 22F. Frost, "Some Properties of the Almon Lag Technique when One Searches for Degree of Polynomial and Lag," Journal of the American Statistical Association 70 (1975): 606-612. 195 approach taken in this study was to select the degree and length of the polynomial lag according to the knowledge of beef and milk in- dustries in Greece as opposed to using statistical criteria. Further- more, constraints were used which require the distributed lag pattern to pass through zero in the second and last period of lag. These two zero constraints are based on the argument that no significant simultaneous or single period response in output generally occurs in livestock production, and, hence, after a given time period, a given incentive no longer affects production. Moreover, a number of non-restricted curves has been tried in this thesis to see if the best (curves) polynomials fitted to the data are the appropriate ones. Polynomial lag models were tried here since an inverted V'-lag model capable of generating lag structure was desired. The rationale of an inverted-V'-lag structure for livestock supply re- sponse is discussed by J. Trapp.23 Some Decision Criteria Fitting polynomial lags may involve making some arbitrary decisions, and to deal with such a problem some guidelines were borrowed from economic theory and judgement. Economic Theory Economic theory suggests that the effect of one variable upon another will be in the same direction over time. For example, 23N.J. Trapp, "A Polynomial Distributed Lag Model of Pork Production Response," Dept. of Ag. Econ. Staff Paper No. 75-29 (East Lansing: Michigan State University, 1975). 196 theory specifies that price has only a positive influence on supply ' over a period of time. This, translated in the results obtained from the empirical analysis results, means that the estimated coefficients should all have the same signs. But this is not the case in practice. Price increases exert a different influence on supply than price de- creases . Judgment When the lag distribution estimated from period 0 to period L has a coefficient similar to that of the last variable, the lag length is increased. 0n the contrary, when the lag distribution has a co- efficient with a different sign than its last variable, then the lag length is decreased. If estimated coefficients obtained from the empirical analysis are statistically significant, this is taken as an indication that this procedure is correct. If, on the other hand, estimated coefficients are not statistically significant, then the procedure is justified' on the grounds that the variables included can be considered logical, since the lack of statistical significance may be due to high inter- correlation between two variables or a linear combination of variables in the particular sample of data on which the estimates are based. The technique cannot clearly distinguish the separate effects of the variables, given the available data. Selecting the Polynomial Lag Model In selection of the polynomial lag model, the following pro- cedure was used. First, an arbitrary degree of polynomial was chosen 197 which had to account for the peculiarities of the data at hand. Then, for this arbitrarily chosen degree of polynomial different lag lengths were tried, and that one was selected which minimized the standard error of the regression. For that lag length different degress of' polynomial were tried, the one which minimized the standard error of the regression was chosen, and, finally, the necessary zero con- straints were imposed. The aforementioned procedure was followed since it was soon revealed that no one lag length was necessarily best in terms of minimum standard error of the regression for all degrees of poly- nomial and vice versa. In a sumnary of what has been said in this section, it should ' be repeated that the difficulty in using the polynomial lag model is that the researcher has to specify the degree of polynomial and the length of the lag. As far as the latter is concerned, the data at hand can help a bit. Kmenta states that one possibility is to keep on extending the length of the lag until the contribution of the additional X's to the regression sum of squares is no longer statis- tically significant. But if the X's are highly correlated, this criterion may not work very well. ' Another possibility is to choose that lengt of the lag which results in the highest value of R , i.e., of the coefficient of determination corrected for the "number of degrees of freedom." However, this also may not always work we 1 since the differences between several values of R may be very small. Nevertheless, one or the other of these criteria, plus other considerations (for example, that all weights should be positive), may help in choosing the "best" lag for the problem at hand.24 24kmenta, pp. 494-495. 198 Empirical Results of Distributed Lag Models The Beef Cattle Slaughter (BCSLt) Eguation There are several factors which influence the number of cattle going to slaughter. The most important factors seem to be: Price of beef: Replacement theory suggests that a cow will be slaughtered when its value for slaughter exceeds its present value which includes discounted future earnings in milk production. The higher the price of beef, relative to milk, the greater will be the number of cows going to slaughter. This theoretical view, however, considers cattle slaughter as an instantaneous response to any change in price of beef. But this is not the case in the cattle industry. As Hicks pointed out, the elasticity of expectations will be negative if farmers perceive the price change as a permanent one and they decrease their production. Thus, in response to an upward movement of price of beef fewer cattle are expected to go to slaughter, while in response to a downward movement in the price of beef more cattle are expected to go to slaughter. . Price of milk: As milk price increases, it is expected that beef cattle going to slaughter will increase, thus favoring the building up of the dairy herd. Price of feedstuffs: Feedstuffs fed to the herd as a whole come from two sources: the market and feedstuffs grown on the farm. The last category of feedstuffs costs less to the farmer than the first one, but climatic conditions and severe competition for land and other resources from other farm activities result in a chronic problem of inadequate production of feedstuffs. Given that and the 199 fact that cattle activity is a complementary farm activity in Greece, it is expected that an increase in price of feedstuffs will cause an increase in the number of cattle going to slaughter. Beef-feed price ratio: Since this factor embraces two variables, the expected result is not always clear. Under Greek conditions, how- ever, prices of feedstuffs have remained remarkably stable, while beef prices had ups and downs during some years of the sample period. Thus, farmers are expected to respond to beef prices rather than to feed- stuff prices whenever this price ratio is used, and a negative rela- tionship is expected between the beef-feed price ratio and the number of beef cattle going to slaughter. Number of cows milked: As cows become older their productivity (in terms of quantity of milk produced) declines. Farmers take care to replace these old cows with younger ones of higher productivity. Thus, a negative association is expected between beef cattle going to slaughter and number of cows milked the previous year (short-run). In the long-run and in absolute levels more cows are expected to be slaughtered as the inventory increases. Milk-feed price ratio: The data reveal that this ratio has been kept relatively stable, and, whenever there was a disturbance it was positively associated with the number of cattle going to slaughter. Given the peculiarities of the data at hand, the biological production conditions of beef and the dual purpose nature of the national herd in Greece, second and third degrees of constrained polynomials were tried. The results of the empirical analysis are given in Tables 20 and 21 below. POLYNOMIAL LAG MODEL PARAMETERS AND STATISTICAL RESULTS SECOND DEGREE POLYNOMIAL CONSTRAINED. Equation (6-28) 200 TABLE 22 TIME LAG: 10 YEARS Regression Coefficients, t-values, d*, Standard Error of _t, Regression_(SER)_p . FPMK FPB - * TTme Intercept -———— R d SER Period FPFG FPFG 126.766 .72 1.06 15.68 ((21.39) i t-l ! 0.0654 -O.9216 E (5.82) (6.19) t-2 0.1157 -l.659 (5.82) (6.19) t-3 0.1570 -2.212 (5.82) (6 19) t-4 2 0.1832 -2.581 (5.82) (6.19) t-5 0.1962 -2.765 5 (5.82) (6.19) t-6 ; 0.1962 -2.765 3 (5.82) (6.19) t-7 5 0.1832 -2.581 3 (5.82) (6.19) t-8 l 0 1570 -2.212 (5 82) (6.19) t-9 0.1170 -1.659 (5.82) (6.19) t-lO 0.0654 -O.9216 (5 82) (6.19) 201 TABLE 23 POLYNOMIAL LAG MODEL SIMPLE AND CUMULATIVE (é) ELASTICITIES SECOND DEGREE POLYNOMIAL. TIME LAG: 10 YEARS Time Dependent Explanatory Variables Period Variable FPMK - Fpg . BCSL W6 p e “F'P‘FE‘ e t-l 0.13 0.13 -O.l4 -O.14 t-2 0.23 0.36 -0.26 -O.4O t-3 0.31 0.67 -O.34 -0.76 t-4 0.36 1.03 -0.40 -l.16 t-5 0.38 1.41 -0.43 -l.59 t-6 0.38 1.79 -O.43 -2.02 t-7 0.36 2.15 -0.40 -2.42 t-8 0.31 2.46 -O.34 -2.76 t-9 0.23 2.69 -0.26 -3.02 t-lO 0.13 . 2.82 -O.l4 -3.l6 ~202 Polynomially distributed weights were estimated simultaneously (within a single function) for the following variables: the beef-milk price ratio and the farm price of feed-grain, the farm price of beef and the farm price of feed-grain, the milk-feed price ratio and the number of cows milked the previous year, the milk feed price ratio and the beef-feed price ratio. Second Degree Polynomial(s) Diagramatically, the distributed weights for the two variables included in the model are as they appear in Figures 6 and 7, respectively. When the function parameters and statistics are observed several points are noteworthy. The distributed weights associated with milk-feed price ratio are all positive; however, the first weight and the last weight are rather small but still significant by the t-value criterion. All weights give a good statistical level of significance when the t-value criterion is employed. Hence, it may be concluded that significant response is indicated to milk-feed price ratio for the fourth year and onwards up to the ninth lagged year. This response is not surprising under the average typical situation prevailing in Greece, especially until the 19605 when farmers used to keep their cows until they were almost physically, though not economically exhausted. Since milk was mostly consumed on the farm by the family, the time lag - for this period - seems not to be very long. With regard to beef-feed price ratio distributed weights it can be observed that they are all negative. This result may seem to be a little puzzling to some readers who are familiar with pure business oriented cattle operations. A closer look at the data 203 moemwo: ooosoweomwo omuma women townimmmm .m meamwm mmw .moemwoz oooacweomwo owoam .oi owoaa oowca oooa-uprz .m ocsmwo ‘ cowuom .Vomwma . 0H9muwmfinmh 0H mew n-m mid n m H o cownom .mcwumn moH.nv 204 reveals, however, that farm price of beef had followed a kind of erratic movement (experiencing ups and downs). Thus, for an upward movement of beef price farmers' elasticity of expectation was negative; they perceived this upward beef price movement as a permanent one and decreased their production. Given the fact that prices of feedstuffs (feed-grain) were kept remarkably constant any farmer's response was associated with changes in the price of beef. The corrected multiple regression coefficient R2 was found to be 0.72, which is rather high considering the fact that only two vari- ables participate in the model. It seems that the inclusion of farm price of feed-grain in the calculation of both price ratios contributes somewhat to the problem of serial correlation. To correct for the presence of serial correlation, the Cochrane-Orcutt25 procedure was used. A clear-cut way of summarizing the impact over time of a given price change is to observe the cumulative elasticity of oUtput re- sponse to changes in a given price. Cumulative elasticities were calculated for each period and are presented in Table 21 and displayed graphically in Figure 8. Negative beef-feed price ratio elasticities have been converted to positive values in Figure 8 to provide easier graphical comparisons. The cumulative elasticities for the final period can be interpreted as the total or long-run elasticity, while those for intermediate years represent varying degress of short-run elasticities. 25D. Cochrane and G.H. Orcutt, "Application of Least Squares Regressions to Relationships Containing Autocorrelated Error Terms," Journal of the American Statistical Association 44 (1949): 32-61. Cumulative Elasticity 205 3.00; FPB/FPFG* I‘.r'1"" _ a 1." .40 2.50. taFPMK/FPFG 2.00 1.50 1.00 0,50 0.10 i 2 3 4 5 6 7 8 9 IO Legged Period Figure 8. Cumulative Elasticities: Second Degree Polynomial * Negative beef-feed price ratio eleasticities have been converted to positive values to provide easier graphical comparisons. 206 The graphical display of the elasticities shows the general pattern of marginal responses to price changes for several years in the two price ratios included in the model. Third Degree Polynomial(s) What follows is a brief comparison between the second degree polynomial and the third degree polynomial for the same length of run. It seems that in terms of signs of response of the variables involved the two degrees of polynomials give the same result consistent with each other. Thus, a positive sign was obtained in both of them for all the estimated coefficients of the milk-feed price ratio and a negative sign was obtained for all estimated coefficients of the beef- feed price ratio. This finding seems to verify what was said at the beginning of this section about output and input response in the number of beef-cattle going to slaughter. In terms of significance of the estimated coefficients - when a t-statistic is employed - it seems that in the second degree poly- nomial all of the estimated coefficients are significant. However, in the third degree polynomial estimated coefficients corresponding to t-8, t-9, t-lO years appear not to be statistically significant by the t-value criterion. This result seems more plausible and it led to the employment of an eight year run length polynomial. Results for the third degree polynomial with a time lag of ten years and with a time lag of eight years are given below in Tables 21, 22, 23 and 24. In terms of R'zs, it seems that the third degree polynomial 2 gives higher values (R2 = .84 and R = .72, respectively) compared with those of the second degree polynomial, while in terms of standard 207 errors of regression the third degree polynomial seems to give standard errors of regression of a lower value than the second degree polynomial (12.31 and 15.68, respectively). From Table 21 it is revealed that the intermediate short-run elasticities, although they look the same, differ; and so do the final long-run elasticities of the two variables involved. The beef-feed price ratio cumulative elasticity rises faster than the milk-feed price ratio cumulative elasticity which may indicate that output prices cause greater fluctuations in the number of beef-cattle going to slaughter. While this happens when a second degree polynomial is employed, events occur in the reverse when a third degree polynomial is employed (see Table 23). Table 23 reveals that the short-run elasticities of the milk-feed price ratio are higher than those of beef-feed price ratio in the first four years become equal at the fifth year; the beef-feed price ratio elasticities begin to rise faster thereafter, i.e., after the sixth year. Table 22 shows equation (6-29) which was tried for a third degree polynomial with a length of run of ten years. The t-values underneath the estimated regression coefficients show that para- meters corresponding to years t-8, t-9, t-lO are not statistically' significant for the milk-feed price ratio, while the same parameters are significant for.the beef-feed price ratio. The non-significance of the parameters corresponding to years t-8, t-9 and t-lO for the milk-feed price ratio is in accord with 26 other findings elsewhere where it was found that milk-cows 26Kitsopanidis, p. 12. 208 which yield an average of 1501-2000 kg of milk per year have an average economic life of 6 to 7 years. This lead for an eight years length of run to be estimated for the variable number of beef cows slaughtered which is given later in this section. Equation (6-29) in Table 22 gives an R2 = 0.84 and that com- bined with the fact that most of the parameters are statistically significant gives some reliance in using it. The various elasticities are shown in Table 23 and the shape of the various lag coefficients is shown in Figures 9 and 10. The shape of the elasticities is 'shown in Figure 11. Trying a third degree constrained polynomial with eight years lag in the variables FPB and FPFG gave the results illustrated in Tables 24 and 25- It is interesting to note that this third degree polynomial which utilizes the farm price of beef and the farm price of feed- grain along with the NCMt_1 variables gives a supply response which is negative for the first variable (FPB) and for the first four years and is positive thereafter. For the second variable (FPFG) the supply response is negative only for the first two years and positive thereafter. The same is true for their respective Cumulative elasticities A An explanation for such behavior by farmers as far as the price of beef is concerned may be that because any change in beef price is conceived by them as permanent, they keep their beef-cattle from slaughter. However, this perception by farmers seems to last only two years. After two years they are ready to accept change in beef 209 TABLE 24 POLYNOMIAL LAG MODEL PARAMETERS AND STATISTICAL RESULTS THIRD DEGREE POLYNOMIAL, CONSTRAINED. TIME LAG: 10 YEARS Equation (6-29) Regression Coefficients, t-values, d*, and Standard Error of Time Regression (SER) Period Intercept FPMK FPB 2 * —_FPF FFF6 R ‘1 SER 125.813 .84 1.83 12.31 (28.900) t-l 0.0900 -0.ll72 (1.622) (.164) t-2 0.1465 -0.4656 (1.907) (.470) t-3 0.1745 -0.9603 (2.419) (1.026) t-4 0.1795 -1.516 (3.531) (2.255) t-S 0.1665 -2.049 (5.670) (4.984) t-6 0.1405 -2.473 (3.289) (4.377) t-7 5 0.1070 -2.705 ; (1.497) (2.942) t-8 0.0709 -2.658 (.779) (2.281) t-9 0.0376 -2.248 (.413) (1.929) t-lO 0.0123 -l.390 (.193) (1.714) 210 TABLE 25 POLYNOMIAL LAG MODEL SIMPLE AND CUMULATIVE (é) ELASTICITIES Ifllgg_DEGREE POLYNOMIAL. TIME LAG: 10 YEARS Time Dependent fl Explanatory Variables _ Period Vagggtle %%¥%. é gggg' é t-l 1.17 0:17 -0.02 -0.02 t-2 . 0.28 0.45 -0.07 -0.09 t-3 ;O.34 0.79 -0.15 -0.24 t-4 E0.35 1.14 -0.24 -0.48 t-5 $0.32 1.46 -0.32 -0.80 t-6 0.27 1.73 -0.38 -l.18 t-7 0.21 1.94 -0.42 -l.50 t-8 : 0.14 2.08 -0.41 -l.91 t-9 1 0.07 2.15 -0.35 -2.26 t-lO g 0.02 -0.22 -2.48 2.17 211 TABLE 26. POLYNOMIAL LAG MODEL PARAMETERS AND STATISTICAL RESULTS THIRD DEGREE POLYNOMIAL. CONSTRAINED. TIME LAG: 8 YEARS Equation (6-30) Time Regression Coefficients, t-values, and Standard Error of Period Regression ,, - * Intercept NCMt_1 FFB FPFG R2 d SER 107.423 0.0440 .86 1.76 11.98 (2.540) (.3231) t-l -389.6 -1012.0 (4.7) (1.2) t-2 -526.9 -628.0 (4.9) (.5) t-3 . -478.2 660.0 (5.2) (.6) t-4 -309.9 2635.0 (5.1) (3.0) t-5 -88.56 3997.0 (1.5) (5.5) t-6 119.5 5066.0 (1.3) (5.5) t-7 ‘ 247.9 5083.0 (2.4) (5.0) t-8 230.2 3557.0 . (2.8) (4.5) 212 AyaweoesFoa oocmoa ecaeev .moemwoz uwuznwgumwo ompmm mowed tmmmiwmmm .op meamwu “RH w m s~ m 3.28m towmcq firmweocapoa mmemmo uewzev . osmoeowoz nooseweompo .oem cowea coco-xptz .m ocsmta coahwm tommmg OH m m m .w m J m N , 1 . H o wou.m1 C an .5 .L a 3 4+ .+ J J +r .t o. o. n n 4+ n+ e a o. 0. hm Rm 8 8 ++ .+ o“ .03 u. u. o+ 4+ muH.o 213 2.50 FPB/FPFG* 2.00 I FPMK/FPFG 1.50 1.00 0.50 T V v V I I 2 3 4 5 - 6 7 8 9 IO 11 12 Figure 11. Cumulative Price Elasticities: Lagged Period Third Degree Polynomial * Negative beef-feed price-ratio elasticities have been converted to positive values to provide easier graphical comparisons. 214 TABLE 27 POLYNOMIAL LAG MODEL SIMPLE AND CUMULATIVE (é) ELASTICITIES THIRD DEGREE POLYNOMIAL. TIME LAG: 8 YEARS Time Dependent Explanatory Variables Per1od Vagagfle FPB t FPFG é t-l -0.56 -0.56 -0.09 -0.09 t-2 -0.76 . -1.32 -0.05 -0.14 t-3 . -0.69 -2.01 0.06 0.08 t-4 -O.48 -2.49 0.22 1 0.30 t-5 -0.14 -2.63 0.34 0.64 t-6 0.18 2.45 0.43 1.17 t-7 ‘ , 0.39 2.84 0.43 1.60 t-8 0.36 3.20 0.30 1.90 215 prices as temporary and, as the price of beef goes up, more cattle are slaughtered. As far as price changes in the farm price of feed-grain are concerned, it seems that it takes only two years for farmers to realize that any change in feed-grain is not a permanent one; during these two years whenever feed-grain prices move upwards the number of cattle going to slaughter moves downwards. Yet for the rest of the years farmers perceive changes in feed—grain prices as permanent ones and positively respond to them by slaughtering their cattle. Their actions are in accord with the a priori knowledge revealed by the data. The distributed weights of the coefficients for this model are displayed in Figures 13 and 14, and the simple and cumulative elas- ticities for such a model are shown in Table 25. - From Table 25 it is obvious that both short-run and long-run elasticities (taken in absolute terms) are greater for the farm price of beef variable than for the farm price of feed-grain variable. Be- cause this result is in accord with findings of previous models, there is a strong indication that it is true. In only three years was the elasticity of the price of feed-grain higher than that of the price of beef. 1 Figure 15 graphically displays long-run cumulative elasticities - in absolute values - calculated from the model in Table 24. Number of Cows Milking (NCMt) Eguation In what follows in this section the number of cows milking variable is examined and, following the previous procedure, two degrees of polynomials are tried in a polynomial lag formulation. There are 216 meEocapoo moemmo vewgp "musmwmz cwuzaweumwo mowed cameoiummm .mp mezmwu 7‘ uoaumm zomuaa m P F ’1' ' F 1‘ 4 NH mesocx—oo mmemmo vows» “mp;m_oz vmuznmeumwo woven comm wommmi . znfinan 994nQTJ4ST0 .N_ oc=m_o qqfltam peanutlistd coauom nomumq mmm.m wn.mm- mouam 217 3.00 2.00 1.00 0.20 12 34 5678 Legged Period Figure 14. Cumulative Elasticities: Third Degree Polynomial * Negative price elasticities have been converted to positive values to provide easier graphical comparisons. 218 several variables which are thought to have an influence on the number of cows milked. a) b) C) d) Price of milk. It is expected that as price of milk in- creases the number of cows milked will increase. Yield per cow. As yield per cow increases, as is the case here, the number of cows will decline if, of course, the farmers' target is to produce a given quantity of milk. Price of inputs. As input prices increase, other things being equal, it is expected that the number of cows milked will decrease. Profitability of alternative enterprises. These enter- prises can be found either within the farm and within the agricultural sector or outside agriculture, i.e., off- farm employment of resources used in dairy activities. The profitability of the farm products which compete for resources within the farm and the sector can be expressed in terms of the prices of those products which compete with dairying for the same resources. Such products could be lamb and mutton, pork and chicken. The profitability of off-farm employment opportunities could be expressed as the wage differences in the farm and non-farm sectors. Inventory of heifers. According to the definition of a beef- cow given earlier and according to the biological fact that at least two years is required from the time a heifer is born until it begins to produce milk, the number of milking cows two years from now will be 219 influenced by the current inventory of heifers. This is generally regarded as a stage in milk production and not as a factor except when it is used to forecast cow numbers a year or two ahead. Though these points about the variables influencing the de- pendent variables NCM come from economic theory, all these variables t were not used. Instead, an effort was made to keep the models as simple as possible. Thus, the variables FPMK + FPFG', FPB + FPFG‘ and the variable NCM were utilized in the polynomial distributed t-l lag formulation. The equations fitted here are shown in Tables 26, 27, and 28, Second Degree Polynomial(s) Here, again, the polynomial lag model suggests that the rate of output (i.e., number of cows milking) response to milk-feed price ratio first increases and then declines (see equation 6-31). This type of output response is in accord with other studies like the one carried out by Chen, Courtney, and Schmitz in which they state that ...The polynomial lag model suggests the rate of output response from a price change first increases and then declines. This output response does not seem to be un- realistic for milk production since it appears unlikely that the greatest marginal output from a given change in price is forthcoming in the immediate period after the price change.28 The response to the two price ratios examined here is shown in Figures 16 and 17, and the simple and cumulative elasticities are provided in Table 27. 28Dean Chen, R. Courtney, and Andrew Schmitz, "A Polynomial Lag Formulation of Milk Production Response," A.J.F.E. 54, No. 1 (Feb. 1972): 77-83. 220 ' TABLE 28 POLYNOMIAL LAG MODEL PARAMETERS AND STATISTICAL RESULTS SECOND DEGREE POLYNOMIAL, CONSTRAINED. TIME LAG: 10 YEARS Equation (6-31) ——v r1 Time Regression Coefficients, Standard Errp£§, d* and R2 1 . Period FPMK FPMK 2 * (Intercept jNCMt-l FPFG FPFG' R d £98.510 0.7249 .93 1.80 '(50.936) (.157) ' . t-l 3 0.0169 -0.1593 1 (.012) (.183) t-2 0.0304 -0.2867 (.022) (.329) t-3 0.0405 -0.3823 (.029) (.439) t-4 0.0473 -0.4460 (.034) (.513) t-5 é 0.0506 -0.4779 1 (.037) (.549) t-6 1 0.0506 -0.4779 1 (.037) (.549) t-7 { 0.0473 -0.4460 . (.034) (.513) t-8 E 0.0405 -0.3823 g ( 029) (.439) t-9 f 0.0304 -0.2867 g (.022) (.329) t-lO 0 0169 -0.1593 (.012) (.183) 221 TABLE 29 POLYNOMIAL LAG MODEL SIMPLE AND CUMULATIVE (é) ELASTICITIES SECOND DEGREE POLYNOMIAL. TIME LAG: 10 YEARS I Time Dependent Explanatory Variables Per1od Var&:ale Eggé- é gggg' é t-l 0.03 0.03 -0.00 0.00 t-2 0.06 0.09 -0.01 -0.01 t-3 0.09 0.18 -0.02 -0.03 t-4 0.10 0.28 -0.02 -0.05 t-S 0.11 0.39 -0.02 -0.07 t-6 0.11 0.50 -0.02 -0.09 t-7 0.10 0.60 -0.02 -0.11 t-8 0.09 0.60 -0.02 -O.13 t-9 0.06 0.75 -0.01 -0.14 t-lO 0.03 0.78 -0.07 -0.21 222 TABLE 30 POLYNOMIAL LAG MODEL PARAMETERS AND STATISTICAL RESULTS SECOND DEGREE MODEL. TIME LAG: 4 YEARS Equation (6-32) Time Regression Coefficients, Standard Errors, d* and R2, Simple Period (e) and e Elasticities . * - Intercept NCMt_1 FPMK d R2 Simple Cumulative e e -219.4802 0.9663 - 3.45 .79 (88.36) (.2116) t 610.8126 0.03 0.03 (681.58) t-1 1223.00 0.07 0.10 (360.6) t-2 1834.0 0.10 0.20 (541.0) t-3 1834.0 0.10 0.30 (541.0) t-4 1223.0 0.07 0.37 (360.0) 223 moem_o2 noosoweomwa owpmm m -~ omen ummuimmmm .o_ mezawu e mocm.o: moecwoz eoosoweomwo owoem ooeea eaoo-¥_wz .m_ messed .« 10H H O . mmHo.o mmmm.o-; a 224 In observing the function parameters and statistics, several points are worth mentioning. The distributed weights associated with milk-feed price ratio are all positive; however all the weights have questionable levels of significance as indicated by their standard errors values. Hence, it may be concluded that this type of formula- tion gives no significant response to milk-feed price ratios examined in this model. The distributed weights to beef-feed price ratio are all negative; however the weights are not statistically significant and it may be concluded that no significant response is indicated to beef-feed price ratio. The largest weights for both price ratios appear in t-5 and t-6 years. It is interesting to note that here the two elasticities appear to have quite different slope. The milk-feed price ratio is steeper, while the beef-feed price ratio appears to show a more gradually smooth trend (see Figure 18). If milk-feed price ratio changes by 1 percent at time t-l, the simple supply elasticity is 0.03 and 0.00 for the beef-feed price ratio. This value reaches a maximum of 0.11 at t-5 and t-6 years for the milk-feed ratio, while a maximum of 0.07 appears to be at t-lO for the beef-feed price ratio. It can be generally observed that both short-run and long-run elasticities are small, verifying the fact that the number of cows milking does not respond directly to only two economic factors but to non-economic factors as well (i.e., feeding the family). 225 .mcomwcanou quwsamem emwmmm muw>oea ow mmzpo> m>muwmoa op umpem>cou coma w>og mmwuwuwwmmFm woven m>wpmmmz+ Powsocxpoa mmemmo ucoumm "mmwuwuwpmmpu m>wumF3530 .n_ mesmwo vowhmm zcmmma AH RH HH OH 6 m e a m s m m H II IIIl‘I I4] 1 1541.1. 7" -0... Human mum Hummm xzmh OO.H thotqscta aatqetnmno 226 A Second Degree Model of Four Years Time Lag In this model a four years time lag is incorporated using only one price variable, i.e., the farm price of milk variable. It seems that this simple model gives good results in terms of sign of the ‘parameter estimated and statistical significance of it and of R2. The results of this model are shown in Table 28. In the last model all the lagged parameters are statistically significant, bear a consistent sign, and have important economic significance. The R2 = .79, which is high enough, and the cumulative elasticity at t-4 year is + 0.37, which is very acceptable. It is interesting to note here that the introduction of the variable NCMt_] not only increases R2 from 0.83 to 0.93, but also gives a negative type of response for the beef-feed price ratio and, at the same time, corrects the deficiency for the milk-feed price ratio appearing in other formulations. It is recognized that live- stock inventories will increase when favorable prices and feed costs are expected to occur. In contrast, inventOries are expected to decrease as a result of a rise in feed cost. The introduction of the current livestock inventories in this long-run supply formulation response was not statistically significant, and, thus, Tryfos'29 point about the significance of a current inventory variable appearing among the independent explanatory variables in a short-run supply function cannot be verified here. 29Peter Tryfos, "Canadian Supply Functions for Livestock and Meat," A.J.F.E. 56, No. 1 (Feb. 1964): 107-113. 227 Third Degree Polynomial In what follows a polynomial of the third degree is tried. The difference between the second degree polynomial and the third degree one is that in terms of R2 the latter gave an R2 higher by 0.01 percent than the former and that result occurred only in the case where the variable NCMt_1 was included in the model. The first of the many differences between the two models exists in the pattern of response of the two price variables included in the model. While the second degree polynomial gives a clear positive response for the milk-feed price ratio and a clear negative response for the beef-feed price ratio, the third degree first gives a negative response to the first price ratio (FPM:FPFG) for the first three years and then a positive supply response thereafter (see Figures 19 and 20). On the other hand, this third degree polynomial gives a positive supply response to the beef-price ratio for the first four years and a negative one thereafter. The second difference between the two polynomials is the magnitude of the elasticities corresponding to the two variables in- cluded in the model. Although the individual year elasticities differ, the final long-run cumulative elasticity is almost the same for the two models and for the two variables included in the model [(+0.78, +0.67) and (-0.21, -0.29)], respectively (see Tables 27 and 29). The cumulative elasticities calculated from the third degree polynomial are portrayed in Figure 21. 228 meEocapoo mwemmo newgh moemwoz eoosaweomeo owpom woven vmmoimmmm OH .m_ oesmwo chEocxroo mwemmo usage moemwoz noosaweomwa owpmm mowed ummuix—wz OH .m_ oesmwo ac.ou mmH.o 229 TABLE 31 POLYNOMIAL LAG MODEL SIMPLE AND CUMULATIVE (e) ELASTICITIES. THIRD DEGREE POLYNOMIAL. TIME LAG: 10 YEARS gleeOd 13:3?2512t Explanatory_Variables ‘ NCM gig-13% é - $133 w e t-l -0.09 -0.09 0.10 0.10 t-2 -0.13 -0.22 0.13 0.23 t-3 0.12 0.10 0.11 0.34 t-4 0.08 0.18 0.06 0.40 t-5 0.03 0.21 -0.02 0.38 t-6 0.03 0.24 -0.09 0.31 t-7 0.09 0.33 -0.16 0.15 t-8 0.13 0.44 -0.20 -0.05 t-9 0.14 0.58 -0.20 -0.25 t-lO 0.09 0.67 -0.14 -O.29 230 .mcomweoasou Fm0wcamem memmm muw>oea op mmsFm> m>wpwmoa ow empem>cou coma m>oc mmmpmuwumopm woven m>wpomwz .1 Pcwsocxyoo mmemmo usage .mwwuwuwpmmpu m>wpmpasau .om mesmwo noanom momwmg . 0H m 4 1 .umoexmae. .omomxxzom no.0 om.o om.o 05.0 231 Summary From a look at the elasticities derived from the various models, it can be observed that model specification indeed plays a considerable role in specifying the magnitude of the elasticities. However, the polynomial lag models give a rate of output response (i.e., the number of cows milking) from a price change which is first increased and then decreased. This type of output response does not seem to be so unrealistic since it is derived only if a flexible price lag structure, such as the polynomial lag formulation, is used. The problem faced here is that the distributed lag models are derived from time series data in which it is likely that more than one price change occurred. It seems very unrealistic, for example, to assume that the output response (i.e., the number of cows milking here) is due only to a once-and-for-all change in one price, and that the other prices during this long period (one to eight or ten years) remained unchanged. Thus, it is more realistic to assume that the aggregate supply function estimated indicates the responsiveness of production from a whole set of price changes. Even in the case of equation (6-32) where one price, the price of milk, is examined, it is believed that the condition of ceteris paribus holds and refers not only to other factors (technology, etc.) influencing the milk production, but also to the rest of the set of prices which affect milk production. These findings here are in accord with the findings of H. Halvorson (1958, p. 1104 for the U.S.A.)30 where he reports a 0.35 30H. Halvorson, "The Response of Milk Production to Price," Journal of Farm Economics 40 (1958): 1101-1113. 232 long-run milk supply elasticity for price, while W.W. Cochrane reports a short-run price elastidity of supply for whole milk ranging from 0.3 to 0.4.31 31W.W. Cochrane, "Conceptualizing the Supply Relation in Agriculture," Journal of Farm Economics 27 (Dec. 1965): 1161-1176. CHAPTER VII . TRENDS AND PROJECTIONS Most projections of the demand for agricultural products are made primarily on the basis of assumptions as to the rate of popula- tion and income growth. Population in Greece has not grown very much, nor is it expected to grow very much on the near future. The rates at which population changes annually in both absolute and in percent- age terms during 1963-1973 are shown in Table 30 below, along with the numbercfl’tourists coming into the country from 1969-1972. From Table 30 it can be seen that increase in population does not account for very much in the aggregate increase of consump- tion of beef and veal. Tourism, on the other hand, is a seasonal problem to be tackled. A separate empirical analysis carried out by the author indicated that when annual data were used along with the numbers of tourists as a spearate independent variable, statistically significant results were not produced for the parameter of tourists coming into the country. Consider that these tourists stay 20 days in the country; if it is assumed that they consume the average quantity that an average Greek consumer did daily, i.e., 21 grs of beef and 31 grs of veal per day, then multiplying these numbers by the number of tourists going into the country, one gets 1023 tons of beef and 1510 tons of veal. These amounts are not considered to be that much, given the fact that annual beef production accounted 233 234 TABLE 32 MID-YEAR ESTIMATES 0F POPULATION IN GREECE: l963-1972 AND NUMBER OF TOURISTS COMING INTO COUNTRY FROM 1969-1972 Mid-year Population Changes Tourists Absolute Numbers Percentages % 1963 8 479 625 31 392 0.37 - 1964 8 510 429 30 804 0.36 - 1965 8 550 333 39 900 0.47 - 1966 8 613 651 63 318 0.74 - 1967 8 716 441 102 790 1.19 - 1968 8 740 765 24 324 0.28 - 1969 8 772 764 31 999 0.36 1 139 400 1970 8 792 806 20 042 0.23 1 407 500 1971 8 831 036 38 230 0.43 1 981 300 1972 8 888 628 57 592 0.05 2 436 400 Source: National Statistical Service of Greece, Statistical Yearbook (Athens, 1974). for 15 000 tons in 1972, and veal production accounted for 76 000 tons in the same year. Thus, major increases in demand for both beef and veal are expected to come from an increase in per capita income. Given that income elasticities for both domestically produced and/or imported beef and veal are high (5.90, 3.75 and 1.18, 0.10 for domestically produced beef and veal and for imported beef and veal, respectively), it is expected that this will cohtinue to be the major source of an increased demand in the years to come. 235 Given the facts that in 1972 consumption exceeded production by 45 000 tons (which accounts for a 68.8% level of self-sufficiency for Greece) and that income import elasticities were +1.18 for beef and +0.10 for veal, means that Greece would continue to rely heavily on imports of these two products to fulfill domestic demand. It is interesting, though, to look at Table 31 which gives per capita total meat consumption and beef and veal consumption in Greece as well as in some West European countries in 1960 and 1972. It is worth noting that per capita total meat consumption in Western Europe increased from 54.1 kg in 1960 to 72.8 kg in 1973, an increase which represents Slightly over a one-third rise. If beef and veal are considered separately, it is seen that per capita consumption only increased from 18.3 kg per head in 1960 to 22.2 kg per head in 1972. When beef and veal are considered as a percentage of total meat consumption, it is shown in Table 31 below that beef and veal's share of the total consumption fell more than 3 percentage points for all of whole Western Europe over the twelve year period covered here. The explanation lies in the following three reasons a) consumer expenditure for beef and veal reached a high level; b) competition was present from other meats with lower prices, especially poultry; c) a different pattern of consumption was established by different consumers in the different European countries. From Table 31 it can be concluded that Greek consumers will still continue to increase their consumption of beef and veal since 236 TABLE 33 PER CAPITA TOTAL MEAT CONSUMPTION AND BEEF AND VEAL CONSUMPTION, WESTERN EUROPEAN COUNTRIES, 1960 AND 1972, IN kg Country Consumption 1960 1972 Total Beefl Beef & Veal PTOtal Beef Beef & Veal & as % of 8 as % Meat Veal Total Meat Meat Veal Total Meat Belg.-Lux. 63.6 23.0 ' 36.2 83.6 26.9 32.2 France 76.4 27.4 35.9 96.3 28.5 29.6 W. Germany 65.1 19.7 30.3 86.9 23.4 26.9 Italy 31.6 13.6 43.0 61.8 25.4 41.1 Netherlands 48.1 17.6 36.6 60.9 18.0 29.6 Denmark 62.4 16.2 26.0 63.5 16.2 25.5 Ireland 62.1 15.2 24.5 86.0 19.6 22.8 U.K. 71.6 24.6 34.4 77.0 22.5 29.0 EEC-9 60.6 20.9 34.5 79.4 24.3 30.6 Greece 23.5 4.9 20.9 54.0 16.9 31.3 Western1 Europe 54.1 18.3 33.8 72.8 22.2 30.5 1Including Ireland. Source: Organization for Economic Cooperation and Development and Foreign Agricultural Service and Econ. Res. Service (U.S. Dept. of Agri.). Adopted from: Western Europe's Beef Production, Consumption and Trade. E.R.S., U.S.D.A. ERS-Foreign 367, September 1974, p. 15. 237 the domestic (Greek) consumption level for total meat was around 54.0 kg/capita in 1972. This means that Greek consumers will be demanding beef and veal as their incomes rise until they at least reach the EEC-9 consumption level which in 1972 was 24.3 while that of Greece was 16.9 (almost 2/3 of that of EEC-9). But this means I that for beef and veal consumption to increase the total meat con- sumption has to increase following the income increase. The OECD 1985 projections for total meat consumption of 55.3 kg per capita was easily surpassed by Greek consumers in 1972 which means that, except for income, there have been other factors which influence consumption of meat, the most probable one seeming to be relative prices. Trends in Domestic Demand and Supply of Feed-Grain The quantity of feed-grain demanded largely depends on its relative price, the prices of livestock products, total domestic production of feed-grain, the rate of its utilization both inside and outside the livestock industry and the composition of the live- stock industry. This last factor plays a crucial role in feed-grain demand. It is well known that for one kilogram of red meat to be pro- dUced higher quantities of feedLgrain are needed than the amount required to produce one kilogram of pork or poultry meat. This fact has begun to be realized the last few years in Greece, and Table B-1 in Appendix 8 reveals that production of pork and poultry meat almost doubled between 1969 and 1974. This doubling means that policy makers - through advertisement or various kinds of policy supports - 238 are trying to divert meat production from red-meats to other meats as well in order to meet increased demand for meat in Greece and, at the same time, to reduce the quantities of feed-grain required to produce the meat demanded. Feed-grain, and particularly corn, greatly contributed to the financial burden which Greece carried all the years of the sample period. Imports of corn had a low peak of 96 000 metric tons in 1961 but had increased since then to reach a high peak of 439 000 metric tons in 1971. In money terms, this meant that Greece had to pay 55.7 million dollars in 1961 and $29.3 million dollars in 1971 only for imports of corn. Equations (5-11) and (5-12) in Chapter V gave an elasticity of the quantity of imported feed-grain with respect to animal units fed with feed-grain in the range of +0.2 to +0.3. This elasticity means that as the animal units increase by 1 percent imports of feed-grain quantity demanded will continue to increase by +0.2 or +0.3 percent in the same direction. The production, consumption, imports and self-sufficiency rate for feed-grain and for the years 1965-1972 are given in Table 32 below. Table 32 reveals that the self-sufficiency rate in feed- grain was 65 percent in 1965 and increased to 95 percent in 1972, and, if the latest data received from the National Statistical Service of Greece is correct, imports of corn reached a level of 960 000 metric tons in 1974 and a level of 464 000 metric tons in 1975; Given the rates of feed-grain utilization and the import elasticity of demand 239 TABLE 34 PRODUCTION, CONSUMPTION, IMPORTS AND SELF-SUFFICIENCY RATES FOR FEED-GRAIN IN GREECE: 1965-1972 IN THOUSAND METRIC TONS OF TDN AND IN PERCENTAGES Year Feed-Grain Production Consumption Imports Self-Suff. Rate 1965 1 108 l 612 504 .68 1966 2 010 2 366 356 .85 1967 2 180 2 670 490 .82 1968 l 610 l 963 353 .82 1969 l 916 2 304 388 .83 1970 1 845 2 615 770 1 .70 1971 2 452 3 221 ‘ 769 .76 1972 2 701 2 829 128 .95 Source: Calculated by the author. -fln~feed-grains,it is expected that feed-grainswill continue to be imported in Greece. As far as milk is concerned, the Agricultural Bank's estimates show that milk production from cows will reach a level of 900 thousand tons in 1980, while its consumption will be at the level of 715 thousand tons, thus creating a gap of 135 thousand tons by that year. Thus, the prospects of increasing milk production look good, assuming that other things will remain unchanged. 240 The Model as a Whole In the preceding analysis the individual models have been specified and estimated, the aim of the specific objectives of this thesis. In what follows in this section the individual models which refer to each particular product are put together, and, thus, a sub- system is formed. Each subsystem is thus formed from a demand equa- tion, the two components of the supply equation, an import demand equation* and, finally, an identity equalizing supply and demand. The direct supply equation is given as an outside equation, which provides a separate piece of information and, in particular, the derivation of the direct supply model. Hence, for the five products investigated here five sub- systems were formed, all of which were linked in such a way as to make up the whole system of the feed-grain-livestock economy in Greece. It should be clarified here that in the preceding analysis the basic theoretical model for supply and demand was of the form: 30 _ __t_ _ QtTBoislPt+Ut B03°:aPt<0 (711 instead of the actual one: 3Pt * In the case of milk and roughage, no import demand equations are given since, by assumption, no imports of these two commodities have taken place. 241 This reversibility of estimation was done since the estima- tion of price elasticities was among the major elements of the specific Objectives of the study. The actual model, however, is the one given by (7-2) when single-equation models are utilized. But it is a simple matter to transform model (7-1) to model (7-2). 1 Karl Fox1 has claimed that single-equation models are valid for perishable and other agricultural products, livestock products included. The argument about the use of these equation models runs as follows. For many agricultural products the quantity supplied in the market for these commodities is predetermined from some previous pro- duction period, since the quantity is partly dependent on factors determined in a previous time period when the production activity was initiated. The process of producing an agricultural commodity is assumed to take place in an irreversible manner with each step de- pendent on previous predetermined stages. The quantity that reaches the market depends, then, on some predetermined variables (exogenous variables). If the above mentioned argument is accepted, it follows, then, that the price of the commodity is determined by the quantity Offered in the market, especially in a perfectly competitive world and under the assumption that no storage of supply or other means of withholding supply are in operation. 1Karl A. Fox, EconOmetric Analysis for Public Policy (Ames: Iowa State College Press, 1958)} pp. 75 and 105. 242 When (7-2) is used and when Qt is predetermined, cov(QtVt) = 0, and the OLS estimates can be shown to be unbiased and consistent. In the case where model (7-1) is used, Pt has to be assumed predetermined and cov(PtUt) = 0; then, the OLS estimates are unbiased and consistent. This assumption can be established here if Pt is expressed as a function of other predetermined vari- ables, and, when so estimated, can be asserted as such in (7-1). Only under this assumption cov(PtVt) = 0 holds; otherwise serial correla- tion problems arise. Parenthetically, to be more specific about this last argument, and within the context of this research, a problem arising about the farm price of the products is examined here.' Farm price here has been estimated either fin terms of quantities consumed at the retail level (which are taken to be predetermined) and in terms of income and prices of other commodities which are also predetermined or in terms of lagged prices. This estimation will be more clearly seen later when the discussion on the system is undertaken, and the actual relationships are given. 1 Equation (7-2) above expresses a demand relationship with g pgjp[i_restrictions based on both economic theory and knowledge of the-industries involved. It should also be noted here that the assumption which holds in (7-2) strictly requires that the direction of functional (causal) dependence is from Q to P and not vice versa, while the assumption made in (7-1) requires that the direction of functional dependence is from P to 0 which obviously is not the same. 243 Even though the argument on which model (7-2) is based per- mits the use of the OLS estimation procedure, it must be said that these models recognize the interdependence between demand and supply relations but impose g pgjgg_restrictions on their relative disturbance variances. These assumptions make the OLS method for estimating the demand equation applicable without any explicit recognition of its interdependence on the supply function in the form of simultaneous equation systems. In what follows the structural equations in their general form are given. Variables in the model are given in a "free form" which means that those variables included and/or others as well could be taken to fit the final equations of the system. Dots within parentheses denote that there may be other variables, not specified here, which can be put in as exogenous variables to solve for the unknown variables. The Model, the Subsystems and the "Best" Equations Fitted Equations contained in the model along with each subsystem which they form are presented below. Endogenous variables are identified by asterisks(*). A. Roughage Submodel 1) Domestic Demand for Roughage * . 0R0t = f1(AUFt, PRt, P8t_], PVt_], P(L+M)t_1),...) (7-3) 2) Number of Acres Planted * , NSTRRPLt = f2(NSTRRPLt_], FPRt_], PFGt_], DVr, K,...) (7-4) 244 3) Yield per Acre YRSTR: = f3(PR PFERT, DVr, K,...) (7-5) t-l’ 4) Domestic (Total) Supply of Roughage 0R5: = f4(f3 - f2) (7-6) B. Feed-Grain Submodel 5) Domestic Demand for Feed-Grain * 0F60t = f5(PFGt, AUFt, PBt_1, PVt_], P(L+M)t_],...) (7-7) 6) Number of Acres Planted * NSTRFGPLt = f6(NSTRFGPL PFGt_], th-1’ Kt’ PFERTt_],...) t-l’ (7-8) 7) Yield per Acre * YF6t = f7(T, K, PFGt_],...) (7-9) 8) Domestic Supply of Feed-Grain QFGS:* s f8(f76 . ti7) (7-10) 9) Import Demand for Feed-Grain _ d d QFGIMPt - f9(PFGt, (PFG /PFGw)t, AUFt, GFGPt_],...) (7-11) 10) Total Supply of Feed-Grain TQF65t s flO(fl8 + +19) (7-12) C. Beef Submodel 11) Domestic Demand for Beef 080: = fll(PBt, P(L+M)t, PDI,...) (7-13) 12) 13) 14) 15) 16) D. 17) 18) 19) 20) 245 Import Demand for Beef * QBIMPt = f12(PBt or PBd/PBw, P(L+M), POI,...) Beef Cattle Slaughtered * BCSLt = f13(BCSLt_], PBt-i’ PVt, (PB/PFG)t_i, PFG Dressed Weight per Head _ * YBCt = f14(DVb, T,...) Domestic Supply of Finished Beef d t* a f15(f3 - f4) 085 Total Supply of Beef 0851 a f16(f12 + +15) Veal Submodel Domestic Demand for Veal * QVDt = f17(PVt, PBt’ P(L+M)t, POI,...) Import Demand for Veal QVIMP: = f18(PVd/PVw, P(L+M), POI,...) Veal Calves Slaughtered VCSL* = f19(VCSL t PV _1, PBt, (PV/PFG)t_i, PFG t-1’ t Dressed Carcass Weight per Head ch: = f20(DVv, T,...) t-i” t-i" (7-14) ..) (7-15) (7-16) (7-17) (7-18) (7-19) (7-20) ..) (7-21) (7-22) 246 21) Domestic Supply of Finished Veal d t* a f21(f7 + f8 - f9) (7-23) QVS 22) Total Supply of Finished Veal QVSI 5 f22(f21 + +12) (7-24) E. Milk Submodel 23) Domestic Demand for Milk QMKD: = f23(PMKt, PCHS PBTR, POI,...) (7-25) t, 24) Number of Cows Milked , * NCMt f24(NCM PMK PB -1’ (PMK/PFG)t_i,...) (7-26) t-T’ t-l’ t 25) Cow Milk per Head YCM: = f25(DVm, T,...) (7-27) 26) Domestic (Total) Supply of Fluid Milk d* . QMKSt a f26(f12 - f13) (7-28) In each quantity-demand equation variations in per capita consumption are assumed to be accounted for by changes in the retail own-price of the product, the prices of substitutes and the personal (per capita) disposable income of consumers, all of which are taken as given. 1 A retail price equation was estimated for beef, veal and milk which reflected the relationship between retail and farm prices. In addition to the respective prices and per capita consumption and disposable income, the wage rates in the meat processing induStries 247 were utilized in an earlier formulation of equations to proxy the farm and the retail price spread, i.e., the marketing margins. Import demand equations for beef, veal and feed-grain are thought to be influenced by farm prices of these products in the domestic market, by the price differential between Greece and world markets, by per capita domestic consumption and by an income variable. The retail price of lamb and mutton, which was mainly used here in this analysis as.the price substitute meats, was considered an exogenous variable for the feed-cattle economy, although from the standpoint of the entire livestock economy it may be treated as endogenous. Equations (2) to (4) in A submodel, (6) to (8) in B submodel, (13) to (15) in C submodel, (18) to (20) in D submodel and (22) to (24) in E submodel describe the supply situation prevailing in the roughage, feed-grain, beef, veal and milk industries correspondingly. The last equation of the above mentioned equations expresses the domestic supply of the five products considered here. As far as beef and veal are concerned, supply is considered as being the equivalent Of the weight of beef and veal which arises from inspected slaughter. No attempt has been made to capture uninspected slaughter. The domestic supply of beef, veal and milk is considered as the product of the number of cattle slaughtered or milked and the average cold carcass weight or the average per cow fluid milk delivered by farmers to milk processing centers. The supply of the finished product in the case of feedstuffs is treated as the product 248 of the number of acres planted times the average per acre yield of roughage and feed-grain. All of the variables which enter in the supply of feedstuffs equations are either lagged endogenous or truly exogenous to the system. The number of cattle slaughtered was explained mainly by the beef-feedgrain price ratio in t-l year and the beef cattle inventory at the beginning of the year. The beginning inventory is that re- ported by the National Statistical Service of Greece on December 31. Therefore, the inventory on December 31 in the year t-l was taken as the value of the beginning year inventory variable for the year t. Early trials with this equation indicated that the inclusion of any variable other than the current beef cattle inventory would not re- sult in a significant improvement in the coefficient of multiple determination. It was therefore concluded that for the cattle slaughter equation the inventory variable makes a substantial contribu- tion toward explaining its variation. A similar conclusion was reached for the number of cows milked. The beginning inventory of cows milked was indicated to be among the variables which greatly contributed to explaining the variation in the supply of cows milked. The dressed cold carcass weight per animal is influenced by factors having to do with improvements in the composition of the national herd and a subsidy variable. In the domestic supply equations of finished beef the animal units at t-l year were used, and the average farm price of beef in 249 (t-l + t-2)/2 years was tried in one formulation, while in a second one the quantity of beef supplied in t-l year was utilized along with the farm price of feedstuffs in t-l year. Generally, there were two equations in the form of an identity in each submodel, except in the fluid milk and roughage submodels where one identity was used since there were no imports for these two commodities. The first identity equation gives the domestic supply of a finished product, which is a product of the number of cattle slaughtered times the dressed weight per animal in the case of beef production and the number of cows milked times the yield of milk per cow. The second identity equation in each submodel gives the total supply of finished product, with imports taken into account. The "best" equations chosen to test whether a significant amount of variation in the endogenous variables was accounted for by exogenous variables in each equation are provided below in a sub- model division manner for the purpose of clearer presentation. A Roughage Submodel 1) Demand for Roughage* QRDt = 603.2417 - 447.7941 FPRt + 0.5638 AUF + 6.3321 T (198.1476) (68.502) (.0312) (1.984) R2 = .98 o"r = 1.96 (7-29)" *Demand relationships are given as quantity of product consumed per unit (human or animal); to find the total demand multiply this quantity times the population (human or animal). ** In these equations here the standard errors of the coefficients are given in parentheses underneath the values of the coefficients. ' (continued on page 250) 250 2) Number of Acres Planted NSTRRPLt = -2458.0452 + 1229.4023 FPRt_1 + 1.1434NSTRRPLt_1 (1040.521) (423.820) (.2542) R2 = 0.98 o* = 2.84 (7-30) 3) Yield per Acre = 42.4423 + 0.347QRP + 0.70570FERT + 15.0297 T (31.699) (.057) (.5044) (6.443) YRSTRt R2 = 0.94 6* = 1.62 (7-31) 4) Total Supply of Roughage QRSt = 201.4080 + 520.1313 FPRt_] + 0.8792 ORst_] (402.6275) (133.980) (.0928) RZ = 0.98 6* = 1.86 (7-32) B. Feed-Grain Submodel 5) Domestic Demand for Feed-Grain QFGD = 912.6617 - 15.3842 FPFG + 0.5953 AUF + 1.8605 T 1 (18.695) (3.648) t (.089) 1'1 (.632) R2 = 0.97 8* = 1.86 (7-33) 6) Numbers of Acres Planted NSTRFGPLt = 2458.0452 + 1229.4023 FPRt_] + 1.1434 NSTRFGPLt_1 (1040.5214) (423.8201) -2 1 * R = 0.98 d = 2.41 (7-340 (continued from page 249) RE: Coefficient of detenmination corrected for the degrees of freedom. d : Durbin-Watson statistic for serial correlation. d*(i): incon- clusive test. Loosely speaking, values of d* aro nd 2 indicate no (first order) serial correlation, and Values of d around and below 1 indicate the presence of statistically significant positive serigl cor- relation in the residuals. The test is often inconclusive for d -values between 1 and 2. 251 7) Yield per Acre YFGt = -107. 8157 + 14.3567 T + 0.5633 K + 16256.173 FPFGt_ 1 (93.140) (2.001) (.453) (9104.1) R2 = .84 6* = 1.60 (7-35) 8) Supply of Feed-Grain QFGSt = -1296. 8261 + 0.6190 AUF + 152 090.4849 FPFGt _] (622. 385) (.096) (49 146.136) 22 = 0.70 6* = l.34(i) (7-36) 9) Import Demand for Feed-Grain QFGIMPt= 30. 7171 - 0.1748 QFGPt_ 1 + 0.2679 AUFt ( 945) (. 089) (.059) -2 * R = 0.54 d = 2.11 (7-37) C. Beef Submodel 10) Domestic Demand for Beef QBDt = 6.0049 - 20.5973 RPB - 0.3270 RP(L+M) + 0.1975 GNP (1.5218) (10.376) (.048) (.016) R2 = 0.90 6* = 1.94 (7-38) 11) Import Demand for Beef QBIMP = -5. 0967 + 0.8024 QBC + 174. 974 (FPB/PIMPB)t t (.899) (.068) t (55. 442) + 12.5456 FP(L+M) (2.232) * = 0.96 d = 1.26 (7-39) = ((1951),, .,22(1972). 252 12) Beef Cattle Slaughtered BCSLt = 26.3955 + 0.8999 BCSL + 0.2113 (FPB/FPFS) ‘ (27.523) (.151) 1‘1 (.805) 1‘1 R2 = 0.70 6* = 1.95 (7-40) ’ 13) Dressed Carcass Weight per Head YBCt = 69.9434 + 16.9888 0vb + 5.7429 T (6.735) (8.733) (.505) R2 = 0.87 3* = 1.01 (7-41) 14) Supply of Finished Beef OBSt = 0.4089 + 1.1278 OBst_1 - 0.3182 FPF5t_1 (.301) (.141) (.021) R2 = 0.98 6* = 1.67 (7-42) 0. Veal Submodel 15) Domestic Demand for Veal QVDt = -6.5543 + 4.5107 RPV + 0.3689 RP(L+M) + 0.0632 GNP (1.850) (14.3973) (.093) (.039) Rz = 0.96 d* = 1.32 (7-43)_ 16) Import Demand for Veal QVIMPt = -3.0645 + 2.4357 RPV + 8.1174 RP(L+M) + 0.0005 GNP (1.055) (10.128) (.298) (.005) -2 * R = 0.89 d = 1.62 (7-44) 17) Veal Calves Slaughtered VCSL = -136.7922 + 0.8916 NCM t + 0.6497 (FPV/FPFG) (65.644) (.204) 1'1 (.286) 1‘1 R2 = 0.87 8* = 1.89 (7-45) 253 18) Dressed Carcass Weight per Head ' cht = 25.3818 + 43.3569 DVr + 5.933 T ' (7.785) (10.7712) (.864) -2 'k . R = 0.97 d = 1.78 (7-46) 19) Supply of Finished Veal qu = 15.2263 + 0.7537 ovst_1 - 0.6824 FPF5t_1 (5.179) (1.742) (.179) §2 t * = 0.96 d = 1.87 (7-47) E. Milk Submodel 20) F1uid Milk Demand QMKDt = 79.7307 - 2 708.7250 RPMKt + 26.8157 RPCH + 0.2238 GNP (31.953) (1 012.245) (10.429) (.117) R2 = 0.87 d* = 1.07(i) (7-48) 21) Number of Cows Milked NCMt = -27.1957 + 0.962 NCMt_1 + 2155.5521 FPMKt__1 (104.104) (.105). (3039.587) R2 = 0.96 6* = l.54(i) (7-49) 22) Average Fluid Milk per Cow YMCt = 728.4024 + 138.5763 DVm + 19.8727 1 (26.048) (36.038) (2.890) -2 * R = 0.87 d = 1.67 (7-50) 23) Total Supply of Milk QMKSt = 235.5269 + 0.6348 QMKSt_1 - 12.968.879 FPFSt_] (7.834) (.197) (605.312) R2 = 0.98 6* = 1.94 (7-51) 254 Projection of the Feed-Grain-Cattle Economy The I'best" equations derived in the previous section can be' uSed to predict the future state of the feed-grain cattle economy in Greece. ’However, before such a prediction can be made, the estimated system must be manipulated in such a way that a unique value for each endogenous variable can be predicted. The procedure is to ob- tain the relevant values of the independent variables for the forecast period which is specified (say, 1973 here), insert them in the equa- tion, and compute the dependent variable. Since data for the independent variables are not available at this time, a kind of an e§_pp§t_forecasting test will be tried here. More specifically, the procedure adopted here is to omit the last period's observations from the original fit, and then the equa- tions were forecasted using the observed but not included observa- tions. This procedure has the advantage of providing an immediate test of the equation's forecasting ability. When an equation is part of a recursive system its sequential nature is helpful in obtaining estimates of independent variables in successive equations. An example here would be that next year's supply may depend on preceding years' prices and/or costs. Thus, using the values of these prices and/or costs, one could forecast supply. The forecast level of supply, then, could be used to fore- cast price in the next period. And, since the system of equations 255 fitted inthe feed-grain cattle economy can be viewed as a recursive system, it is a matter of simple manipulation to make such forecasts. In Table 33 below e§_pp§t predictions are given for the 23 equations of the system. The 1971 values were used for the endogenous values to test the predictability of each equation fitted. Generally speaking, the equations gave rather satiSfactory predictions. But, since only one year's values for each endogenous variable were tested, it could be that predicted and actual values are close by chance. The system was also tested for values of the three last years (1971-1973). The results of such a test are shown in Table 34. In this new test Theil's.inequalitycoefficient was employed since it is believed that in order for forecasting procedures to be verifiable they must be "based upon theoretical consideration-- however simple--and on empirical observations obtained beforehand-- n2 however scanty and crude. The prediction efficiency of each equation was thus tested using Theil's v coefficient which is given by the following formula: v =\/gPi r: Ai)2/( /z(§i)2 H /£(:i)2 Pi : predictable value where Ai : actual value n : number of observations. 2Henry Theil, Economic Forecast and Policy (Amsterdam: North Holland Publishing Company, 1965), p. 14. 256 TABLE 35 PREDICTED AND ACTUAL VALUES OF THE ENDOGENOUS VARIABLES FOR THE GREEK FEED-GRAIN CATTLE ECONOMY, ANNUAL AVERAGE 1972 AND PREDICTED 1972 fit Unit Value 1972 Value 1972 Absolute Variable Average Average Difference Actual Predicted (+) In- crease (-) De- crease 1972 1) QRD, per animal kg TDN 2682 2751 + 69 2) NSTRRPL 000 str 3346 3696 + 350 3) YRSTR kg/str 635 720 + 85 4) QRS 000 tons TDN 2125 2122 - 3 5) QFGD, per animal kg TDN 2682 2897 + 215 6) NSTRFGPL 000 str 6576 4399 ' -2177 7) YFG kg/str 411 337 - 74 8) QFGS 000 tons TDN 2701 2257 - 443 9) QFGIMP 000 tons TDN 128 667 + 539 10) QBD kg/head 3.73 3.76 + 0.03 11) QBIMP kg/head 2.05 3.13 + 1.08 12) BCSL 000 head 76 73 - 3 13) YBC kg 191 206.5 + 15.5 14) 085 000 tons 15 16.20 + 1.20 15) QVD kg/head 11.23 19.43 + 9.20 16) QVIMP kg/head 2.3 2.4 + 0.1 17) VCSL 000 head 257 265 + 8 18) YVC kg/animal 208 193 - 15 19) QVS 000 tons 76 ’ 72 - 4 20) QMKD kg/head 61.43 41.12 - 20.31 21) NCM 000 head 427 469 + 42 22) YMC. kg/animal 1279 1287 + 12 23) QMKS 000 tons 546.1 471.2 - 74.9 257 .mo.o o.mmm o.o~m o.mom F.o¢m o.Pvm m.mnm mcou ooo mxzo Po.o o.FNmF N.ooNF m.om- o.omN_ o.PoNF o.oom— Fmew:m\ox ozw No.o m.mm¢ N.—m¢ m.om¢ o.-¢ o.o~¢ o.ove one; ooo 2oz Fo.o o.~o m—.¢o oo.mo m¢.po om.Fo N¢.mo ommc\ox oxzo mp.o m.oo «.mm m.o¢ o.on o.~n o.¢n mcou ooo m>o mo.o o.nop o.~op N.oo_ o.oo~ o.oo_ o.oo_ Pmavco\ox o>m co.o m.oo~ o.omm n.mm~ o.mmN o.¢v~ o.o~m one; ooo 4mo> ¢N.o m.m m.¢ ¢.m oo.~ eo.~ No.m umm;\ox mzH>o mp.o o.o~ o.ep o.PP mm._F oo.op mo.mp cmmg\ox o>o mo.o o.¢F o.¢F m.mp o.m~ o.ep o.¢F mcou ooo moo mo.o o.oo~ F.No~ m.mo_ o.—op o.omp o.¢np Fmevcmxox oow. mo.o _.No o.mo o.vo . o.on o.Po o.No one: ooo omoo mp.o o¢.¢ n~.e oo.m mo.N o¢.~ m~.¢ omm:\ox azamo Pm.o mo.o No.o mm.m mn.m oo.¢ No.m vmm:\ox ooo mo.o o.omo o.moo o.omm o.o~_ o.oom o.omn zoh mcou ooo mzHomo op.o m.mP—N o.~mop m.~mop o.Fom~ o.~m¢~ o.mvop zoh mcop ooo momo mo.o o.omm m.n~m o.m~m o._p¢ o.oom o.oom Lum\ox omm Fpao o.mooo m.oomm ~.moom o.ommo o.mmmo o.ooo¢ gum ooo oaoumhmz n~.o «.moom m.npmm o.—oNN o.moo~ o.no~m o.n¢nm zoh ox oomo No.o m.N¢NN o.ooNN m.noo~ o.mmpm o.~mFN o.oo~N zoh mcop ooo mmo oo.o m.¢mm N.o¢m o.P¢m o.mmo o.~mo o.omo Lum\ox mhmmw Po.o o.oo_m v.¢m~m N.oopm o.o¢mm o.oo~m o.~¢~m gum ooo Jammhmz oo.o m.m¢o~ o.mmm~ o.o¢o~ o.vFoN o.oo_N o.¢FNN zoh ox omo zuwpmscmcm «mo— PnoF omop mum? pump onop pmco mpnmwgm> m.mech vmpownwgm wzpm> szpu< HzmHonmmoo >h~o<=0mzH m.4~mzh oz< .Nmiomop mm mzh mom >zozoom mohh moozmooozm mxk no mmoo<> 4<=ho< oz< omhomomxm om m4o< mmuHmm mmmp h< m ucm pasta mumuipwo mum“ U-: smwm “66: pmmcmo vmmcm covenanmcou coon Papa» -305 TABLE B-6 MEAT IMPORTS IN THOUSAND METRIC TONS SELECTED YEARS. GREECE, 1962-1972 Beef Mutton Poultry Pork Total Meat Year and andv Veal Lamb 1962 17 1 10 2 30 1963 28 1 l5 6 50 1964 22 l 20 6 49 1965 31 2 32 ll 76 1966 31 - 34 11 76 1967 37 1 35 12 85 1968 45 4 33 8 90 1969 55 l 36 6 98 1970 68 - 40 3 111 1971 51 - 60 3 114 1972 45 - 46 3 94 Source: 0.E.C.D., "Meat Balances in OECD Member Countries," (Paris, January 1974). 306 TABLE B-7 INDICES 0F LIVESTOCK AGGREGATE OUTPUT GREECE. BY PRODUCT GROUPING, 1954-63 (1957/59 = 100) Product Grouping Total Year livestock Cattle Sheep_ Goats Hog§__ Poultry Rabbits Productivity 1954 81 92 85 98 83 132 87 1955 91 100 98 100 98 130 95 1956 96 97 93 95 100 108 96 1957 96 95 93 102 97 112 98 1958 103 101 98 95 100 100 100 1959 101 103 110 103 102 120 101 1960 106 102 103 105 101 132 103 1961 120 107 108 120 106 160 112 1962 138 110 115 133 113 124 126 1963 150 110 116 139 116 128 134 1954/56 89 97 92 ' 98 94 122 93 1957/59 100 100 100 100 100 100 100 1961/63 136 109 113 130 112 124 125 Source: Lawrence H. Shaw: Production (Athens, 1969), p. 219. Postwar Growth in Greek Agricultural 307 .o«« .a .Amomp .mcwsuv4 Pouch «. 1 «.m- ”.m- 1 1 1 «.P- I m.m- «.1 «.1 mpwaamm m.« w.F m.~ P.« F.F «.m m.« m.« m.m m.~ «.m m.« «pupae; «.e o.¢ ¢.m m.« F.m m._ «.F «.c m.« o.m m.m «.o moo: o.m o.« P.m «.P o._ «.o o.~ m.m e.« m.P F.¢ m.¢ mpmou m._ e.p «.5 m.~ m. m. m.« «.F o.m «.P o.« w.« amwcm m.m m.o m.m o.¢~ ~.« o.PP ¢.op m.m e.m m.m m.m m.« mpuumu 4%..“ é...” 1w“. .3“wa 3%.. my «New 2%: Peach cowmmm pwzuoca buaoomm oz< mzowwmm >m .mummmc .Hamhao mpH4 zH zhzomw mo mmp pmuoh_ . co? mm Aoop u mm\nmmpv moivmmF monwmm >m .mummmw .hsmhzo mhH4 no mmumnzH mum u4m< u>< .mcou vcomao:u cw copuuauoca ”a .vcomaogu c. mocu< u< .mop .n .—~mp .m:~:u< .auuogw ma xooagmo> poupumpuoum .o.m.m.z A—V "cannon a«m.« Fae cup m«m.p map ow m—o.« m—s PvN mm«.p m « mmo.p onm._ mam . asap —~«.« «Fe ov— ¢F«.p «op em mam.p «we «w« m«p.p m m mom." m««.p mso.— mom— e«m.« «cm mew moc.— mm «m .mpc.p pse «mm ooo.— m w m«e.p mom.p mac.p wom— mmm.« m—n mnp mum.p mm— —pp mo«.« es“ pmm pap.— mp —p men.p omm.p pmo.— «om— smm Fmp m~m.~ m«« amp mpv.p «o— m—p «om.p «an em« o~o.— mp op emn.p o«o.« «m—.— comp ~oa our m««.p mem cep cm«.— amp o«p moc.p «mm mom ooo.~ o— op «vo.p «no.« wm«.— mom— pnm.p wwN o_« cop.F mew m«p m«m.p oq«. pwp mom «N a« «mv.p mme.— «vp.— com— 3 moo.Fi em« Mm« «wo.— amp mmp m—o._ mm« p—« «pm pm «0 «op.p 9pm., Timeo.r ommp mafia—Wa— LLWPUvOm K mwmmm gnom— >§ mNM—GZ <. FK mama—O K $< Nun—«Lam < E uh“ < >< any; {I Lu0> mm omhumgmm z~ mw<¢u>< oz< zouhuaoomm .