‘ I " 11‘93‘13‘5Jtfl. ‘J‘W [J‘fikkhfiofi "‘ :fIJifim'J‘A L -I I'm-e", I. ,IIIILIIIII . .. ”If? ‘ MP 3:;ch :JJ‘JIJJ EIJ, :J- ' ’1' “MI 4"; J ..... . ';’!fJ"I:'b JI‘ I 1“I,II‘I'I‘"I.IAI I:- I ( III IIIISIIJ: IIJI IIIII IJ’, ,IJIIHI '(IIII IIJ’III} :bCI-III, 1!;ng LJIJJJ IIJ‘ IIII'IIII I , . II III . ':\J' .I: L'JI‘I‘I-IJY. who”. 58'“ III JV 1’ ”‘J-h I“ ’J'J‘JIWI II'JJI JwJJHWJJJ 5w 34'} “9‘3? ’I~ WW «II. ‘ M" ‘J‘I‘ IIIIIIIIIII‘I I2: ,. . IJJIJJWHJ'JII. Iu1r\Ji;:1J-i II‘IJJI'IJJIIIIJI NIH III I” 1 ,. I“ I’i‘u‘ I4 .;'IJ:;' ‘5’ Tit-7mg- ”17““: L. L- .I. 2—: —$ M6:— 4" «we, I; 4.1: :45 13:??? .JL . I I I ““53“ ' ‘IJ‘J'JJ 'LdJé‘EJ'I‘I‘ II,“ ' 9;?! I', 7".“ ' '- ’ ”JV“ .' ,I , . [I 'I I; I ‘ . ,I 'I . . .I. II II. . .. ,. .IZ’II; . I... I"? . - s II- III I- I- III IIIII‘II IWII. , i 1. I"). II. W I \ I I, ”J, I I . 2 _ VI W. H , V (J I I I II\ - 1'“ {IV I . . - . Q . (1| IJLI L . J . ‘JJ'IIJ ”‘1': I‘( (IAIII EJ‘IJ‘I J mg" [3‘3 JJEI “Jul . M i. .1 {J I J I 'fi' 1““ fl h J, .. I III IJII'II Ji' I IWIJIIIILII {mud II J21“ IIJII {JIIIIII {FI‘IIWILIJI -.I . , 11‘ .' JI‘J‘I (I; «J . I I1 I I?" II.I JIJJJIIJII EJJJJII” ”I IIJI' IIiIJJ 1 JII‘IIIIJ JIIIJU’J {QT ‘ ‘I. 'II ’ In?“ . =qu I ) J , ’I 1 I‘ J" H IJ IWJ I I I' I I‘ A ' '. I'II IIIJJJ JJJII . I IJIIII I I II 2 :‘."2‘.£: VJ . .IIIIIIIII IIIIIr,I,.II I IIIIJI ”I IIII‘J IJJJI ‘ II|I ‘( "4 III- III"? III . . I I ‘JJ IIJI JI‘JI I.‘ IIIIII IIIJI :II‘J‘ .I, ‘JIJJIJI' 'L“: I I ::_::;’- ‘IJ ‘? '~ ‘9 ‘JJI I “I: .II J“ JI I 11" Id III-J'fIJLIII EJ‘J’ IRE: WE? IJ fAII; "Md“ ' - ”SJ IIIJ I 1”“ JJJIJ limb IJIJII'JJ IJJII‘J ”2““ W EJ‘III'IIIHJ‘ )I I I .HII' III M III 'I M, 2...: ..—~ M r." V I HJ' (iJ JIJI’ ’I'I‘J ‘1 I‘ ' mu.“ . III ‘11:?!“ JJ’IJJ‘ 'JJI' I ,v-I‘JI Ii .4. J‘JI. J 11‘3” JJ JIIJIJJJ ‘JZJILJJ‘IE I'JI , IIILTJI‘HIIII .4 (“"“'IJJI“ II . I II . J3 II I'EIJILfme JI'I IIIIIJ .5353. “’2' II‘JIII l‘ili ‘(14 M, ‘ I‘J " ' JIJJIIII“ JfIIII JJJ (It: tJJIII :I; llIIJIIIJI' I | I d"! I I III M 53:1: 5 ,L V II: '.‘_'J "I '. {- Iii}! ‘I’III. WJN II I") :7: —£* 5*"? 4";"1 v“. t’ 3&3: 21“" L.“ 'c.‘:_ ,N , :3? - 5-K; “P“ ;;L ’E‘Ei'vri'st '7 . 4‘32 7";er can ’94:: 323—9: 4 "it: ,L .2: “‘" ’ *5— .L__._. ._I._ , ' , i'hfgt: L. fake-’67. 0:2. ILS‘EL LII. - .‘ir‘fir 5:503“ __ A g TL"??? _ 35-: 21-5: » f:»-;-::.L W‘w::_rl.- '- L‘ ”V..- $- 5% ‘ ’55)“ ’ 9:4"? I,;: 5'. r:- L W «M. 1.7, .. . vgg. ,WILL, .I A ‘ “It: 1:; {is ; .4 M .L_L.L.._,G V-.. 9L1" r L” I-~.4. '..L_LL . é ‘3’??? "P L. $153? ,7" .73" 0.4.4;— — E y’ _ . . PL? .. .32.. Ls: F - - ~ "52:; agar 9.13.: . '1;-.‘5:‘” V J...“ L31" ‘4‘ 22.65“; V" + L if?” :7 L. .- __I d? 37' ..::.,I_ .r‘ ' . ‘ “I «a» ‘1‘?“ 4 :73: a «5-35 gar... ”.L‘rfi' . r; -.. . 7.3;: ‘2'" 43.... {.122- :32} I 5%? 4319:: .1: {23'- I}: fJJJIIJ {4‘ ‘3'? J ‘IJII‘IJI‘RLJ‘I‘ Imfi ‘ JJ‘JIJIJ!‘ IJJJIJM "3". J' .T‘IIJ I“ ‘ JIJIJI I u.- E;— v . a L. 55':- ~ 1.; L. . J '2 I L 1‘43. -- ' L < .' Lirnzigflg'} an N.)- ..J“— ‘uI %‘L-: . - .3; .52.; *3», LI . :13, “9‘53" an if: “1:33:14 . . 1 v — ‘7‘, - F “us-L. A .~ .3 .1' my}; .- 71' .’ I :1 f¥..é~‘:'::1hf.._~ . ~ a. LI . ,2”. _ : 5: t“ .8- -": o—a. ‘L-‘tfigfig‘ "11:12:?“- ._—‘>’:—?‘ .. 151.. k- «P . . . L . .L._..._. Q» ‘ L-vI _-:I . 53315 >. Ema 14,353 4.5.? L- ‘2. ,- w:;:=~?*~‘;. - ~I 1, W; ”9—. LIA-M: - .LL $5- -*v a.-. L, - ”,4: - --,_:\t: .3: if’A-VTMWY I - 4 ‘t. I «C‘ ”L: “1:? .3, 944* -tq~ L .4 “.7431, .6 ,L- .gés yr ”- . 4—,-.‘42 ~95" “VI: arctic. fig‘ ;‘ M . JI. WT.” .wv '. ‘ :- “ L:.L'.£9--£ “ I I I ' «" '.-. .v-5 ‘x ,4 "— 2‘. .LLLL r ' .~. ' '1'". -. 4: r4 237’" :33“ LL. _ _ L.. $33; ‘2. a _-4~5-:=‘< ' "‘~< ‘ . ‘—-:J-~l“2 '“A ‘5 ‘i‘§5?% {catgy' ’ .L I.,7,,L—\,I _ —-A. w #955” "— ”£533 ' a» I ..u m . . A as: 55:" .f’ I m .3 '4. . ' .... Iii-’1 w ‘ am cg»: mars-A .1: ’° 3... JEJ—Y; ‘"_ L I I "L: L' '3 Ir? It” 555% ‘ 41$. ~..~;:‘~:~.= ..._ «Jig-A3: I. AM». ~ .. .L ». Ln, _ :r‘ _ AIL . .55.; L ‘4'- __L -fi' ~ m NJ}? ;_.J. _ ._ :1" $53: '7‘. - _-—_—L: 3:5.“ J‘s-II“: II 1«"‘I I'I'III " -V. text. L‘. .III _J.__.I> .— 93'!“ “— .-L--I..-=--=—_, 3331.! fig»; .. ' ".I-J‘. .' :3 . L' ' I “Le-4:2. ‘ "‘7.“ 4 ‘th—{fi :13“: Riv)" A v ””3 LL... :: _ ,L V’V' ‘ - 4"?“ c is" L ‘g‘v— d ;.L-"-’ ' ', - ' .: gar-"'4‘"- - L _.,~‘ ‘12. ' '32::5FTLT- I, A L , _ . . I.‘ A J? YE; . .1 5 .II J I ifJierI " " I} a J :5 K‘J' m 2‘; "Jf‘w‘gj’fi in ‘NJI‘JJ .~ . 47— L...— 43L -3273 ‘ 55¢» ’tfijo‘n a: I; w . '2‘?’ ’ $1 ---.;&S‘5¢ 300 A187 MIrrjIIg SOURCES OF AGRICULTURAL MARKET INSTABILITY By Ian Lennox Dalziell A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1985 Copyright by IAN LENNOX DALZIELL 1985 ABSTRACT SOURCES OF AGRICULTURAL MARKET INSTABILITY BY ‘Ian Lennox Dalziell Agricultural markets are less stable than those in most other sectors of the economy. Indeed instability is often seen as the most pressing policy problem facing agriculture. This dissertation is dedicated to the examination of the meaning, extent, and sources of price and quantity instability. Particular attention is given to the role of marketing institutions in contributing to, or alleviating this instability, and to the relationship between instability and the effectiveness of marketing coordination. The analysis of the study is in three parts. First, an annual instability measure is developed for the measurement of a dimension of marketing coordination effectiveness. This measure is then applied to markets for more than 100 different commodities in order to distinguish and identify possibly poorly coordinated markets. Great variation is evident in the degree of instability among different commodities. In the second part of the analysis, a cross-sectional study is undertaken to examine the sources of agricultural market instability. It was found that, for annual crops, the greater part of production instability is due to potentially controllable factors rather than uncontrollable yield variation. In addition the contribution of various conventional supply and demand shifters are shown to make only a negligible contribution to total instability. In contrast, marketing institutions are an important explanator of differences in annual instability between commodities. Evidence is also provided that instability itself contributes towards further instability. The third part of the analysis considers recent changes in instability. It is found that most commodities have experienced increased instability since the early 1970s (especially price instability), but it is the field crops and animal products that have shown the greatest increases. Whilst macroeconomic factors are identified as a source of the increase, the wide differences between commodities and the explanatory power of other sources provides evidence that other factors also are important: in particular, the changes in Government support policies and institutional changes in fruit and vegetable industries. ACKNOWLEDGEMENTS I wish to express my sincere appreciation to those who have made my graduate studies possible. I am especially grateful to Dr. Stan Thompson for his guidance and personal help during my course work, and to Dr. Jim Shaffer for his supervision and encouragement of this dissertation research. I also wish to thank the other members of my guidance committee: Drs. Jim Bonnen, Harold Riley and Norm Obst, for their assistance. I am also grateful to the Australian Public Service and the Australian Bureau of Agricultural Economics for the financial support of a Commonwealth Public Service Scholarship to pursue this doctoral study. In addition, I thank the United States Agricultural Marketing Service for their support of this dissertation research. However, it is my family who have borne the daily brunt of my studies, most of all. I thank Rosamund for her constant encouragement, understanding, and hard work, without whom this would not have been possible; and our children Anastasia and Andrew. And thanks be to God. ii TABLE OF CONTENTS 1. Introduction The problem of instability Orderly marketing Instability and marketing coordination Study rationale and method Summary and plan of study 2. what is instability? Introduction Mathematical instability System instability Instability and market coordination Predictable and unpredictable instability Instability over time Conclusion 3. Historical approaches to instability in agriculture (.00) 1 2 3 4 \10‘ Introduction Traditional Marshallian approach Asset fixity theory Modern risk approaches 3.4.1 Risk and uncertainity 3.4.2 Risk in Economic models 3.4.3 Instability matters, even without risk aversion failure Perfect information Information and efficient markets Incomplete markets Non—competitive markets Market imperfections elsewhere . Other externalities 3.5.7 Concluding comments Market coordination approach Conclusions 3 ’1 wwwwwwo UIUIUIUIUIUIE O‘Ulothl-‘fi' iii Page U!».A(o+» H (D 10 11 14 16 17 19 19 19 23 24 25 27 30 31 31 32 34 35 35 36 37 38 40 4. Measuring Instability 4 4 4 4 4 .1 .2 .3 .5 .6 Introduction Instability measures used in the literature 4.2.1 Characteristics of instability measures The INS measure 4.3.1 Decomposition of INS 4.3.2 Characteristics of the INS measure Comparison of INS and CVT measures 4.4.1 Statistical analysis 4.4.2 Simulation study Deflating the price series Conclusions 5. The extent of instability among commodities (JIUIUIUIUIUIUIUIUIUIUI .1 .2 .3 .4 .5 .6 .7 .8 .9 .1 .1 Introduction Data Classification methods Aggregate index of instability Price and quantity instability Greatest quantity instability Greatest price instability Most stable commodities Recent instability 0 Comparison with other industries 1 Concluding comments iv 42 42 43 43 49 51 54 56 56 58 61 65 67 67 68 69 72 74 77 79 80 82 84 85 6. Sources of instability 6.1 Introduction 6.2 Sources of instability 6.3 Yield 6.3.1 Yield and instability 6.3.2 The relative importance of yield instability 6.3.3 Is high production instability mainly caused by yields? 6.3.4 Indirect effect of yields on production instability 6.3.5 Summary of yield effects 6.4 Demand and supply factors 6.5 Institutional factors 6.5.1 Annual versus perennial production 6.5.2 Processing 6.5.3 Price supports 6.5.4 Gross value of production 6 5.5 Futures markets 6 5 6 Marketing orders 6 5 7 Long run changes in prices and quantities .5.8 Trade . .9 Simultaneous model 6.5.10 Conclusion 6.6 Price instability as a source of area instability 6.7 Summary and conclusions 050‘ 87 87 89 89 91 99 100 101 102 103 109 115 116 117 118 119 121 124 125 125 128 130 133 7. Changing Instability: extent and sources 135 7.1 Introduction 135 7.2 Has instability increased? 136 7.2.1 Changes in aggregate market instability 136 7.2.2 Increases in instability in commodity groupings 138 7.2.3 Increases in instability by commodity 142 7.3 Explanations for increasing instability 146 7.3.1 Increased export demand 146 7.3.2 Less stable international monetary policies 147 7.3.3 Flexible exchange rates 149 7.3.4 Changes in US price support arrangements 150 7.3.5 Unusual weather 150 7.3.6 US export embargos 151 7.3.7 Narrowing of the genetic base for grains 151 7.3.8 The transfer of risk to farmers 152 7.3.9 Expansion of production on to marginal lands 153 7.3.10 Growth in protectionist pressures 153 7.4 Empirical analysis of sources of increased instability 154 7.4.1 Factors behind the increase in quantity instability 156 7.4.2 Factors behind the increase in price instability 160 7.5 Summary and concluding comments 163 8. Conclusions and implications 166 8.1 Rationale for study and approach 166 8.2 Major findings 168 8.3 Some implications 171 8.4 Future research 172 Appendix A: Commodity codes and data coverage 178 Appendix B: Definitions of variables used in tables and regressions 182 Appendix C: Instability measures by commodity 186 Appendix D: BASIC program for simulation study Appendix E: A comment on Myers and Runge’s article on supply and demand contributions to U.S. corn market instability Appendix F: Derived data for the analysis of the study Bibliography 200 vi Table Table Table LIST OF TABLES Page Characteristic Statistics of Components of Index 72 Commodities Grouped by Quartiles According to Composite Instability Index 73 Unstable Commodities: Aggregate Index 76 Quantity Instability: INS and CVT Upper Quartiles 78 Price Instability: INS and CVT Upper Quartiles 79 Stable Commodities: Quantity and Price Upper Quartiles 80 Changes in Representation of Upper Quartile in Recent Period 83 Commodities in the Top Instability Quartile for Both Periods 83 Instability in other Economic Sectors 85 Decomposition of Production Instability among Yield, Area and Interaction Components 93 Percentage Decomposition of Production Instability 94 Summary of Decomposition of Production Instability 95 Sources of Quantity Instability: Reduced Forms 112 Sources of Area Instability: Reduced Forms 113 Sources of Price Instability: Reduced Forms 114 Sources of Area Quantity and Price Instability: Two-Stage Least Squares Estimation 127 Mean Instability Measures for Three Periods 137 Changes in Quantity and Price Instability by Commodity Type 139 Mean Proportional Changes in Instability for Commodity Groups 141 Proportional Change in Instability by Commodity 143 Commodities with Highest Increases in Price and Quantity Instability 145 Regression Coefficients Explaining Increase in Quantity Instability: 1961—71 to 1972-82 157 Regression Coefficients Explaining Increase in Price Instability: 1961-71 to 1972-82 159 vii Table Table Table Table Table F.1 Page Commodity Codes, Type and Data Coverage 179 Instability Measures by Commodity 187 Ratio of Variance in the Demand Intercept to Variance in the Supply Intercept (DSR) for Corn under a range of Elasticity Assumptions: 1971- 72 through 1982-83 192 Decomposition of Corn Price and Quantity into Supply, Demand and Interaction Components 195 Derived Data for the Analysis of the Study 197 viii LIST OF FIGURES Page Figure 4.1: Graph of SDDQ and CVTQ Instability Measures: Monte Carlo Simulation (lines) and Actual Data (points) 60 Chapter 1 INTRODUCTION 1.1 THE PROBLEM OF INSTABILITY Many commentators identify instability as being the major long term problem facing United States agriculture. The subject occupies about a quarter of Heady’s textbook (1952). Tweeten (1979) gives more space to this topic than any other in his book on farm policy. In addition there is evidence that instability in agriculture is increasing especially for export crops (see for example Tweeten, 1983; Hazell, 1984; Firch, 1977; Mangum, 1984; Myers and Runge, 1985). Why is agriculture so prone to instability that so many see it as a problem. And is it really a problem? One reason for the concern with the issue is that agriculture is unusual in that input—output relations are relatively uncertain. In most industries the employment of known input quantities is almost certain to lead to the predicted output. The relationship is given by the specifications of the chosen technology. This is not true for agriculture where weather and disease can affect yield. This undoubtably introduces a degree of uncertainty in agriculture, which is absent in most other industries. To be sure other industries face uncertainty in needing to forecast the behavior of competitors and demand patterns of consumers. But these are also uncertainties bourne by agricultural producers. If weather and diseases were the only factors making for the high instability of agriculture then little could be done to improve the situation apart from the development of disease resistant strains and perhaps the development of long term weather forecasting techniques or extensive irrigation. However, there is another difference that makes for greater instability in agriculture than other industries. This is the coordination system within the marketing chain. In many non-agricultural industries marketing arrangements, such as vertical integration and contracts, ensure that only enough product is produced to meet the expected demand. As a result the supply and demand of intermediate products is at least effectively coordinated. This is not so in the food system. This is not necessarily a coordination based on sequenced perfectly competitive markets, but rather one based on mechanisms established to deal with the potential uncertainty of the system. In agriculture, market coordination is perceived to be a problem. The demand for forms of orderly marketing is evidence of this perception. However, most economists see this demand as one based on ignorance or as a thinly veiled request for subsidy assistance. This dissertation focuses on this issue. How much of the instability evident in agriculture might be explained by exogenous, uncontrollable factors, such as weather and demand shifts, which are difficult to alleviate; and how much is due to potentially controllable failures in coordination. It is difficult to analyze these things for any individual commodity market. However much can be elucidated from a comparison of a number of commodity markets with different coordination mechanisms. 1.2 ORDERLY MARKETING While orderly marketing seems to be desired by participants in agricultural markets, but it is not a well defined concept. It is associated with market stability and some measure of certainty. It also represents some distinct disatisfaction with the coordinating role of the market, where there is evidence that certain participants benefit from an asymmetric distribution of information or unequal power. Orderly marketing therefore has something to do with ensuring that the correct amount of product is produced and distributed in a timely fashion to consumers, when and how they want it. But what is the correct amount. The neoclassical model implies that prices will convey the appropriate market signals to ensure this coordination. However to do so requires conditions of perfect competition with complete certainty, the absence of transactions costs, well defined property rights and intertemporal markets. Some of these conditions are so completely lacking that attempting to promote those conditions can be prohibitively expensive. Hence there is a demand for alternative market 4 coordinating structures (to so called free markets) as articulated by the desire for orderly markets. 1.3 INSTABILITY AND MARKETING COORDINATION There is a relationship between price and quantity instability and orderly marketing. In fact intertemporal failures in marketing coordination are reflected in market instability. In this sense a measure of instability may be taken as a measure of marketing coordination effectiveness. Of course, such a measure is only a proxy, for there are other factors that cause instability. Indeed prices and quantities must vary somewhat to ensure market coordination. It is of course excessive instability that is both symptomatic of a coordination problem and also contributory to it. This study is directed at the measurement and analysis of market instability (prices and quantities) of agricultural commodity markets with the purpose of identifying poorly coordinated markets. 1.4 STUDY RATIONALE AND METHOD Thus the approach of this study is principally diagnostic. The intent is to identify possible failures in market coordination and to provide some quantitative estimate of the contributions of various sources of instability, with particular attention given to institutional factors which may be potentially alleviable. The breadth of scope of this study necessitates that it be 5 mainly exploratory and heuristic in its approach. However, the focus upon one dimension of market performance, namely instability, permits the application of quantitative techniques for ‘measuring’ the relative institutional performance of individual commodity subsectors as evidenced by their instability. It is intended that this study provide both a context and a source of hypotheses for future research into orderly market questions for United States agriculture. 1.5 SUMMARY AND PLAN OF STUDY The plan of this study is as follows. In the next chapter consideration is given to what the term instability actually means. It will be shown that the mathematical concept of instability is unhelpful and in fact has confused discussion of the issue. Although instability can be thought of as being partly predictable and partly unpredictable, it will be shown that both types are important for market coordination but for different reasons. Instability has been given extensive coverage by the agricultural economics profession and Chapter 3 is devoted to consideration of this literature. Particular attention is given to reasons why market prices may not both efficiently allocate existing production and convey appropriate marketing signals for future production decisions. Little of the historic interest of the profession in instability has been directed at market questions, apart from remedial problems like buffer stock 6 schemes. In Chapter 4, various instability measures are discussed and appropriate measures are developed for the analysis of this study. Measures of instability for more than 100 commodities are examined in the fifth chapter to determine which commodity markets exhibit the most and least market instability (i.e., of prices and quantities). This cross-sectional analysis of commodities allows comparison between markets so that not only can the more unstable markets be identified but also the relative instability of markets can be determined and a context is provided for their relative magnitudes. Then in Chapter 6 consideration is given to the sources of this instability and the relative importance of each. First, the effect of yield instability is examined and shown, as one might expect, that it can explain a major part of observed instability. Area instability is also examined and shown to represent another major source of instability. It is seen to be associated with yield instability indicating that producers of products with unstable yields have difficulty coordinating production decisions. Then the possible influence of various supply and demand shifters, such as population, income and input prices are considered and shown to be of relatively minor importance in explaining year to year variation. The next part of Chapter 6 is devoted to analysis of that part of instability that can be explained by various institutional and physical characteristics of the commodity or market, and 7 a final section considers the empirical evidence for price instability creating further instability. In Chapter 7 the question of increasing instability is investigated. It is shown that whilst price instability has increased, the increase has been more concentrated in particular commodity groups than others so that it is not as general as some previous analyses have implied. The sources of this increase are then investigated. The final chapter provides conclusions to the analysis and examines directions for further research. Chapter 2 WHAT IS INSTABILITY? 2.1 INTRODUCTION Instability is a concept that everyone professes to know but few can define. Moreover, it quickly becomes clear in any discussion that people have very different interpretations of what instability really is. The word itself has a negative aura and instability seems implicitly undesireable. However, there is often a difference of opinion between professional economists and market participants whether the concept is at all relevant to economic markets. Partly the disagreement reflects different understandings of the meaning of the word; and partly it reflects differences in confidence about the ability of markets to operate efficiently or fairly. In this chapter I will describe some different interpretations of the concept and suggest a definition that is appropriate for this study. 2.2 MATHEMATICAL INSTABILITY The usual mathematician’s or physicist’s use of the term instability describes a property of dynamic systems in relation to a steady state or equilibrium. A system may be stable if it converges on an equilibrium or unstable if it diverges from an equilibrium. An intermediate case is where 8 9 a constant cycle is maintained about an equilibrium: neither converging or diverging. Mathematically whether a system is stable or not can be determined by whether the eigenvalues of the equations describing the system fall within the unit circle in cartesian real-imaginary coordinates (Chiang, 1974). An alternative, and equivalent requirement, is that the poles of the equations describing the system must fall in the negative quadrants of the Laplace s-space for stability (Manetch and Park, 1984). This definition of instability is not very useful for analysis of existing markets. If the structure underlying a market had unstable characteristics in the sense described above, then no market would be possible. If there was an initial equilibrium, the first shock would ensure displacement from equilibrium and prices and quantities would then go to either zero or infinity. Such markets do not generally exist.1 Thus in this mathematical sense of instability, existing agricultural markets are stable. Clearly this is not a useful description of instability for the purposes of this study. However, it provides the reason why many will assert that agricultural markets are ‘stable’. And in this sense they are. However it is clear that agricultural markets exhibit instability of a different 1 Some markets may experience these characteristics for a period. For example, the dynamic processes underlying "tulip manias" can be described as unstable. In these cases price expectations are based on rates of change in prices so that a spectacular boom is followed by a bust. The source of this descriptive name is such an occurrence in the Dutch market for tulip bulbs in 1634-37. 1% quality. I will call it economic instability in contrast to the mathematical instability described above. 2.3 SYSTEM INSTABILITY One approach to describing economic instability is to consider it as the dynamic adjustment path towards equilibrium within a systems perspective (Butler, 1979). This approach recognizes three sources of instability: exogenous input shocks (both controllable and uncontrollable), random shocks in the feedback mechanism, and exogenous behavioral influences. Instability is then measurable in terms of the dynamics of the estimated system using characteristics of the ‘transient response’ of the dynamic system. These characteristics are typically ‘rising time’ - the time taken in response to a shock to reach a specified proportion (e.g., 90%) of the new equilibrium; ‘overshoot’, the percentage by which the maximum value of the actual series overshoots the equilibrium; and ‘settling time’, the time taken for the series to settle within a specified width band of the equilibrium value (say 5%). (See Manetch and Park, 1984). This approach is not very useful for choosing a measure and is difficult to operationalize for analysis of actual (as opposed to conceptual) markets. However the approach is very helpful in understanding what the idea of instability might encompass. In particular it distinguishes exogenous and endogenous sources of instability. It also highlights 11 the role of the feedback mechanism for understanding endogenous instability. The marketing institutions are part of this feedback mechanism. 2.4 INSTABILITY AND MARKET COORDINATION The discussion above provides some insights into the relevance of instability for market coordination. In line with the systems approach, markets exist in an economic environment which generates exogenous shocks to a market system. These shocks may be weather conditions, changes in income growth or government policy changes. Within the supply and demand framework it can be expected that the market clearing equilibrium of prices and quantities will exhibit a degree of variability in response to these exogenous shocks. This variability might be considered as the ‘normal’ variability or instability that facilitates the allocation of already produced goods within a market. In addition to this market adjustment source of instability there are other sources identifiable within the systems perspective. These include the dynamic process as economic variables converge on a new equilibrium. The overshoot and settling time characteristics of the transient response are components of the instability generated by the dynamic process which exceed the demands of ‘normal’ instability described above. On the other hand the rise time characteristic of transient response may be a component that reduces observed instability. Such lags however, probably .... I’IJ represent either excessive production or failure to take advantage of potential profitable production opportunities. For example, if an export embargo is imposed on a crop which makes it less profitable to grow, and farmers continue to produce it in unprofitable quantities, then resources are wasted relative to their alternative uses. However, observed market instability is less than if they had reduced production immediately. When they do decrease production they may do so to an excessive degree causing overshoot and with a longer subsequent settling time. These later characteristics will be observed as instability. These responses are properties of the feedback mechanisms of the market. In other words, they are functions of the institutional design and management of the market. For example market intelligence services, the structure of the market and the existence or lack of futures markets will all affect the transient response to a shock. In this way instability can be seen to be dependent on the coordinating mechanisms of the market, and on the endogenously induced changes in participant behavior. Thus instability of markets has both desireable and undesireable components. First, some part of instability functions to coordinate markets. It represents the observance of the market signalling function of prices and the subsequent adjustment of quantities. Another part H [.1 represents instability in excess of (or possibly less than, or different than) that required to achieve this aim. The discussion above views instability of a market as a measureable output or as a performance indicator of a system. However instability is also an input into the feedback mechanism. Unstable markets are more difficult to predict than stable markets and hence production, marketing and consumption decisions are more difficult to make in such systems. It is reasonable to expect that instabiltiy in these circumstances may breed further instability as participants in uncertainty respond by making poor decisions. In market failure terminology instability can be said to generate an informational externality. In other words participants in unstable markets can not help but make some wrong decisions about production, marketing and purchases, which affect others as well as themselves. Thus instability may be considered as both a cause and a consequence of poor market coordination. However it is only an indicator. For example, it might be desireable to have increased price instability if holding prices constant masks market signals so that inappropriate levels of production are induced or markets do not clear. In fact a number of situations can be conceived where increased instability might have desireable alternative consequences. However, it seems reasonable to consider this dimension as one indicator of market coordination effectiveness despite these difficulties. Most, (except, as shall be seen, some 14 economists!) consider instability or ‘excessive’ instability to be a bad thing. 2.5 PREDICTABLE AND UNPREDICTABLE INSTABILITY Instability can be detrimental to market coordination in that it is detrimental to prediction. However some instability is in fact predictable, at least by some participants and to some degree. For example the well attested cycles in beef and hogs may permit a greater degree of predictability than shorter run fluctuations in markets such as for soybeans. In situations where there is predictable instability then participants may adjust to their known economic environment. Hence, they will not produce product for less than the cost of production nor fail to exploit profitable production opportunities. Thus it might be argued that no resource allocation problems are created in this situation. There are two problems with this argument. First, this reasoning is static and ignores the existence of fixed capital inputs in production, processing and marketing. When these inputs are product specific then there will be unutilized capacity at certain times. Second, intertemporal storage costs will likely be higher under more unstable (although predictable) markets than under stable markets. Thus entirely predictable instability will be of concern for market coordination as well as unpredictable instability. However, when real world commodity markets are examined it is difficult to find much predictable instability. In the examples given above, it is clear that a large proportion of the instability evident in these markets is very difficult to predict. Moreover the cyclical variability of these ‘predictable’ markets is indicative of the market instability which helped to generate the cycles.2 Thus although predictable variability may be of less concern than unpredictable variability, both forms of variability may be symptomatic of existing market instability. This discussion has immediate relevance to the choice of an empirical measure. For the reasons decribed above it seems reasonable to prefer a measure of instability rather than attempt to measure predictability. However, among instability measures there seems some justification to choose a measure that emphasises shorter term variability rather than cyclical phenomena although that too is important. I will return to this issue again when consideration is given to the choice of a measure for analysis in Chapter 4. 2 Although the evidence from the theory of partial equilibrium analysis is supportive of the contention that stabilization of a single market is desireable, an example can be provided of a two commodity market where cycles are optimal. Assume that corn can be either consumed, stored (at a cost), or fed to hogs; corn harvests are stochastic: and consumers maximize an intertemporal discounted and concave utility function: then prices oscillate along the optimal path. In this model hogs become an efficient means of utilizing excess grain during good seasons compared with costly corn storage. The hog-corn price ratio oscillates in the optimal strategy (Burmeister, 1978). 16 2.6 INSTABILITY OVER TIME It is now obvious that instability is many faceted. One of these facets is the periodicity of the instability. The beef cycle is an example of a relatively long period (10 years) cycle. Other industries also exhibit cyclical instability, for example hogs, eggs (Hartman, 1974), lemons and watermelons. Some of this cyclical phenomena is a result of the length of the production period. This may extend from some weeks in the case of eggs to about a thousand years for the life of an olive tree. The longer the production period the longer the period of possible cyclical instability for an industry. In addition nearly all agricultural industries exhibit annual variation in response to seasonal factors. Thus although olive trees have a long productive life, they produce olives each year, and the annual crop is partly a function of weather conditions. The annual crops (e.g., corn) fit this case most clearly. Eggs are of course an exception in that they are little influenced by weather. Some quickly growing vegetable crops, such as lettuce, also exhibit instability where the week to week movements are probably of more relevance for market coordination than the year to year changes. Even though corn is an annual crop, corn prices fluctuate by the minute. And corn production is not only an annual phenomenon in that production is dependent on fixed costs or on previous investment that has a longer time horizon than one year. Thus for each commodity it would be possible to find a spectrum of instability 17 measures classified by the periodicity of the measure. This discussion is suggestive that there may be virtue in considering instability measures based on autoregressive moving average models (Box-Jenkins) or upon spectral fourier techniques (eg Hannan). In the interests of simplicity and cross commodity compariability I will choose one time frequency and choose one which seems to offer the best means of comparing commodities. This would seem to be the annual frequency. 2.7 CONCLUSION What does all this mean then for the definition of instabilty and for its measurement? In this chapter I have argued that mathematical instability is an inappropriate concept for the analysis of markets. A more useful approach is provided by systems concepts of transient response or, alternatively, measures of variability. Such a measure can provide a dimension of the effectiveness (or rather the lack of effectiveness) of market coordination in an agricultural market. It is not a perfect measure but it has some virtues for comparing marketing arrangements between industries. In addition the degree of market instability will be of interest to others concerned with broader agricultural policy issues. It is problematic, however, exactly what instability is. I have argued that it is related to, but not limited to, unpredictability. Instability may exacerbate unpredictability and vice versa. I have also argued that 18 instability may be measured according to different time frames: some of more relevance to certain industries than others. However, it would appear that an annual measure is the most useful one for making comparisons between commodities. Thus, for the purposes of this study, I will define market instability to be some arithmetical summary measure of the annual variability of price and quantity for each agricultural commodity. The exact choice of measure will be discussed in Chapter 4. Chapter 3 APPROACHES TO INSTABILITY 3.1 INTRODUCTION In this chapter I will give a brief review of the treatment of instability in the economic literature. This will provide a rationale and context for the analysis described in this paper. The major contributions may be classified in five groups as follows: 1. Traditional Marshallian approach 2. Asset fixity 3. Modern risk aversion approach 4. Market failure 5. Associated with poor market coordination 3.2 TRADITIONAL MARSHALLIAN APPROACH The traditional approach to instability uses the theoretical framework of Marshallian analysis. This approach allows conclusions to be reached about the consequences of instability for welfare of producers and consumers and for society, as measured by producer and consumer surplus. It also permits some conclusions to be reached about the advantages or disadvantages of price stabilization. Perhaps the first to use this approach was Waugh (1944). He assumed linear demand and supply schedules, a perfectly competitive market and no uncertainty. He showed I9 0 fl} that consumers benefit from price instability relative to prices remaining stable at their mean. Apparently unaware of Waugh’s contribution, Oi (1961) proved the analogous result for producers, i.e., that producers benefit from unstable prices relative to the mean price. The implication of these results was counter-intuitive: that price instability is desireable. However, each of these approaches considered the welfare of one group, ignoring the effects on the other participants. Subsequently Massell (1969, 1970) synthesized Waugh’s and 01’s results to conclude that who gains and who loses depends on the source of the instability: producers gain (lose) and consumers lose (gain) when the source of the instability is on the demand (supply) side. However the gains from instability are insufficient to compensate the losses so that total welfare is always reduced under instability. Samuelson (1972) made a similar point in noting that these analyses are partial and that consideration of general eqilibrium factors demonstrates that the price instability described was infeasible. Unstable prices would not have as their mean the price that would be maintained under stability. So that total welfare is reduced under instability relative to the stable case. It should be noted that these results are obtained on the basis of a competitive market with linear demand and .supply schedules, where the instability is reflected in additive horizontal shifts of the schedules and where prices are certain although variable. Various analyses have been done to investigate the sensitivity of the results to these assumptions. Tisdell (1963, 1978) took issue with the assumption that prices were known when production decisions were made. He demonstrated that Oi’s result was not maintained when actual prices (as opposed to the price distribution) were not known at the time of the production decision. Turnovsky (1974) extends this approach by investigating the implications of modelling price expectations with alternative lag structures, namely Nerlovian adaptive and Muthian rational formulations. He shows that the Oi result (namely that producers gain from instability originating from demand fluctuations) depends crucially upon how the expectations are generated. Rational expectations do not change the 01 result, although adaptive expectations may. Not surprisingly his models demonstrate that price instability creates greater losses to net welfare when supply is based on expected prices than on actual prices (i.e., perfect information). Thus instability which is not predictable (uncertain) has greater welfare costs than predictable instability. This has relevance for the measuring of instability in the next chapter. Turnovsky (1976, 1978) also considers the implications of alternative specifications of the stochastic elements of the model. He uses a model with multiplicative shifts and finds that the price elasticity of demand becomes critical in determining the distribution of benefits from reduced instability rather than the source of the instability, as was the case for .:_ l— I'I 'I linear schedules. He shows some modification to Massell’s conclusions. In particular the adaptive formulation gives indeterminant results and the extent to which producers lose from reduced price instability depends upon the autocorrelation of the stochastic shifts that provide the instability. However, in all these partial equilibrium analyses, reduced instability provides net welfare gains. The interested reader will find many of these results summarized in Adams and Klein (1978), and Newbery and Stiglitz (1981). Perhaps the only attempt to use the Marshallian welfare approach in the context of alternative market structures to the perfect competitive model is provided by Bieri and Schmitz (1974). They examine the case of an intermediary who is either a profit maximizing monopolist-monoposonist (‘pure middleman’) or a producer controlled marketing board. Their example might apply to the international grain trade. They find that the pure middleman can gain by actively manufacturing price instability or by not stabilizing price fluctuations when they are a result of natural causes. This is not true for the marketing board where stabilized prices are desireable for producers. The general conclusions of this literature is that there seems to be agreement that instability is socially undesireable (as measured by producer and consumer surplus methods) and that there is a redistributive effect among rI_I Lu economic agents as a result of the instability. However the form of the instability, the way prices are predicted and the nature of the demand and supply schedules can all modify some of the general conclusions outlined by Massell. Criticisms of these Marshallian approaches can be made on the basis of their partial equilibrium assumptions, the measurement of welfare in terms of uncompensated surpluses (Currie et al, 1971; Willig, 1976), how they deal with uncertainty, the assumption that ex post and ex ante supply are the same, the lack of consideration of dynamic factors in production, their failure to consider costs of stabilization and their reliance (in all but one case) on the assumption of perfect competition. 3.3 ASSET FIXITY THEORY One of the assumptions in all of the above analyses is that of a static production process. But dynamic factors are particularly important in the study of instability. The commitment of inputs in production, including marketing, may have a stabilizing effect as production decisions may not be revised rapidly. However, the difficulty of revising production decisions where there are fixed factors can lead to long term disequilibrium. Johnson and Quance (1972) argue that this can result in an overproduction trap as producers make optimum decisions for variable inputs based on previous decisions about fixed inputs. Supply functions therefore 34 take different forms depending on whether prices are increasing or decreasing. This theory demonstrates the importance of dynamic factors for instability. Fixed factors can explain some stabilizing role in the short term coincident with long term disequilibrium and hence instability. 3.4 MODERN RISK APPROACHES In the last chapter, the relationship between instability and predictability was considered. In summary it was shown that observed instability is partly predictable and partly unpredictable, and that these elements had different consequences for market coordination. In addition instability encourages unpredictability so that there is a complex interrelationship between these elements. In this section I wish to review some of the ‘risk’ literature as it pertains to market instability. Clearly ‘risk’ is directly relevant to instability. This section is in three parts. First, the meaning of risk as it is used in the literature is discussed. The second and longest part provides a brief review of how the concept of risk has been developed, how it has been used in economic models of behavior, and some of the principal conclusions of this literature as it pertains to agricultural markets. Third, I mention some of the recent economic modelling studies, which indicate that instability matters also for ‘risk-neutral’ behaviors although some common perceptions (and statements) often suggest otherwise. m Ul 3.4.1 RISK AND UNCERTAINTY The classic examination of risk in economics is provided by Knight (1921). He notes, inter alia, that a distinction can be made between risk and uncertainty. The distinction between these concepts is that a probability distribution can be formulated for risk whilst it is not possible to do so for uncertainty. Since Savage (1954) this distinction has fallen into disuse on the basis that every individual will be able to form some subjective probability distribution over possible outcomes even though the objective distribution may not be known. Moreover, when one is considering an individual’s own utility the subjective, rather than the actual, probability distribution is the relevant one. Despite this, analysts often make some distinctions based on the degree of uncertainty. For example, Newbery and Stiglitz (1981) distinguish systematic and unsystematic sources of instability. Moreover empirical evidence does suggest that lack of knowledge of the underlying probability distribution does affect behavior 26 (e.g., Ellsberg, 1981 who distinguishes ‘ambiguity’).1 While there is much discussion of risk, little progress has been made on incorporating uncertainty (or unsystematic risk, or ambiguity) in economic models. It can be argued that market participants will have some idea of the distribution of prices and outputs to be expected in a market. But it can equally well be argued that participants do not form subjective probability distributions. In a later chapter I will present evidence that instability of agricultural markets has increased. In such a situation it is probable that there is a large measure of the instability in these markets that represents uncertainty and can not be described only as risk. Thus the concept of risk, as it is used in the literature, can be considered to be relevant to only a component of the unpredictable part of instability, as I 1 Another example is provided by the so called coin-tack game. Here participants in a controlled experiment are asked how much they would pay for the opportunity to gamble where they can nominate the way a fair coin will fall. They are also asked the same question for a tack which may fall point up or on its side. Typically participants will offer less for the tack game even though they have the choice of calls. As they have the choice of calls the probability of success must be at least as good as a 50-50 chance of the fair coin game. In fact the objective probability favors the point on its side and most will choose this. However the fact that they offer less for the tack game suggests that the underlying subjective probability distribution is insufficient information to predict behavior: the uncertainty of the situation also affects the price offered. Real world situations are likely to be more complex than this simple experiment and ‘uncertainty’ might be expected to dominate ‘risk’. E7 have defined the term in the last chapter. Nonetheless it is useful to review the literature and this is done below. 3.4.2 RISK IN ECONOMIC MODELS Early studies that attempted to incorporate risk in agricultural models used various versions of mean-variance analysis. These postulated decision makers making a tradeoff between expected outcomes and the expected variance (or standard deviation) of the same outcomes. However a major thrust of recent conceptual research into the influence of risk has concentrated on the effect of risk averse behavior within the framework of the expected utility hypothesis (see for example the extensive bibliography of this literature in Machina, 1983). This has followed the development of a generally accepted measure of risk aversion by Arrow (1971) and Pratt (1964); and an analogous definition of increasing risk as defined by Rothschild and Stiglitz (1970)2. When these concepts are applied in a comparitive static framework they suggest that behavior will often differ in the presence of risk than would be the case otherwise. For example the familar result that fixed costs are irrelevant for a profit maximizer is not generalized for a utility maximizer in the presence of risky output prices (see Sandmo, 1971). In the context of agricultural production decisions, risk averse 2 The Rothschild-Stiglitz definition of risk can be described as a ‘mean preserving spread’. Unfortunately it does not provide a complete ordering and hence is not useful for the comparison of different markets as is done in this study. B |'|_‘I individuals (who are assumed to maximize expected utility) apply a discount to risky outcomes. Thus it might be expected that producers will apply less inputs to a product which is more risky than otherwise and that supply would be lower as a result, so that risky enterprises would have higher prices and lower output than would be the case under lower levels of risk. In fact, this expected result is not always forthcoming from models that utilize the expected utility hypothesis, once they are extended to include more realistic formulations. First, the argument of the utility function is unlikely to be prices or even revenue: producers are concerned with income or more likely with consumption, rather than these intermediate parameters. Second, the presence of alternative outputs allows farmers to form a portfolio of production possibilities, which may lead them to choose higher production of the more risky product rather than the reverse.3 There may, of course, also be markets for shifting risk (eg futures or credit markets or private storage) which make risk of less consequence for individual farmers. In fact when complications are added to economic models of risky markets, there are few unambiguous answers. 3 An example can show this. If a farmer can produce two crops: one with no risk and another more profitable but riskier crop. Then he may decide to spread his risk by producing both crops. Now if a stabilizaton scheme reduced the price risk of the second crop, the farmer may increase production of this crop as he has less need to spread his risk across the less profitable but sure first crop. If the second crop has elastic demand (say an internationally traded good) and the first is not, then aggregate supply response by producers may lead to them being worse off than before. 9 IL" This is clear from the extensive study on the economics of risk as applied to commodity price stabilizaton schemes undertaken by Newbery and Stiglitz (1981). This work provides a comprehensive review of recent results in this area and can be recommended to the interested reader. Notwithstanding these difficulties, the risk literature demonstrates the importance of one type of instability for economic behavior. It highlights the importance of risk management and the possible desireability of institutions to manage risk, either through remedial schemes such as buffer stocks or through the development of appropriate institutions. One result of these studies is to show that risk to one participant is not the same as to another (see, for example, Sharpe, 1964). Thus institutions to shift risk may reduce (or increase) it. The literature also questions many of the received results of neoclassical economics. Before leaving this literature it is important to note that many of the results are dependent on the validity of the expected utility hypothesis. This was developed by von Neumann and Morgenstern (1944). It can be proved from a series of axioms that the expected utility of an uncertain event is the sum of the utilities of each possible outcome weighted by its probability (see, for example, Savage, 1954; or Luce and Raiffa, 1957). The validity of the theory depends upon the acceptance of the axioms, the acceptance of the concept of utility maximization as a description of O DJ human behavior and it is generally also required that there is constant marginal utility for money. The underlying axioms seem ostensibly reasonable but behavior is sometimes contary to them. The Allais (1953) and Ellsberg (1981) paradoxes provide examples of instances where most individuals do not observe the axioms. The utility assumption while debateable is a common one in economics. On the other hand the last assumption (of the constant marginal utility of money) is difficult to support and without it the theory cannot differentiate between risk aversion and declining marginal utility of money (see Fleisher, 1985). 3.4.3 INSTABILITY MATTERS, EVEN WITHOUT RISK AVERSION It is often asserted that risk averse behavior is necessary for instability to make a difference to outcomes (see, for example, Biswanger, 1979 p392). However it is not necessary to postulate risk aversion to obtain results showing that instability makes a difference. Just (1975) demonstrates a simple model that shows risky costs will alter the optimum production level for an expected profit maximizer. In fact non-linearity of the objective function in static models is a sufficient condition to obtain this result. Antle (1983) shows that risk also matters for dynamic formulations (multi-stage or multi-period) even with risk-neutral agents. The conclusion to be drawn from this review is that risk or instability does affect the behavior of economic agents with or without the presence of risk-averse behavior in all but the simplest static economic models. Thus there is possible potential to improve economic performance if risk can be managed. 3.5 MARKET FAILURE The neoclassical model provides abundant rationale for ignoring instability all together. The argument runs as follows. Theory postulates that, under the assumptions of the perfectly competitive model, the operation of markets will ensure the attainment of pareto optimal efficiency. Hence the observed variability in prices and quantities is merely evidence that markets are doing their job. There is therefore no reason to be concerned with instability. The market failure approach considers what deficiencies there may be in a particular market which violates these assumptions and that may justify appropriate intervention to overcome the failure. It seems worthwhile to consider some of the candidates for market failure in agricultural markets in some depth. 3.5.1 PERFECT INFORMATION It is clearly counterfactual to assume perfect information. However it is not so clear whether the real world might not approach a situation where ‘it acts as if’ the perfect information requirement was satisfied. In this section I will examine some of the empirical and theoretical 7.1, x -" I' 3 I evidence on the subject to ellucidate the relevance of this restriction for the question of instability. The theoretical work of Rothschild (1973, 1974) shows that relaxation of the assumption of perfect information can have drastic effects on the allocative role of the market. He introduces a consumer search cost into the simple neoclassical model and finds that monopoly pricing can be expected despite unlimited numbers of sellers. Another approach to the role of information costs is provided by Heiner (1983). He finds that the most economical policy in the presence of uncertainty is to economize on the acquistion of new information. In fact his model finds ‘uncertainty to be the origin of predictable behavior’. Moreover he finds that there is little reason to expect firms to act as if the perfect information assumption was satisfied. Successful firms will be satisficers with respect to information requirements. 3.5.2 INFORMATION AND EFFICIENT MARKETS The efficient markets literature provides some basis for the empirical examination of whether there is sufficient information available for markets to operate efficiently. Samuelson (1965) was the first to prove ‘that properly anticipated prices fluctuate randomly’. Fama (1970) formulated this result in tests of the relevant price series as to whether they are thereby consistent with the efficient [-1 [d market hypothesis. A difficulty with this approach is that price series will only be expected to move in a random walk if all available information is utilized. However the acquistion of information, including the knowledge of the model, is not costless. The inclusion of only ‘economically’ rational expectations need not be unbiased and so correct forecasts on average need not characterize efficient markets (Smith, 1978). The empirical evidence on commodity markets is at best mixed and inconclusive as to their ‘efficiency’. While tests of financial markets were rarely able to reject the hypothesis of ‘efficiency’ the results have been more mixed for commodity markets. Moreover, there is abundant evidence that relatively simple rules can be profitably employed in commodity markets which is supportive of information impactedness rather than market efficiency (see Smith, 1978). In addition international data on stocks and production are often scanty and of dubious quality, and while there is some modelling of commodity markets, they have not proved to be especially accurate nor are they particularly extensive especially outside the US. Thus the empirical evidence seems to suggest that information may be insufficient for commodity markets to operate ‘efficiently'. A particularly apt illustration of the importance of information is the distinction between ex ante and ex post 4 [-1 prices. Where producers have perfect knowledge the market mechanism will perform its allocative function on already produced goods. The resulting price will be an efficient and unbiased market signal for future production decisions. However in the presence of uncertainty, (as Smith shows) the resulting price will be (in general) biased. Hence the variability of the price obtained in the market will reflect not only stochastic elements, such as weather, but also the mistakes of market participants. Thus prices, in such conditions, can not simultaneously perform their allocative function (for already produced goods) and their signalling function (for future production) in an efficient manner. Clearly this is a potential source of instability in agricultural markets. 3.5.3 INCOMPLETE MARKETS As is well known, one of the contributions of Arrow and Debreu (1959) to the understanding of the theoretical requirements for pareto-efficiency in general equilibrium is that there must exist a complete set of markets. They show that these are a necessary condition for efficiency. Where there is a temporal dimension and lack of perfect forecasting then complete sets of futures and risk markets are also required. However, only some commodities have futures markets and these typically extend only a short distance into the future so that they only allow producers market information for their short run production decisions. 35 Market signals, in the form of future prices, are not available for longer run production decisions, such as capital equipment or land purchase decisions. The situation for risk markets is similar. The problems of moral hazard and adverse selection have discouraged the formation of adequate insurance markets. As a result prices are called upon to not only provide market signals but also to carry risk. They can not do both functions at once and hence they are inefficient (see Newbery and Stiglitz, 1982). 3.5.4 NON COMPETITIVE MARKETS It is sometimes argued that agriculture provides the closest example of the conditions for perfect competition. However this assertion ignores other parts of the marketing chain where there are fewer participants and many institutions which affect market performance (Parker and Connor, 1979). 3.5.5 MARKET IMPERFECTIONS ELSEWHERE Where there is at least one imperfection in the economy, the theory of the second best asserts that there is no guarantee that the lack of market imperfections elsewhere will contribute to a pareto efficient outcome (Lipsey and Lancaster, 1956). As all economies have lots of imperfections (including those described above and specific government interventions) there is no presumption that a less regulated market will be more efficient. This is a rather general and iconoclastic demolishment of the argument for laissez faire. But is there specific evidence that agricultural commodity markets are adversely affected by the absence of the conditions of perfect competition? Wage and price rigidities, and imperfections in the capital market are good candidates (Newbery and Stiglitz, 1981). The commodity boom and bust of 1972-75 were very likely in response to the instability of international liquidity (Bosworth and Lawrence, 1982; Bond et al, 1984). Other sectors of the economy, which have wage and price rigidities, were more insulated from these developments. Hence the agricultural sector experienced the brunt of this instability. Undoubtably these ‘imperfections’ not only affect instability but also the level of prices and quantities. 3.5.6 OTHER EXTERNALITIES Other external effects have been noted for agricultural commodity markets (Smith, 1978). These include the macroeconomic effects of agricultural instability, especially inflation and the associated increase in inflationary expectations; and the effect of instability on the exchange rate (especially in developing countries). The difficulty of forecasting future prices in the presence of instability can also be considered an external effect and therefore a source of market failure. Lu] \J 3.5.7 CONCLUDING COMMENTS The conclusions to be gained from this review of the relevance of market failure theory are straight forward. There is clearly abundant evidence that agricultural markets are likely to exhibit market failure. Thus laissez faire prescriptions are unlikely to be optimal. In addition this review highlights some of the possible areas which may be candidates for institutional reform. These include the possible development of futures and risk markets. Moreover instability is isolated as both symptomatic of market failure (lack of information, incomplete markets, imperfections elsewhere in the economy) and as a characteristic of agricultural markets with undesireable external effects. Hence policies that are directed towards alleviating instability may have desirable external effects. However the market failure approach, whilst examining the failures of the assumptions for the neoclassical paradigm to apply, still uses that paradigm. Pareto- efficiency becomes the performance measure for analysis, and little attention is given to other performance criteria. An alternative approach will be discussed in the next section. 8 DJ 3.6 MARKET COORDINATION APPROACH An alternative approach to those described above is described in Shaffer (1980). Shaffer adapts the familiar structure-conduct-performance (SCP) paradigm used to study the industrial organization of markets (Bain, 1959) and generalizes it to apply to policy situations in general and market coordination problems in particular. His approach is to develop a paradigm in terms of environment, behavior and performance. Like the SCP paradigm this formulation allows the analyst to consider multiple performance measures and to choose ones that are appropriate to the analyticl purpose. Moreover it allows examination of behavior in its particular environment rather than what it ‘should’ be. This approach permits direct policy analysis of the environment in order to direct behavior so that some desired performance is achieved. The environment of the paradigm is a series of overlapping opportunity sets which are physical, politico- economic and determined by an individuals position in the economy. Each individual’s opportunity set is constrained by the organization(s) to which she belongs, market factors, property rights, technology, internal operations of the organization(s) and pervailing uncertainty (especially information impactedness (Williamson, 1979, 1981) etc. The response of individuals and organizations to their environment can be described as their behavior. Important 39 characteristics of this behavior are bounded rationality in the presence of opportunism (see Williamson, 1979, 1981), satisficing rather than maximizing behavior with consequent development of standard operating procedures, multible goals (Cyert and March, 1963), slackness (Hirchman, 1970; Liebenstein, 1979), selective perception of the environment and the importance of collective action (Olson, 1965) and learning. Performance is then the outcome of the behavior of the sum of all the relevant participants. What counts as performance will be dependent on the political articulation of preferences. Performance outputs will also be part of the new environment. The analysis problem is to understand the linkages, and the policy problem is to redesign the environment with its structures of incentives and distribution of power, to achieve desired performance. In the context of this study, instability may be treated as an undesireable performance characteristic, or as an instrumental variable influencing a number of performance characteristics of the marketing system. Clearly the structure of certain commodity markets is such that the behavior of individuals and organizations in the environment leaves something to be desired. Moreover this undesireable performance is likely to reinforce and produce further instability (Skinner, 1974). A part of the problem is that some marketing systems are poorly coordinated. 4% Some of the heuristic implications of this approach to the study of market instability are: 1. it directs attention to instability as an undesireable performance characteristic in its own right and as an instumental variable influencing performance; 2. it suggests that the analyst consider alternative market coordination mechanisms than spot markets (eg vertical integration, contracting etc) rather than using the perfectly competitive market as a norm with its counter- factual assumptions: 3. it suggests that explicit attention be given to the role of uncertainty rather than assuming it away. 3.7 CONCLUSIONS This brief review of the extensive literature on instability shows it to be a large and complex area of research which is still far from resolution. This research has concentrated on the possibility of gains to be achieved by stabilization and particularly through remedial policies such as buffer stocks. Agricultural economists have often concentrated on the benefits or otherwise to producers: hence their concern with income (or even producer consumption) stability as the major focus. Host approaches use the neoclassical model of the market as the starting point for analysis. As a result many of the causes of market instability are assumed away before analysis begins. Thus 41 little attention is given to the problems of coordination in markets under real world conditions of uncertainty and hence to the question of institutional reform to improve coordination. In this study the focus will be on the question of instability with the aim of identifying deficiencies in market coordination. The approach will therefore be to consider the existing instability of agricultural commodity markets; to isolate commodity marketing systems which show excessive instability and hence provide evidence of poorly coordinated market processes; and attempt to come to some understanding of the sources of instability across commodities. To do this it will first be necessary to find a measure of instability. The next chapter is devoted to this task. Chapter 4 MEASURING INSTABILITY 4.1 INTRODUCTION In this chapter I will consider the problem of selecting an empirical measure that can be used for the description and analysis of instability. In a previous chapter it was suggested that an instability measure is desired as a proxy for ‘coordination effectiveness’. It was also suggested that such an instability measure should have characteristics both of a measure of variability and also of unpredictability. Given this goal, in this chapter, I describe and discuss a number of the single variable instability measures used in the literature. It will be seen that each has certain strengths and weaknesses as appropriate instability measures for this study. 0n the basis of this discussion I will suggest a new measure which has certain desireable characteristics. I will then provide a detailed description of this method and give an illustration of its use. I will then compare this measure with one of the more common measures in use. This comparison will be undertaken using part of the data set for this study. It will be seen that no single measure is entirely adequate for our task. 0n the basis of this analysis and 42 4 [-1 discussion I will reach some conclusions about the choice of measures for the rest of this study. 4.2 INSTABILITY MEASURES USED IN THE LITERATURE Even a quick perusal of the relevant literature reveals a multitude of methods for measuring instability. There is no generally accepted method. This reflects a lack of consensus on what instablility is as well as a desire to match the method to the problem and purpose at hand. In this section I will list and describe some of these methods. Some of the single variable methods to appear in the literature are: 1. Variance 2. Coefficient of variation (CV) 3. Coefficient of variation about a trend (CVT) 4. Absolute coefficient of variation formulation 5. Firch measure 6 Coppock index 7. Average percentage change measures 8. Moving average measures 9. Tweeten’s uncertainty index 10. Percentage range 4.2.1 CHARACTERISTICS OF INSTABILITY MEASURES 1. Variance Variance is the most commonly used measure of variability. It is simple to calculate and to interpret. Moreover mathematical and statistical techniques are well developed to manipulate it. It has however a number of drawbacks as a measure of instability in the present 44 context. First it is dimensioned in the (square of the) units of the original series. The fact that it is the square is easily solved by using the standard deviation in cases where that is desireable. However the use of a measure dimensioned in the units of its own series makes comparison with other series difficult. For example the variance of the US population is many magnitudes greater than the variance of corn price. However most would agree that the corn price is more ‘unstable’ in some sense than the US population. Also a change of units will change the variance without changing the underlying character of the series. Even comparison of variances of a series at different times may present difficulties if the relative ‘size’ changes. Another difficulty with the variance measure is that it implicitly includes trend in its measurement. For example, if there is a constant increase in a series each year, then the variance measure will register this as deviation from the mean, and hence contributing towards the variance measure. For many purposes such a series may be considered very stable: certainly it is a very predictable. In this case the variance measure will be a measure of relative trend rather than a measure of instability. Some analysts alleviate this difficulty by choosing short time periods over which to calculate the statistic (eg Tweeten, 1983). This entails some cost in terms of accuracy and is only marginally successful in solving the difficulty. Another characteristic of the variance measure (and measures based 45 on it) is that the squaring accentuates the effects of outliers. This implies a quadratic loss function which may be appropriate but is clearly a disadvantage if an outlier is an error. Despite these difficulties the variance is the most used measure of instability (eg Tweeten, 1983; Piggott, 1978; Myers and Runge, 1985). 2. Coefficient of variation (CV) The coefficient of variation is the standard deviation of a series divided by its mean. The measure is dimensionless and standardized (by the division by the mean). It therefore overcomes a significant disadvantage of the variance as a measure of instability. Instability of quite different series (such as US population and corn prices) may be compared using this measure even though the underlying units of the series are very different. However this gain is obtained at a cost. This measure is not nearly so easily mathematically manipulated as the variance measure. For example the decomposition of instability undertaken by Myers and Runge (1985), and described in Appendix E, is not possible using this measure. Like the variance, the coefficient of variation does not abstract from trend. So this difficulty is not solved by this method. 46 3. Coefficient of variation about a trend (CVT) To overcome this last difficulty some researchers (e.g., Mehra, 1981) remove the trend from the series and then form an instability measure as the standard deviation of the residuals divided by the mean of the original series. (Not the mean of the residuals which would have to be zero or close to zero). There are many ways, however, to detrend a series. Commonly an ordinary least squares regression is used to remove a linear trend from the raw data or an exponential trend can be removed from logarithmically transformed data. One difficulty with these methods is seen statistically by the autocorrelation of the residuals. Consequently the measure will give accentuated weight to series where there are long cycles or where trends change. It would seem to have disadvantages as a proxy for ‘predictability’ as it implicitly assumes that agents know the long term trend before the trend is established and that they expect at each period for an immediate return to the long term trend. 4. Absolute Coefficient of Variation Formulation This is another variation of the CV. Instead of including the sum of the squares of the deviations from the mean this measure substitutes the sum of absolute deviations. This measure gives a lower weight to outliers than the CV. It is more sensitive to the difference between dispersed and compact series. 47 5. Firch measure This measure is described and used in Firch (1977). It is one of the more useful measures for the study of instability in this list, and the measure which I will use is very similar to this one. He uses the variance of the first differences of the natural logarithms of the data series. The Firch measure is dimensionless, abstracts from an exponential trend, gives a lot of weight to short term movements and can be decomposed among multiplicative components. A problem with this measure is that the mean change about which the variance is calculated depends exclusively on the first and last data points. Hence it can be rather unstable depending on the choice of end points. 6. Coppock index Coppock’s main concern is trade instability (Coppock, 1962). His measure is the antilog of the square root of the Firch measure above. The index has the same problem with the end points (see Offutt and Blandford, 1983) as the Firch measure but its added complexity makes it less manipulatable than the former measure. 7. Average percentage change method There are a number of these methods of which Offutt and Blandford (1983) describe three. These are: (i) the average of the absolute value of the percentage period to period change; 48 ii) the average of the square of the percentage of period to period changes; iii) the same as ii) except that the percentage is calculated over the beginning or the end of each interval depending on which is greater each period. Each of these methods measures the period to period (i.e. short run) variability. They are thus partly indicies of unpredictability. None of the three make any allowance for trend in the data series. The second measure is the most manageable for manipulation. The first measure gives more moderate treatment to outliers than the others. The third measure gives symmetrical treatment to increases and decreases which the others do not do. 8. Moving average method This is the average of the absolute value of percentage differences of each data point from its (centered) moving average. The period over which the moving average is calculated is typically 3 or 5 years. Even more than others, this method is a measure of short run instability. It gives very little weight to intermediate-run and cyclical fluctuations. 9. Tweeten’s uncertainty index Tweeten uses this index in Tweeten (1981) to describe the increasing instability of aggregate excess demand for US farm product. The measure is the absolute average annual 49 percentage change minus the algebraic average percentage change. The second term is deducted to compensate for ‘the extent that average changes are part of a predictable upward trend that does not surprise market participants.’ In fact the method will give a zero value for any monotonically increasing series and is therefore not a good measure for this study. 10. Percentage range There are two versions of this simple measure. The first is the difference between the lowest and highest values expressed as a percentage of the midpoint of the extremes. The second version is the difference between the smallest and largest absolute percentage changes. Both measures give relatively little information about the series. Both are likely to be dependent on the length of the series, and to be strongly affected by outliers. Neither makes any allowance for trend. 4.3 THE INS METHOD It is clear from the above discussion that no measure is perfect for the purposes of this study. Any measure that might be used can only provide a proxy for instability. In this section I will describe yet another measure which has certain advantages for this study. I will call it the INS measure in the absence of a more descriptive term. It draws on some of the features of the Firch and the average 5% percentage change methods. The measure may be defined as the variance of annual percentage changes. Mathematically it is Var(100'dO/Q). Computationally it will be useful to approximate dO/O as follows: 92 = Q;;QL;i Q (Qt+0t-1)/2 This is analogous to (half) the definition of an arc elasticity: d0 P = (Of‘Qf—1) (Pt+P+-1) Q d? (Qt+Qt—1) (Pt‘Pt-1) The variance of (do/0) is calculated in the usual fashion: Var A = z (A -z A/n)2/n where A = (do/0) The use of the midpoint of the change as the base for calculating the change has two advantages. First it gives symmetrical treatment (and bounds) to increases and decreases; so that dO/Q lies between -2 and +2. Second, it allows decomposition of a variable, such as quantity, into yield and area components with less residual error than would occur from using the initial point as the base. This measure effectively exponentially detrends the series. Thus if the series increased by a constant percentage each year, then there would be a zero variance. The economic implication of this sort of measure is that market participants can readily adjust to constant percentage increases each year, but they will have difficulties if period to period percentage changes are highly variable. It has therefore some of the qualities of an index of unpredictability as well as being an index of variability. This measure shares a disadvantage with the Firch measure in that much weight is given to the end points when the mean change is calculated, although more weight is given to the intermediate points with the INS measure than is true of the Firch measure. 4.3.1 DECOMPOSITION OF INS An advantage of the INS measure for analyzing sources of instability is that it can be decomposed into multiplicative components. For example, production quantity of crops is a product of area and yield. 0 = A H Y hence d0 = AdY + YdA and dQ/Q = dY/Y + dA/A Now the formula for variance of a sum (D = B + C) is Var D = var B + var C + 2cov(B,D) thus var(d0/O) = var(dY/Y) + var(dA/A) + 2 cov(dY/Y,dA/A). (See Goldberger, 1970; Sackrin, 1957; Bohrnstedt and Goldberger, 1969.) To illustrate the INS measure, and its decomposition, it is useful to look at some simple data. Suppose quantity, area and yields for a commodity for 4 years are as follows: Year 0 A Y 1 100 100 1.00 2 120 110 1.09 3 140 120 1.17 4 135 110 1.23 In this illustration yield is moving gradually, though not uniformly, upward. Most of the variability in quantity is due to variability in areas. Now consider the period to period percentage changes, where the growth is computed at the mid point of each pair of periods. Hence the first datum for quantity is: 100u(120-100)/((120+100)/2) = 18.2% i.e., between period 1 and period 2 quantity increased by 18.2% calculated over the midpoint of the periods (at 110). This can now be done for each datum, and the transformed data are now: Year dQ/Q dA/A dY/Y 2 18 9 9 3 15 9 6 4 -4 -9 5 These have means, and variances about those means as follows: dO/O dA/A dY/Y Mean 10 3 7 Variance 94 71 2 Covariance (dA/A,dY/Y) 10 Note that the percentage changes for area and yield each year add to the percentage change in quantity whilst in the original data the relationship was multiplicative. This Ls! 5 also holds for the means. The means also show that areas have grown 3% per year whilst yields have grown by about 7% per year on average. The variances of the transformed data show that there is relatively little instability in yields and most of the quantity instability can be explained by instability in areas. This was also observed from the original data but this methodology allows us to apportion the instability using the formula for decomposition of a sum: var(dQ/O) = var(dY/Y) + var(dA/A) + 2 cov(dY/Y,dA/A) i.e., 94 = 71 + 2 + 2 a 10 or, in percentage terms, the source of instability in quantity is as follows: area 76% yield 2% interaction 21% Total 100% The interaction term arises because of some correlation between changes in areas and yields and therefore not all the quantity instability can be uniquely apportioned between the two components. It is interesting to consider the means, variance and coefficient of variation of the original data: 0 A Y mean 124 110 1.121 variance 242 50 .0072 CV .125 .064 .076 54 The variance of a product 0 can be related to the variances of A and Y but the arithmetic for a multiplicative relationship is complex and depends on mixed moments (see Goodman, 1960; Hazell, 1982, 1984; Burt and Finley, 1968). The variances give a measure of instability but they are not comparable between series; they need to be scaled. The coefficient of variation provides a scaling. However the CV measure can not be directly apportioned between area and yield components. Note that the CV for yield is almost as high as that for area. The reason is that the yield CV implicitly includes a trend component. It is calculated around the mean of the series. Thus most of the variability is trend rather than instability. The area data, on the other hand, show little trend. This example shows some of the characteristics of the INS measure and reveals some of the difficulties of instability measures in general. 4.3.2 CHARACTERISTICS OF THE INS MEASURE The previous discussion allows some comments to be made about this measure for the analysis of instability. Its chief advantages are that: - it is dimensionless and so can be used to compare different series and different commodities; - it is a detrended measure; removing an exponential trend from the data; - it can be decomposed into multiplicative parts relatively easily: 55 - it has some intuitive appeal as an index of unpredictability because it implicitly assumes that the next period will grow from the current period at the average rate of growth of the series; - it gives more weight to period to period fluctuations which are possibly what most would mean by instability, and relatively less weight to long term cycles. Among the measure's disadvantages are the following: - it gives excessive weight to outliers and hence to data errors; - the detrending process gives greater weight to the end points; — it requires data to be relatively precise, eg to have about three significant figures, since percentage changes must be calculated each period; - it implicitly assumes that agents know the long term exponential trend. As described earlier in this chapter the CVT measure is among the more useful of the existing measures for our purposes. When these characteristics of the INS measure are compared with the CVT measure we see that they both share some advantages and disadvantages. However the INS measure is possibly superior on the grounds that it is more empirically manipulable, it has elements of an index of ‘unpredictability’ and the fact that it gives more weight to short run phenomena which are more easily described as 56 instability. For these reasons I will use the INS measure as the major measure for analysis in this dissertation. 4.4 COMPARISON OF INS AND CVT MEASURES In this section I make an empirical comparison of these two measures to come to some better understanding of the relationship between them. I do this in two ways. First, I conduct some statistical tests on the actual data set of the study to see how the two measures relate and how they differently rank the commodity data. Second, I create some synthetic data with known statistical characteristics and analyze the two measures as applied to this synthetic data. Then I will overlay these results with actual commodity data points to see how the two measures compare. 4.4.1 STATISTICAL ANALYSIS In this section I will compare the CVT and INS measures using some simple statistical tests. First a simple statistical regression is done on two series (production and deflated prices) and then the ability of the two measures (INS and CVT) to rank the data is compared using a non- parametric test. For each of 108 commodities, the INS instability measure is calculated for annual production for the period 1950-82. For the same data set the CVT instability measure is also calculated. When the INS measure is regressed on the 57 CVT measure for both these quantity data, the results are as follows: INSQ = -316 + 4.51 CVTQ R2 (-4.0> (10.7) N .52 Corr = .72 108 A similar regression may be done for the 105 commodities with deflated price data: INSPD = -202 + 3.28 CVTPD R2 = .49 Corr = .70 (-3.1) (8.6) N = 105 The numbers in parentheses are t-values. The correlation coefficients are Pearson measures. They are reported here for comparison with the Spearman rank correlations below. While there is clearly a correlation between the two measures it is not very strong. Moreover the significant intercept term is suggestive of misspecification. It is possible that a stronger relationship between the two measures might be obtained if the INS measure was replaced by its square root. The INS is a variance, and hence a square, measure and it might be expected that the root would show a closer relationship with the CVT measure, which is based on the standard deviation. When the square root of the INS measure was regressed on the CVT measure for both the quantity data and the price data, the results were as follows: INSQ = 1.80 + .0930 CVTQ R2 = .64 Corr = .80 (1.4) (13.7) INSPD = 2.49 + .0825 CVTPD R2 = .63 Corr = .79 (2.0) (13.3) 58 There is an improvement in the relationship, however the correlation is not perfect. It is clear that the two measures are quite different. Does this difference extend to the way the two measures rank the instability of commodities? To answer this question the Spearman non-parametric test is conducted on the way the two measures rank the instability of commodities. This is of concern for this study because one of its objectives is to rank the instability of commodities in order to identify possibly poorly coordinated market processes for future research and analysis. Spearman Rank Correlations Quantities .82 Prices .83 These are similar to the Pearson parametric correlations and again demonstrate that while there is a high correlation between the way the two measures rank the commodities, the measures are not at all identical in what they are identifying as instability. 4.4.2 SIMULATION STUDY In this section I will describe a Monte Carlo simulation study using artificial data to examine the relationship between the CVT and the INS measures. It would seem possible that the existence or otherwise of 59 autocorrelation between successive data points in a series might provide some of the explanation for the difference between the two measures. With this in mind, a series of random numbers were generated with a normal distribution and known means and variances. Serial correlation was imposed on successive data points. No trend was imposed on the data. In this way the empirical relationship between the two measures can be established. Each case consisted of 30 random numbers generated as described. These represented 30 period’s (or year’s) data, roughly the length of period of the actual data. For each case coefficients of variation and INS measures were calculated. These were then averaged over 100 runs and recorded. It was found that the serial correlation parameter did indeed make a significant difference to the relationship between the two instability measures. As to be expected from the previous analysis it was found that the square root of the INS measure provided a better, and an almost linear, relationship with the CVT measure. The resulting linear relationships are graphed in Figure 4.1 for various values of the serial correlation coefficient. Superimposed on the figure are the scatter of the CVT and (the square root of the) INS measures for the 108 commodity quantity data. Most of the commodities fall in the region where the serial correlation coefficient lies between 0 and .8. It would seem that the simulation does a reasonable Job in modelling important characteristics of the data for this purpose. Again the importance of serial correlation in the 6O Ampcsoav oboe ~m3p0< ecu Amazesv gasps—:Ewm 0—188 08:02 ”mmm:mkssHmu az< comm so mzas oo.vow oo.wvm oo.mmv 00.0mv oo.nsn oo.v—n oo.omu oo.om— oo.Ov— oo.~o oo.vn +----+-u--+----+----+----+----+-u--+-u--+----+----+-n--+----+----+----+-n--+una-+----+a---+----+----+. 4 1.8.uczm . m.uozs 0.01m i 4 +----¢----+----+----+«---+----+----+----+u---+----+----+-.-u+----+---u+uu--+----+-u--4----+n--u+-n-u+ Oo.mhm oo.h.m OO.mmv OO.—Ov 00.nvn oo.mmu 00.5NN Oo.mw— 00...? oo.nm Ov.w. Ow.vu Om.wn Om.mn Ou.bv Om.vm Ow.Nw On.oh comm 61 data is confirmed as an important difference between the two measures of instability. The BASIC computer program that generated these data is attached in Appendix D. Thus, this simulation study suggests three conclusions. First, a major reason for differences in the two measures is the different way they treat serial correlation within the data series. This was suggested in an earlier section and is confirmed by this analysis. Second, it confirms that there may be advantages in using the square root of the INS measure in doing analytical work rather than the raw measure. If the objective is ordinal, i.e., to purely rank commodities, then whether the square root is taken or not will make no difference. But when the measure is used cardinally then the root seems called for. Thirdly, there is a significant difference in the way these methods measure instability. It is suggested that sometimes it may be useful to use both measures. 4.5 DEFLATING THE PRICE SERIES A question which often arises when economic analysis is done is whether it is appropriate or not to deflate the price series. This study is no exception. Is the instability observable in nominal prices more or less appropriate for market coordination issues than the variability observed in real (deflated) prices? 6 ['33 In a classical monetarist world, where money is neutral, inflation represents a response to the money supply with no effects on real variables. Money is a veil and all input and output prices would move with the inflation rate. Under these assumptions, deflating the price series would reduce the amount of observed price instability equally across all commodities. Moreover deflation would be entirely appropriate as inflation would have no effect on predictability (of quantities or real prices) and would be an irrelevant part of variability. This is the usual, albeit implicit, assumption made by economists analyzing markets. Demand and supply studies are usually calculated in real variables and the estimated elasticities are therefore essentially real parameters. In this sort of world, the rate of inflation will be irrelevant to decision making concerning market decisions. The assumptions would appear to be particularly applicable to the long term where most empirical studies support the neutrality of money and the overiding importance of money supply in determining inflation rates. This argument has a proviso in that some evidence suggests that a relationship exists between real agricultural prices or costs and inflation. Institutional factors may lead to sticky adjustment of real prices to inflation. A particular example is the capital market where often nominal interest rates are fixed and repayments are Lu 6 made in nominal dollars so that capital costs, and hence investment decisions, are sensitive to the inflation rate. In addition, it is possible that agricultural prices may respond to variations in the money supply which also affects the inflation rate. Some agricultural prices are formed in auction type markets that respond more quickly to inflationary trends than other parts of the economy. In fact the auction-contract market literature would suggest that these markets are likely to have accentuated the movements in prices as money supply feeds into these markets first. (These issues are given more atttention in Chapter 7.) Moreover there is evidence that overshooting occurs (Frankel, 1984). One consequence of this is that the response to inflation might be expected to differ between commodities depending on their market structure. For these reasons it is not entirely appropriate to assume that nominal price instability can be broken down into two non‘ interacting components of inflation and real factors. However, it also seems likely that the instability of real prices is more relevant to most market participants than that of nominal prices. Prices and costs do tend to move together. (See Gardner, 1977 and Tweeten, 1983 for some analysis of the relationship between prices and costs in the most inflationary period of the mid 1970s.) Certainly a large portion of the fluctuation in nominal prices is due to fluctuations in inflation and probably most of that fluctuation is of minimal concern for coordination and for 64 predictability. For these reasons the presumption is that deflated prices are probably more relevant to questions of price instability and will be measured that way in this dissertation. In fact deflation makes little empirical diference for this anlysis. This can be seen from consideration of the relationship between nominal and real prices and from the correlation between the measured instability of the two data sets. If nominal prices are decomposed among inflation and real prices as follows: P = PD * CPI and, using the decomposition technique described in the last chapter: var(dP/P)=var(dPD/PD)+var(dCPI/CPI)+2*Cov(dPD/PD,dCPI/CPI) For the 33 year period, the average values of these terms are as follows: var(dP/P) 397 var(dPD/PD) 397 var(dCPI/CPI) 12 2*cov(dPD/D,dCPI/CPI) ~12 Thus the instability measure is not very much different on average whether for nominal prices or for deflated prices. Moreover the simple correlation between the nominal and deflated price series is .999. Thus analysis will be very insensitive to whether prices are expressed in nominal terms or in deflated terms. 65 This is not always the case. Myers and Runge (1985) have recently done a series of studies decomposing price and quantity instability among supply and demand components for corn, wheat and soybeans. They use an adaption of a methodology first used by Piggott (1978). They choose not to deflate their price series. However their results are particularly sensitive to this methodological choice as can be seen from the discussion in Appendix E. 4.6 CONCLUSION In this chapter I have discussed a number of contenders as measures for the analysis of instability. The INS measure was developed and seen to be a useful one for the purposes of this study, but it too has deficiences. Comparison of this measure with the CVT measure showed that although there were similarities, they were quite different and that it might be useful to use both methods on occasion. Consequently it is intended to use both the INS and CVT methods for the classification study in the next chapter. However the analysis in Chapters 6 and 7 requires the choice of one method: so I will use the INS method which has the advantages that it is mathematically tractable and that it has more the characteristics of an unpredictability measure than its rival. While there may be some debate about the appropriateness of deflating price series, in this analysis there is little empirical difference. However, it is decided 66 that the balance of the argument is in favor of deflation and so the analysis is pursued in real terms. Chapter 5 THE EXTENT OF INSTABILITY AMONG COMMODITIES 5.1 INTRODUCTION This chapter is devoted to a classification of commodities according to their degree of market instability. A major reason for doing this study is to identify unstable commodity markets. Such an understanding could focus analysis on the marketing arrangements for some of these commodities that may be amenable to institutional reform. But it is not only the unstable markets that are of interest. Analysis of stable commodities may enable understanding of what contributes to successful coordination in agricultural markets. The chapter is organized as follows. In the next section the data sources for this study are discussed. Then quartile analysis is employed to identify stable and unstable commodity markets using the CVT and INS measures: the two instability measures described in the last chapter. The initial description uses an aggregate index of market instability. This is then decomposed among price and quantity components to enable an appreciation of the type of market instability that is present. The following two sections consider quantity and price instability separately 67 68 giving attention to what the INS and CVT measures reveal about the choice of the set of unstable commodities. Then a listing is provided of the more stable commodity markets identified by the analysis. Another section asks whether the choice of time period makes a difference to the instability rankings. The penultimate section compares these quantitative measures of agricultural commodity market instability with the experience of other non-agricultural markets. This provides some perspective to the magnitudes of the instability evident in agricultural markets relative to other economic series. Some concluding comments are made in the final section. Since a large number of commodities are considered, the presentation of the results is necessarily rather unwieldy. To facilitate interpretation, the results are presented in two places. Summary tables are provided in the text of this chapter whilst the comprehensive results are presented in Appendix C. 5.2 DATA The data for this chapter are mostly taken from various issues of Agricultural Statistics. The quantity series are total utilized production. Price series are gross farm prices deflated by the Consumer Price Index. The time period for data analysis is from 1950 to 1983. When this length of data is not available, then a shorter period was used. Thus, where a trade-off was necessary between comprehensiveness 69 and comparability, the choice was generally made in favor of comprehensiveness. The decision to do so was made on the basis that comprehensiveness will facilitate the work of those identifying commodity markets for future analysis. However only those commodities that had an annual gross value of production in excess of one million dollars in some part of this period were included in this study. The non- agricultural data are taken from The Economic Report of the President. Details on the length of each series are provided in Appendix A. For each agricultural data series two instability measures, the INS and CVT measures, were calculated. This was performed using the Times Series Processor (TSP) programming package. In each case a linear trend was removed in calculating the CVT measure. The cross-sectional analysis of this and later sections was done using the Statistical Package for the Social Sciences (SPSS). 5.3 CLASSIFICATION METHODS There is, of course, no unambiguous means of deciding which commodity markets are more stable than others. Nor is there any objective bench mark against which the instability of individual commodity markets can be compared. The best that can be done is to provide a number of ways of ranking these commodity markets according to some measures of instability. In this section I will suggest a number of 7% criteria that are based on the instability measures discussed in the last chapter. The first measure used is a composite index of ‘market instability’. This measure incorporates both price and quantity instability measures and utilizes both the INS and CVT definitions. It therefore has four components. This index is one fourth of the sum of each measure divided by its individual mean. This procedure is used to scale the two measures and give them equal weight in a composite index. Hence commodities with average instability will have an index value of 1.00. Unstable commodities will have higher values and stable commodity markets will have lower values. Before presenting the results it may be useful to discuss the implications of such a measure. In the previous chapter it was shown that there was a high correlation betwen the CVT and INS Vmeasures of instability. To do such a transformation it is necessary to assume a somewhat stronger measuring rod than a purely ordinal measure. It is not possible to claim that the measures proposed have a one to one cardinal representation with either ‘market instability’ or ‘coordination effectiveness’. Nor do the arguments advanced in the last chapter make it obvious that the two identified measures should have equal weight in the proposed representation function. However, in the absence of a better available alternative I propose to use this index initially to rank 71 the 105 commodities of this study. In support of this compostite index the following technical and conceptual comments may be made. First, it was demonstrated in the last chapter that the square root of the INS measure was a more appropriate measure to compare with the CVT one than the original INS measure. Second, it can be seen from Table 5.1 that standardizing the two series by dividing each series by its mean gives similar standard deviations and skewness measures. Subsequent aggregation of the two measures should therefore give a reasonably meaningful index. The question remains whether it is appropriate to give equal weight to each of the components on conceptual grounds. Should the INS and CVT measures be equally weighted? Should the price and quantity components be similarly equally weighted? If the index alone was to be relied upon these would be significant questions. However the index will be used here only as a first approximation. I will consider the information provided by the individual components: price and quantity, and CVT and INS, later in this chapter. In addition it will be seen that it is possible to use quartile analysis to group commodities without resorting to composite indicies. \1 [TI Table 5.1 CHARACTERISTIC STATISTICS OF COMPONENTS OF INDEX Standardized Standard Skewness Range Series1 Deviation INS (Price) .54 .89 .22 - 2.84 CVT (Price) .50 .40 .15 - 2.41 INS (Quantity) .68 1.56 .10 - 3.92 CVT (Quantity) .61 1.46 .15 - 3.66 Composite Index .50 1.04 .19 - 2.63 1The series are as described in the text. Each series is standardized by dividing by its mean so that all standardized series have means equal to 1.00. 5.4 AGGREGATE INDEX OF INSTABILITY As can be seen from Tables 5.1 and 5.2, the aggregate instability index on the commodities under study range from .19 to 2.63: a very wide range. The instability values for each of the 105 commodities are presented in four quartiles in Table 5.2. The results are instructive in their diversity. The first five commodities have double the measured amount of instability of the ‘average’ commodity and 13 times the instability of the most stable. Indeed the wide differences in instability between commodities is supportive of the efficacy of an inter-commodity approach to the study of market instability. The two most unstable commodities are both subtropical tree crops, namely olives and avocados. Perhaps it is not surprising that these should head the list as they are crops with very long lags from ’73 Table 5.2 COMMODITIES GROUPED BY QUARTILES ACCORDING TO COMPOSITE INSTABILITY INDEX III QUARTILE I QUARTILE II QUARTILE No COMNAME POINSTAB ”0 COMNAME POINSTAB 1 OLIVES 2.53 27 SUGARCAN 1.31 2 AVOCAROO 2.54 28 PLUMS 1.28 3 SUNFLowR 2.45 29 ORANGES 1.27 4 TRT CHER 2.25 30 VEAL 1.25 5 PECANS 2.09 31 NECTaRIN 1.22 5 DRY PEAS 1.98 32 PEPPERMT 1.22 7 ALMONOS 1.95 33 PEARS 1.20 8 TEMPLES 1.54 34 vaLNUTs 1.18 9 SPEARMNT 1.53 35 FIGS 1.14 10 FLAXSEED 1.51 35 ONIONS 1.13 11 LIMES 1.57 37 GRAPEFRT 1.13 12 FILBERTS 1.57 38 POTaTO 1.12 13 POPCORN 1.50 39 sucaRBET 1.10 14 TANGELOS 1.49 40 COTTON 1.09 15 GARLI 1.48 41 SVT CHER 1.09 15 BUSH BER 1.47 42 GRAPES 1.09 17 vs 1.42 43 WOOL 1.08 18 TANGERIN 1.38 44 WHEAT .1.05 19 PaPava 1.37 45 DATES 1.05 20 LEMONS 1.35 45 RICE 1.05 21 PRUNES 1.35 47 sz POT 1.04 22 CTONSEED 1.35 48 MACADAMI 1.03 23 APRICOTS 1.34 49 ALFALFA 1.02 24 DRY BEAN 1.34 so BARLEY 1.00 25 SORGHUM 1.33 51 SOYBEANS .97 25 POMERGRN 1.33 52 CRANBERR .95 IV QUARTILE 53 CABBAGE .92 79 PICKLES .62 54 PR TOMAT .92 8O BEEF .62 55 CORN .90 81 CARROTS .58 55 oaTs .89 82 STRaw85R .58 57 ARTTCHKE .89 83 ASPaRacs .57 58 HOPS 85 84 CELERY .57 59 HONEYDEW .83 85 PR SPNCH .57 50 8Rocc0LI .83 85 BANANAS .57 51 APPLES .82 87 PR SNPBN .55 52 LAMB MUT .82 88 CANTALOP .55 53 8Rus SPR .80 89 FR CUCUM .54 54 FR SPNCH .80 90 FR TOMAT .52 55 BEETROOT .79 91 MUMS .51 55 PORK .79 92 HAY .49 57 EGGPLANT .77 93 MINI MUM .49 58 PR sw CN .75 94 ORN PEPP .45 59 LIMA BEN .75 95 Tosacco .45 7o ESCAROLE .75 95 LETTUCE .45 71 PEANUTS .75 97 555 .45 72 MAPL SIR .75 98 CARNATNS .43 73 WATERMEL .73 99 TARO .43 74 PEACHES .71 100 F SNPBEA ~41 75 TURKEY .71 101 FR sw CN .35 75 BROILERS .71 102 MILK .28 1; gr 182 584.1333 ~32 AULIFLR .53 - 105 TEA ROSE .19 1 The composite instability index gives equal weight to prices and quantities, and the INS and CVT measures. It has a mean of 1.00 where high values represent less stable commodities. 74 planting to production. Olives may continue to produce for 2000 years. But then the third most unstable market is an annual crop namely sunflower seed. Both the most stable and the most unstable markets seem to be commodities with relatively low gross value of production. The more important commodities tend to be in the middle of the list. But then milk, with a very stable market, is an exception. In fact it is relatively difficult to make general statements about this list without finding many exceptions. The next chapter will provide multivariate analysis to determine what order can be gleaned from these instability measures concerning differences between commodity markets. 5.5 PRICE AND QUANTITY INSTABILITY The last section provided composite (price and quantity) market instability indicies. In this section the series will be disaggregated and information on both price and quantity instability will be presented. Again I will examine indicies using both the INS and CVT measures. As might be expected there is correlation between the price and quantity instability series. The Pearson correlation coefficient between the price and quantity instability indicies is +.58 and is significantly different from zero at the .001 level. This indicates that the more quantity unstable commodities also tend to be the more price unstable: a not unexpected result. In practice the correlation means that any listing of commodities appearing 75 in the upper quartiles of either series will have a high degree of overlap. This is indeed the case. Table 5.3 presents a listing of the 40 commodities that appear in the top quartiles for either quantity instability or price instability. The first column lists those 12 commodities which appear in both the top quantity and price quartiles. The second column lists those 14 which appear in the top quantity instability quartile and not in the corresponding price instability quartile. The third column provides a listing of those which appear in the top quartile for price instability but not for quantity instability. Most of these commodities also appear in the top quartile of the composite measure described in the last section and listed in Table 5.2. These include all those in the first column and those with asterisks in the other two columns. Of all the 40 commodities listed here only broccoli does not fall in the first two quartiles of the composite instability index shown in Table 5.2. The listings are in descending order of instability in each column. For the interested reader, Appendix C gives values for all commodities from which these tables are compiled. 76 Table 5.3 UNSTABLE COMMODITIES: AGGREGATE INDICIES1 Both Quantity only Price only Olives Spearmint! Sugar cane Avocados Filberts! Bush berriesP Sunflower seed Garlicu Onions Tart cherries Rye“ Potatoes Pecan Papaya! Dry edible beans9 Dry edible peas Apricots.1 Wool Almonds Sorghum! Grapefruit Temples Prunes! Sweet cherriesl Flaxseed Macadamia nuts Lemons. Limes Tangerines* Oranges Popcorn Pomergranitesa Figs Tangelos Plums Pears Broccoli Sugarbeet Nectarines Peppermint 1 Commodities listed are those in the top quartile for instability as measured with the aggregate index of the INS and CVT measures. They are listed in descending order. Commodities in the first column appear in the top quartiles for both quantity and price. Commodities in the first column and those with asterisks in the second and third columns appear in the top quartile of the composite measure as listed in Table 5.2 77 This analysis allows identification of markets which are quantity and/or price unstable. Commodities which are price unstable but not quantity unstable may be so because of highly inelastic demand relations. These may provide interesting cases for marketing studies. However for some of these commodity markets the price instability may not arise from this source but rather from inherently unstable marketing mechanisms. These will also be of interest for marketing policy. 5.6 GREATEST QUANTITY INSTABILITY In this section the commodities that exhibit the greatest quantity instability will be considered. A similar methodology for identifying these commodities will be used as was done in the last section. Using both the INS and CVT measures those commodities that appear in the top quartile according to both measures will be identified, as will those that appear in the top quartile of one of these measures and not the other. These results appear in Table 5.4. 78 Table 5.4 QUANTITY INSTABILITY: INS AND CVT UPPER OUARTILES Both Measures INS Only CVT Only Olives Sunflower Prunes Macadamia Nuts Avocados Pecans Tangerines Broccoli Almonds Filberts Pomergranites Nectarines Dry Ed Peas Tart Cherries Plums Fresh Spinach Spearmint Garlic Sweet Cherries Honeydews Flaxseed Rye Lemons Veal Limes Popcorn Cotton Walnuts Temples Papaya Apricots Sorghum Tangelos This analysis is instructive for a number of reasons. First it identifies the commodities which exhibit the greatest quantity instability. Secondly, the use of two measures provides an indication of the type of instability experienced by some of these commodities. For example, those commodities that have a high ranking on the INS measure and a lower ranking on the CVT measure can be expected to exhibit greater instability from year to year than those that do not. While those that score highly on the CVT measure and not on the INS measure will have experienced changing trends or cyclical behavior over the period of the study. Thus it is not surprising that almost all those in the INS-only column are tree fruit with the more geographically concentrated growing areas. A characteristic of many tree crops is that they have a two year cycle as a good production year causes depletion of the sugars in the 79 plant which makes lower production in the following year more likely. 5.7 GREATEST PRICE INSTABILITY In this section those commodities that exhibit the greatest price instability are presented. The methodology is identical to that of the previous section. Those commodities that appear in the upper quartile for price instability for both (in the first column) or for either of the INS and CVT measures (in the second and third columns) are noted in Table 5.5. Table 5.5 PRICE INSTABILITY: INS AND CVT UPPER QUARTILES Both INS and CVT Measures INS Only CVT Only Tart Cherries Avocados Pecans Flaxseed Sugar Cane Olives Pears Popcorn Dry Edb Peas Bush Berries Plums Figs Sunflower Seed Onions Cabbage Sugar beet Dry Edb Beans Temples Pomergranates Peppermint Potatoes Wool Tangerines Spearmint Almonds Tangelos Prunes Cranberries Grapefruit Cottonseed Limes Lemons Oranges 5.8 MOST STABLE COMMODITIES Up to now the discussion in this chapter has concentrated on the identification of unstable markets. But, as was suggested at the beginning of this chapter, it would also be very useful to know which markets are stable. 8% Further research may profitable consider these markets in order to determine what makes them to be stable. This may be useful information in the development of improved coordinating mechanisms. This might not always be the case: often the factors that make one market stable are commodity specific and are not characteristic of other markets. Moreover some markets may be stable at very high costs so that there may be little desire to transfer this experience to other markets. Milk may fall in this category. However these questions should follow detailed examination of individual markets. In this section I will present the group of commodities that have both stable prices and quantities. Selections for further research might be made from these lists. Table 5.6 STABLE COMMODITIES: QUANTITY AND PRICE UPPER QUARTILES Both Measures Milk Hybrid Tea Roses Mushrooms Fresh Sweet Corn Potted Mums Taro Fresh Snap Beans Green Peppers Carnations Quantity stable Eggs Lettuce Celery Hay Broilers Fresh Tomatoes Carrots Turkeys Beef Pork Onions Fresh Cucumbers Wool Sugarcane Watermelon Price stable Tobacco Miniature Mums Broccoli Maple Sirup Peanuts Bananas Proc Spinach Cauliflower Chrysanthemums Fresh Spinach Lima Beans Asparagus Strawberries Honeydew Peaches 81 As in previous tables, the first column lists those commodities that appear in both the most stable quartile for quantity as well as for price. The second column lists those commodities that are in the most stable quantity quartile but not in the most stable price quartile. Similarly, the third column gives price stable commodities. It is noteworthy that the floriculture crops from the commodity set appears in these lists. It would be interesting to know how these industries maintain such stablility. It is also noteworthy that milk, which has possibly the most extensive support and control mechanisms, is the most stable commodity market. This suggests that the program is quite successful in this respect. The tobacco program also appears to have been very successful in stabilizing price. It may be surprising to see some vegetables with very short growing seasons, such as lettuce, in this list. The instability measures used here are annual measures. Lettuce may exhibit high market instability during. the year, but this can not be captured in annual averages. It will be seen in the next chapter that the apparent stability of the lettuce market, implied by these annual averages, hides a coordination problem that annual area data helps to elucidate. Another observation that can be made from this analysis is the appearance of a number of animal commodities among the list of quantity stable commodities. This is reflective 88 of the more certain yield relationships for these commodities. 5.9 RECENT INSTABILITY It is worth knowing whether the commodities identified here as unstable, based on analysis of a 33 year period, are the same ones that would have been identified in the latter third of the period. In other words, to what degree is this classification dependent on the time period chosen? Differences between the two periods may reflect changing relative instability over time, or they may just be reflective of some lack of robustness in the analysis. The question of changing instability will be addressed in Chapter 7; the concern here is whether the more recent period gives a different ranking or identifies different commodities as being unstable or not. In fact the data reveals a reasonably high degree of agreement in the rankings of price and quantity instability between the two periods. The Spearman rank correlation between the two periods for the INS measure is 0.91 for quantity and .89 for price. Examination of the top quartiles for the INS measure for price and quantity shows some changes in representation but is not radically different. Of the twenty-six commodities of the top quantity quartiles twenty-one are common to both periods. For the top price quartiles Table 5.7 CHANGES IN REPRESENTATION OF UPPER INSTABILITY QUARTILE IN RECENT PERIOD1 QUANTITY PRICE Total period Recent period Total period Recent period Temples Cottonseed Temples Flaxseed Tangelos Lima Beans Tangelos Popcorn Tangerines Peanuts Tangerines Sugarbeet Pomergranites Grapes Grapefruit Rice Sweet Cherries Artichokes Limes Figs Oranges Walnuts Pears Wheat Cabbage Alfalfa Prunes Sweet Potatoes 1 As measured by the INS instability measure Table 5.8 COMMODITIES IN THE TOP INSTABILITY QUARTILE FOR BOTH PERIODS1 BOTH QUANTITY PRICE Olives Flaxseed Sugarcane Dry Ed Peas Popcorn Bushberries Pecans Prunes Onions Avocados Filberts Wool Sunflower Seeds Apricots Dry Ed Beans Tart Cherries Limes Potatoes Almonds Spearmint Pomergranites Lemons Cotton Sweet Cherries Rye Plums Garlic Sorghum 1 As measured by the INS instability measure. 84 seventeen of twenty-six are common to both periods. There are no changes in the top half of the quartiles. Again it can be seen that the same group of very unstable commodities reoccurs in every list. 5.10 NON-AGRICULTURAL INSTABILITY The previous sections have described the instability of certain agricultural markets. In this section the INS measure will be applied to a number of non-agricultural industries and series with the intention of placing agricultural instability in context. Many non-agricultural industries experience marked instability. However, such industries experience most of their instability as a result of fluctuations in demand: supply instability pays a much less important role. Some industries are very susceptible to the fluctuations of the business cycle (e.g., cars, new housing) whilst others are less so. The level of aggregation presents difficulties for comparitive purposes. The greater the degree of aggregation, the greater the level of stability to be expected. Production of a certain type of car is likely to be less stable than the car industry in total. Likewise food production in total will be more stable than most of the component commodities. Hence the comparison of individual agricultural commodities with the instability of entire sectors of the US economy is likely to make agriculture look more unstable than otherwise. Therefore the results in the next table should be used with some caution. 85 These data are presented to provide a context for the INS measure itself. Table 5.9 INSTABILITY IN OTHER ECONOMIC SECTORS INS Instability Measure Quantity Series Mean Agricultural Commodity 405 Consumer Cars 228 Defence and Space Equipment 296 Primary Metals Production 178 Chemical Production 33 Transportation Equipment 130 Federal Govt Expenditure 41 Business Equipment 78 Price Series (deflated) Mean Agricultural Commodity 390 Dow-Jones Index 166 Other Series Population .1 Consumer Price Index 12 Disposable Income per capita (deflated) 5 5.11 CONCLUDING COMMENTS The primary purpose of this chapter has been to identify stable and unstable agricultural commodity markets. Various instability measures on prices and quantities were used for this purpose. In addition, consideration was given to whether the ranking provided over a long (33 year) period also is applicable to more recent experience. It was found that there is a relatively high degree of agreement between the two measures and periods as to what commodity markets are unstable. In addition the different measures provide 86 information on the type of instability evident in markets. Where commodities are more unstable on the INS measure than on the CVT measure is indicative of shorter run instability (e.g., tree fruit) rather than longer term cycles or secular taste changes, which the CVT measure emphasizes. Commodities that display more price instability than quantity instability can also be identified by this analysis (and vice-versa). The final section compares the agricultural commodity market instability with that of some non- agricultural series to provide a context for their relative magnitudes. Chapter 6 SOURCES OF INSTABILITY 6.1 INTRODUCTION In the last chapter measures of the extent of instability for a large number of agricultural markets were presented. It was shown that there were wide differences in instability among the different commodities. In this chapter an attempt is made to come to an understanding of the sources of this market instability. An attempt will also be made to measure some of these sources to gain a perspective on their relative contribution to total market instability. To do this a variety of methods will be used. First, possible sources of instability will be listed and various hypotheses are suggested to explain agricultural market instability. The relative importance of yield, the most likely candidate to explain agricultural quantity instability, will then be considered. To do this a particular empirical virtue of the chosen instability measure is utilized to decompose production instability between area and yield factors. This will provide a perspective on the extent to which instability in agriculture is due to this (partly) biological charcteristic of agriculture. Consideration is then given, in the fourth 87 88 section, to the contribution to instability provided by fluctuations in the demand factors of income and population and also the supply factor of input prices. This is done by building a simple commodity market model and inserting likely values for various key parameters and noting how much instability could be obtained from the interaction of these factors. A cross-sectional regression analysis is undertaken in the fifth section to determine how much of the differences in instability among commodities can be explained by their various production and institutional characteristics. This approach has a secondary value in raising interesting questions about the nature and results of these institutional variables. In the sixth section the relationship between area, quantity and price instability is examined to see if there is evidence that price instability leads to area instability. If so then this is indicative of the role of price instability in contributing to market coordination problems. A final section provides a summary of the results and some concluding comments. 89 6.2 SOURCES OF INSTABILITY This section simply lists some possible sources of instability for agricultural markets. These may be summarized as supply, demand and institutional factors. Supply Factors Weather Other yield factors Use of inputs e.g., fertilizer Geographical distrinbution of production Diversity of genetic stock Supply Response of farmers Errors in price forecasts Asset fixity - especially with perennial crops Responses to risk Interest costs, debt equity etc Demand Factors Domestic demand shifters Income, population and tastes Prices of substitutes and complements Export demand shifters Foreign supply and demand conditions Exchange rate volatility Volatility of international liquidity Institutional Factors Structure of markets and coordinating mechanisms Government policy shifts 6.3 YIELD A major source of instability in agriculture is of course the weather. Perhaps no other economic sector has to suffer, not only uncertain demand and input supplies, but also very uncertain production functions with a large 9O stochastic component.1 It is very likely that this uncertain input and output relationship makes choice of input decisions suboptimal. In this section I will investigate the extent to which instability in agriculture can be traced to yield fluctuations. Yield variability is not just related to weather. The theory of production economics posits output as a function of various inputs: archetypically land, labor and capital. Yield, which is output over land input, will then be a function of the same inputs. In an economic context then, yield will be a function of the input prices and the available technology. In practice, however, those who model agricultural markets, usually acknowledge the high correlation of land with the other inputs and hence its overwhelming importance in explaining output. In addition, the stochastic infuence of weather on yields is implied. The usual procedure is to assume that the major decision variable is land, which depends on prices of outputs and inputs. Some analysts have had some success modeling yield as a function of economic variables and occasionally some find a proxy variable for the weather (see, for example, Gadson et al, 1982; or the MSU model). However, the usual practice is to resort to a trend alone to model the non- 1 Other sectors do, of course, experience some uncertainty about production relations. Labor economists have given particular consideration to the quantity and quality aspects of labor inputs in the literature on principal agents. However the instability due to this source is likely to be much smaller than the effects of yield. 91 stochastic elements. The trend variable is intended to encompass technological change among other factors. This practice is pursued also because of the difficulty of determining how inputs are allocated among products in multi-output enterprises. For these reasons the usual procedure, which is followed in this study, is to assume that the major portion of yield variability can be attributed to weather and therefore to treat it as being generally unrelated to other economic variables. It is recognized that this represents some oversimplification and is not entirely supported by the evidence which follows. 6.3.1 YIELD AND INSTABILITY As suggested above it seems likely that yield is a major source of market instability. Partly this instability will be a direct effect and partly it will be indirect, as a result of the poor decisions it induces. To investigate the direct effect, quantity instability for 65 commodities is decomposed among area and yield components. It should be noted that the choice of these commodities is made on the basis of data availability. In particular there is very little data available on areas of perennial crops. Moreover, animal products are necessarily excluded from this part of the study. The methodology described in the fourth chapter allows quantity instability to be decomposed among areas and 9 ['1] yields. There is also a covariance term that represents an interaction effect and which is often quite significant. How much variability in agricultural production is due to the direct effect of yields? For the 65 commoditity markets with yield information production instability can be allocated between yield and areas as shown in Tables 6.1 and 6.2. Table 6.1 provides the instability measures and Table 6.2 gives the percentage of quantity instability explained by the components. 93 Table 6.1 DECOMPOSITION OF PRODUCTION INSTABILITY 1AMONG YIELD AREA AND INTERACTION COMPONENTS COMMOD VAROO VAROY VARDA COV2AY EGP 152. 99. 151. -98. GAR 763. 105. 661. 2. WAT 89. 54. 52. -17. BRU 298. 130. 122. 50. LMA 323. 57. 217. 52. SPF 109. 57. 97. -45. SPP 195. 27. 147. 22. ART 283. 308. 29. -55. BET 306. 114. 213. '20. £50 70. 59. 104. -93. AS 56. 52. 23. '20. 08 105. 47. 37. 21. CP 67. 37. 82. -52. DEW 170. 105. 210. ~144. PEA 160. 80. 62. 18. BC 198. 80. 100. 19. CE 18. 13. 21. -16. CT 65. 54. 59. -47. SUP 233. 36. 131. 67. CR 95. 65. 44. -13. TAR 41. 56. 52. '68. P0 84. 22. 53. 9. SCF 18. 16. 18. ’16. TOP 29. 33. 26. -29. LT 10. 14. 19. '23. ON 89. 44. 82. ~38. SCP 212. 54. 102. 58. SBP 107. 31. 83. ’7. TOP 425. 109. 215. 104. CUF 58. 48. 39. '30. GP 60. 41. 59. -40. CUP 141. 41. 111. -9. SBF 40. 23. 27. '9. STW 68. 71. 59. -62. CRB 138. 137. 5. -4. HOP 179. 33. 129. 17. PEP 232. 98. 139. '4. SPR 947. 159. 563. 240. CTS 458. 154. 298. 8. FLX 1199. 557. 637. 21. POP 993. 249. 611. 151. SUN 1775. 2092. 3185. -3400. MAC 180. 265. 69. -153. PEN 403. 280. 93. 35. MUS 12. 5. 15. '8. BY 205. 49. 175. ~18. OT 215. 92. 80. 44. SB 179. 76. 46. 58. SG 859. 192. 539. 164. UH 200. 81. 124. '3. CN 132. 108. 41. '17. SOC 70. 43. 60. '33. SGB 200. 39. 145. 16. OBE 230. 83. 144. 4. OPE 1614. 1020. 548. 89. RYE 630. 104. 324. 209. RIC 156. 26. 178. -47. T08 121. 33. 60. 29. HAY 31. 18. 6. 7. COT 483. 170. 299. 17. VOL 19. 1. 16. 2. PAP 478. 260. 218. 2. BAN 182. 164. 82. '63. BB 328. 139. 61. 128. ALF 185. 205. 369. -390. l The column headings represent, in order, the commodity code, production instability, yield instability, area instability and an interaction covariance term. 94 Table 6.2 PERCENTAGE DECOMPOSITION OF PRODUCTION INSTABILITY 1 APERO INTERACT YPERQ COMMOD ................................................................. 4 976119736085109294619 0377427631392522525979029237826304341 591 6 111411.33278118721618 N42 . 256 . 29. 198 6.231.14 3322 1 33m . . . _ . — _ . u .1 1 . _ . . . 4|. 9731795009152491716673 00288178987420953298355763216334 1409246599 9854687174432235195426 99947569768 765656732283266387635151684419 1 11 1 1 1 1 1 1 1 1 ................................................................. 54148249743552002358769409596389849827465879243221210637778554021 616415103894565478163281442228625091413421464244248623611253 5941 1 1 11 1 11 1 PRTUAFPTTCSBPWACETanROF—PTNPPPFD.prrWED.FRSXPNCNSYTBGWNCBEEECBVTLPNBF GAAR”ppRESACCEEBCCWCAPCOLOCBOUGUBTROEPTLOUAEUBOSS CGGBPYIOAOOAABL EGUBLSSABE DD. 5 T ST SSTC CSSCHPSCFPSMPM SSDDRRTHCWPB A The column headings represent, in order, the commodity code, I yield instability as a percentage of production instability, area instability as a percentage of production instability, and an interaction term. Note that the percentages add to lOO. 95 These results are summarized in Table 6.3, which shows averages of the instability measures of the 65 commodities and their percentage decomposition among components. Note that the percentage decomposition is made by weighting each commodity equally (i.e., averages from Table 6.2), and not from the averages of the instability measures. Table 6.3 SUMMARY OF DECOMPOSITION OF PRODUCTION INSTABILITY Source of production Instability Percentage Instability Measure Yield 140 55 Area 196 72 Interaction -53 -27 Total 280 100 Thus yield variability directly contributes a major portion of the instability in agriculture. Before discussing these results it would be useful to know how the covariance term should be interpreted. Its presence indicates that yield and areas are related. Four hypotheses might be suggested: 1. Poor weather may lead to some areas planted not being able to be profitably harvested and hence left unharvested (e.g., grazed) when yields are low. 2. Low prices, associated with a large crop area, may lead to less area being harvested with effort being 96 concentrated on the better yielding portions of the planted area. 3. Low prices, associated with a large crop area, may lead to the area harvested being less intensely or less frequently collected. 4. When price prospects are good, production is often extended to marginal, lower yielding, areas. Clearly one factor that has a bearing on the reasons behind these relationships is the definition of area. These data are derived from statistics on “area harvested" for all these commodities except for vegetables grown for the fresh market. In this latter case the area data are ”area for harvest which includes any acreage partially harvested or not harvested because of low prices or other economic factors. Area for processing is area harvested" (Agricultural Statistics 1981 p 151). Those crops whose area measure is "areas for harvest“ are indicated in Appendix A. Each of the four hypotheses above might explain covariance between yield and area. However, the first one is the only one of the four that implies a positive correlation between yield and area instability. Such is clearly counterfactual for these commodities in aggregate. However, Table 6.2 shows that almost half of these commodities, 30 of 65, displayed positive covariance terms. Interestingly, the crops that may be utilized either for grazing or for harvest 97 have relatively high positive correlations. These include rye, hay, oats, and soybeans. The second and third hypotheses describe a common pattern in some vegetable markets. For example, the lettuce industry shows how the choice of area harvested is a means of coordination in response to prices and or yields. If there is a large area planted and prices are low, then not all of the area is harvested and/or the planted area is harvested less frequently to reduce harvesting costs, principally labor (Hammig and Mittlelhammer, 1980). As a result yields are low. For this reason quantity produced is more stable than yields or areas. In this case, yield is a controllable parameter which is used to coordinate supplies to market. However, this is achieved by committing (non- harvesting) inputs that are not reflected in output. Hence resources are wasted. Thus, the relative stability of production, on an annual basis, belies a coordination problem that leads to non—labor input costs which are not reflected in output. Thus costs and prices may be higher than under different marketing arrangements that lead to more stable areas planted. It might be expected that those vegetable crops grown predominantly for the fresh market and for which the area data are "area for harvest" rather than area harvested, might have a larger negative correlation than other crops. The second hypothesis could provide a second explanatory 9B factor in these cases. In addition, fresh vegetable crops are frequently marketed on spot markets which can be very volatile week to week. This is less true of markets for crops grown for processing, where contracting is more common. Examination of the data supports this expectation. The mean of the covariance term is -30 for vegetables reported on an "area for harvest" basis and +12 for those reported on an "area harvested" basis. Again this is indicative that some resources allocated to production do not appear in the composition of the final product since areas are left unharvested. The fourth hypothesis may be more likely to apply to some of the field crops where production is very extensive so that yields might be responsive to changes in areas. This would seem an unlikely explanation for vegetable crops that require relatively small areas individually and thus where an abundance of suitable alternative land is available. However the small negative covariances for corn, wheat and rice may be due to this factor. It seems likely that some of the explanation for the covariance terms may come from one or more of these hypotheses. For each case the covariances are true interaction terms which can not legitimately be allocated to either area or yield instability. 99 6.3.2 THE RELATIVE IMPORTANCE OF YIELD INSTABILITY Returning again to Table 6.3, it would seem that although yield is a very important factor contributing to quantity instability it is clearly not the only factor. In fact, in aggregate, instability of areas dominates instability of yield as a source of production instability. Although this result is not general for all the commodities considered here. For 22 of these 65 commodities (34%), yield instability is more important than area. In particular, the following crops exhibit much greater yield variability than area variability: corn, soybeans, peanuts and hay, and the perennial crops included in this selection, namely asparagus, artichokes, cranberries, bushberries, macadamia nuts and bananas. It might be expected that other perennials, such as tree fruits and nuts, would also show greater yield variability than area variability. However, area data are not available for these crops. If it is naively supposed that yield variability is uncontrollable, whilst area variability is controllable, or potentially controllable, then this analysis suggests that there may be potential for reducing the instability of production instability in American agriculture (of annuals) by more than half. This of course does not mean that this would necessarily be economic or desireable. However, it is indicative of the extent to which potentially controllable factors contribute to instability. 100 6.3.3 IS HIGH PRODUCTION INSTABILITY MAINLY CAUSED BY YIELDS? A question raised by this analysis is whether commodities with particularly high production instability are in this category because they are particularly suspectible to yield variation. For example, if production of some commodities were highly concentrated geographically, then yields might be observed to be very unstable without the alleviating benefit of counteracting yield affects elsewhere. Commodities with high production instability might be high for this reason alone. If this were so, it might be expected that those commodities with greater production instability would have a higher proportion of their instability deriving from yield variability. In fact, the data do not support this hypothesis. Of the 26 commodities with the most yield variability (40% of the sample), only 7 have yield variability predominant. This is not a dissimilar proportion to that for all of the 65 commodities, (27% compared to 34%). A simple correlation of production instability against the ratio of yield to area instability gives an insignificantly negative value of -.06. The data do not support this hypothesis. Hence there is no support for the contention that highly unstable commodities are such because of highly unstable yields. Unstable plantings are also important. 101 6.3.4 INDIRECT EFFECT OF YIELDS ON PRODUCTION INSTABILITY It is clear from the above discussion that yield instability has a direct effect on production instability. However, it is also possible that instability of yields may affect instability of areas and thereby have an additional indirect effect on production instability. It might be expected that if yields are unstable, decisions about areas become more difficult and more prone to readjustment. A positive correlation between yield variabililty and area variability would be supportive of this indirect effect of yield instability on production instability. It is also possible, however, that the causal relationship works in the opposite direction, i.e., those commodities with greater area instability cause them also to have greater yield variability. In fact, the previously mentioned hypotheses advanced to explain the covariance terms might also be used to suggest reasons for a relationship between area and yield variability. Hence, if it was somehow possible to allocate the covariance term between the other two components, then a positive correlation between the adjusted area and yield variability measures would be better evidence of an indirect yield effect than otherwise. However, as already discussed, there is no completely satisfactory way to allocate the covariance terms. In the absence of a better alternative it could be useful arbitrarily to allocate the covariance term equally between the other two terms. The correlation between the yield and area instability measures is +.91. When the IOE covariance term is allocated as suggested, then the correlation is +.61. Both these correlations are significant at the 1% level. Hence, whether the data are adjusted or not, there is a high correlation between yield variability and area variability. This seems to be very suggestive that yield variability not only has a direct effect on production instability but also an indirect effect through inducing poor decisions about areas to be planted. It is difficult to explain this relationship by appealling to beliefs about the ability of producers rationally to discount yield effects when making production decisions. Producers in industries with unstable yields clearly find it more difficult to decide the appropriate amount to plant each year. They therefore make mistakes. This finding supports the contention of Chapter 2 that instability begats instability. 6.3.5 SUMMARY OF YIELD EFFECTS In this section production instability is decomposed among yield, area and interaction effects and the influence of each are investigated. Although yield is an important explanation of production instability (at least for the commodities of the sample) area instability was a more important contributor to total production instability. Thus yield effects can not be blamed for the greater part of production instability. Moreover yield did not show a greater than proportional contribution in explaining the 103 greater instability of the more unstable commodities. It was seen that in some cases (e.g., lettuce) yield is, in fact, a coordinating mechanism. It was also determined that yield and area instability is correlated between commodities. This is suggestive that producers of commodities with unstable yields have difficulty in making appropriate choices of how much area to plant. Thus instability is seen to induce poor coordination of supply with demand. 6.4 DEMAND AND SUPPLY FACTORS In this section I wish to consider some of the sources of instability in agricultural commodity markets. In particular I wish to ask the question: "how much instability can be ascribed to those factors that are general to microeconomic commodity markets?" The factors considered here are: demand shifters, ie income and population supply shifters, is input prices There are of course other sources of agricultural instability, such as weather, taste changes, monetary forces and the instability of particular complements and substitutes in both production and consumption - all of which will impinge upon instability in any individual market. However, in this section I will consider only the above sources to attempt to gain a perspective on their magnitude. It will then be possible to compare the instability derived from these sources with the observed 184 amount of instability in commodity markets to gain a perspective on their relative contribution. To estimate the degree of instability likely to arise from these sources it is useful to build a model of a simple commodity market. Consider first a perfectly competitive commodity market where all participants are aware of the price to be received, the quantity to be produced and the levels of the supply and demand shifters. Prices and quantities may be determined by a two equation model of the form (in logarithms): supply 0 = a0 + a1P + a2Pf demand P = b0 + b1(Q - POP) + bQI where Q = quantity P = price Pf = supply shifter such as input price I = demand shifter such as income POP = population The reduced forms are: Q = D(ao+a1bo) + Da2Pf + Dalbzl - Da1b1POP P = D(b0+a0b1) + Da2b1Pf + Db2I - DblPOP where D=1/(1-a1b1) For analytical purposes it is useful to express the variables as percentage period to period changes: 0’ = (dO/dT)/Q P’ = (dP/dT)/P etc where the right hand side is approximated by: Q’ = 2*(Qt-Qt-1)/(Qt+Qt-1) etc Despite the assumption of perfect knowledge, this market will exhibit variability in response to the variability of the supply and demand shifters. var(Q) = 02822'V8r(Pf) +Dza12bzzivar(1) + D2a12b129var(POP) + D2a1a2b2Icov(Pf,I) - D2a1a2b19cov(Pf,POP) - 02a12b1b25c0v(I,POP) var(P) = D2a22b12Ivar(Pf) + 02b225var(I) + D2b125var(POP) + D2a2b1b2'cov(Pf,I) - D2a2b125cov(Pf,POP) — 02b1b2-c0v(I,POP) The dependent and independent variables in these equations are now the INS instability measure used in this study. Thus, these equations relate the instability of prices and quantities in a market to the instability of demand and supply shifters described above. If I now relax the assumption of perfect foresight and assume that producers know only the structure of this model and can make a forecast of the supply and demand shifters, the model becomes a rational expectations model. The reduced forms are identical to the previous model except that the values of the shifters are replaced by their expectations. If it is assumed that these expectations are generated by an autoregressive process, as is typically done in rational expectations models, then : 106 It“ = co + cIt-1 where the parameters may be estimated by an ordinary least squares (OLS) model: It = co + cIt—l + e In this example the variance of the forecast will be lower than the variance of the actual series: var(It) = var(It‘) + var(e) as the covariance term is zero under the assumptions of the OLS model. Consequently the variances of prices and quantities will also be lower under rational expectations than under perfect foresight. These models differ according to the degree of information available to agents. They demonstrate how increased knowledge can be a source of instability. Under perfect knowledge agents might be expected to react to every small change. However, a more conservative strategy is implied under rational expectations where changes in trends in the exogenous variable must be established before response is made. Indeed Heiner (1983), using a much more general formulation, shows that uncertainty can be the origin of predictable and stable behavior. Using this model it is now possible to estimate the extent of instability attributable to these factors. Let us now choose possible values for a1, a2, b1 and b2. The initial values for the b terms are flexibilities 107 derived from the own price and income elasticities averaged across all commodities from the demand study of George and King (1971). The simple average of the farm level elasticities of the commodities in the George and King study are -.4 and .3 for the price and income demand elasticities respectively. The initial values for a1 and a2 are guestimates. It is important to note that this is a simulation of a representative commodity market rather than of the total farm sector, and hence simple averages are appropriate. a1 = .5 a2 = -.5 bl = -2.0 b2 = .75 and the variables are defined as follows: Pf Prices paid by farmers index deflated by the CPI2 I Deflated disposable income per caput POP Total US population The variance-covariance matrix for these variables calculated in terms of percentage changes over the period 1950-1983 is as follows: Pf I pep Pf 5.15 .31 —.13 I .31 4.75 .02 pop -.13 .02 .10 For these variables, the variance figures on the diagonal represent the measures of instability according to 2 The Pf index is an aggregate one and therefore underestimates the input price instability faced by a producer of any individual commodity. 188 the INS method described in Chapter 4. Hence if these variances and covariances, and the values for the a and b coefficients, are substituted into the previous equations describing the instability of prices and quantities, then the contribution of these factors can be estimated. When this is done the expected instability of quantities and prices measured by the method are .5 and 2.2. These can be compared with the average quantity and price instability measures for commodity markets of 405 and 390 respectively. Thus only a small proportion of annual instability in prices and quantities (in aggregate) can be attributed to fluctuations in real income and population and fluctuations in the aggregate level of real input prices. They are not likely, therefore, to be important sources of instability fOr commodity markets, and the source of agricultural market instability must be sought elsewhere. The next section will consider various physical and institutional factors as possible sources for commodity market instability. 109 6.5 INSTITUTIONAL FACTORS In this section the influence of various physical, economic and institutional factors on market instability are investigated. In the previous chapter it was shown that there were wide differences in instability among commodities. Here I will attempt to explain some of these inter-commodity differences. To do this a cross-sectional regression analysis is conducted with measures of instability regressed against various factors. This analysis also allows the examination of various hypotheses about the sources of instability. Before presenting the analysis some methodological points ought to be discussed. Clearly price, quantity, and area instability are related to one another. I have presented evidence in earlier sections of this chapter to empirically support the theoretical assumptions that each affects the other. Hence there is some simultaneity of these variables, and if they are to appear in the same regression equation then a simultaneous estimation technique is called for. However, for some of the independent variables examined in this study it is not immediately clear whether they most directly influence quantity, area or prices. It is difficult, if not arbitrary, to build an appropriate structural model. Note that this is not a familiar demand- supply model but rather one where price and quantity instability are linked not only by supply and demand factors 41.11, 118 but by other factors as well. It can also be argued that the purpose of this analysis is primarily exploratory rather than an attempt to build a structural model. For these reasons the (initial) discussion will concentrate on the estimation and interpretation of reduced form equations; rather than attempt a structural model that takes account of the interactions of the endogenous variables. The estimated equations will therefore have relatively low explanatory power. Moreover all of the independent variables (listed below) will be included in the equations. Exclusion of non- significant variables would increase the explanatory power of these equations as measured by the diagnostic tests. However, such a procedure would invalidate the supposed significance levels of these tests. For this reason and because the purpose of the analysis is primarily exploratory, the procedure described above will be followed. In later analysis the equations will be reestimated in a simultaneous model to investigate whether any additional information is provided by this formulation, but with the caveats given above. The dependent variables in these regressions are the instability measures for area, quantity and real prices. A composite market instability measure was also tried but did not appear to add anything to the analysis that was not observable from the individual components and therefore it is not presented. As before the square root of the INS measure is chosen. Not only does the square root measure 11: provide a closer data fit, it also has better distributional properties (see Chapter 5). A fuller description of the variables is given in Appendix B, but for ease of interpretation they are briefly described here. The dependent variables for these regressions are: SDDQ quantity instability SDDPD price instability SDDA area instability The independent variables are as follows: MDQ average rate of quantity growth MDPD average rate of growth in prices GVP gross value of production PROC percentage of production which is processed M dummy for an import commodity X dummy for an export commodity ANN dummy if commodity is annual rather than perennial DS dummy for commodities with government price supports DF dummy for markets with futures markets DMO dummy for markets with federal marketing orders DVOL dummy for markets with marketing orders that include volume management provisions DMF dummy for markets with marketing orders that include market flow provisions These regressions are presented in Tables 6.4 to 6.6. In each table models are presented with and without the processing variable for which there is a smaller sample (72). There is also a smaller sample (64) for the area instability. 1'12 Table 6.4 SOURCES 0? QUANTITY INSTABILITY: REDUCED FORMS Independep; . ' Model 1 ' 1 Model 2 Variables- COefTTCTent t-VaTGe COeffTCTent t-Value cons 17.20 7.35 ‘ 14.07 5.15 MDQ .64 2.38 .80 2.74 MDPD -1.45 _ -1.95 -.78 -.70 GVP -551 -1.87 -4082 -1.15 H 3.01 1.03 3.64 1.12 X 7.01 1.89 6.57 .96 0M0 4.40 1.34 5.03 1.49 DVOL -2.03 -.51 -5.74 -1.34 DMF ~3.31 -.86 -1.70 -.47 0F -5.10 -1.52 -6.40 -1.04 05 -.72 -.19 —4.35 -.65 ANN 6.86 -2.77 -$.49 -2.l4 PROC .72 1.94 R2, F3/ .24 4.0 .33 3.9 N§/ 104 72 l/A fuller description of the variables is given in Appendix B. 2 g/Measures of fit are the corrected R _value and the F-statistic. 3/N is the number of observations. 113 Table 6.5 SOURCES OF AREA INSTABILITY: REDUCED FORMS Independeg} “Model 1 'Model 2 _ Model 3 Variables— CoeffTCTent t-Value Coefficient t-Value CoeffTCTent t-VaTue CONS 7.69 3.69 1 8.07 3.39 8.73 8.05 M00 1.18 5.56 1.39 6.03 1.03 4.91 MDPD -1.46 -2.07 -I.19 -I.21 -l.84 -2.96 GVP -740 -2.21 -758 -.23 M -3.42 -.98 -3.10 -.77 X 1.71 .55 -.56 -.08 DMO -.31 -.09 -.27 -.07 DVOL 1.68 .36 -.68 -.14 DMF o6.43 -l.40 -4.44 -.91 -5.66 -l.47 DF .01 .00 -1.10 -.17 -4.60 -1.89 US .30 .10 -1.12 -.17 ANN 1.83 .78 .02 .01 PROC .23 .74 R2, _2_/ .38 4.5 .47 4.5 .35 9.7 My .54 so 65 'l/A fuller description of the variables is given in Appendix B. 2 -§/Measures of fit are the corrected R value and the F-statistic. 2IN is the number of observations. ll4 Table 6.6 SOURCES OF PRICE INSTABILITY: REDUCED FORMS Independeg; 2.- Model 1 - . Model 2 Variables— CoeTTTCTent t-VaTue COeffTCTent t-Value CONS 17.25 9.24 18.83 7.38 MDQ .15 .69 .16 .59 MDPD -.37 -.62 -l.25 -l.20 GVP -601 -2.56 -2186 -.66 M 1.82 .77 3.00 .98 X 6.43 2.17 5.01 .78 DMD 3.68 1.40 1.63 .52 DVOL -.76 -.24 1.87 .47 DMF 5.73 1.86 5.33 1.59 DF 6.59 2.45 13.94 2.42 05 ‘ -2.00 -.57 -5.58 -1.os ANN -6.04 -3.05 -6.52 -2.71 PROC -.22 -.64 R2, 3/ .29 4.8 .31 3.5 N3/ 104 72 l/A fuller description of the variables is given in Appendix B. 2 g-/Measures of fit are the corrected R value and the F-statistic. 2/N is the number of observations. 115 It can be seen from these tables that the major influences on market instability are: annual versus perennial production processing versus fresh production price supports gross value of production futures markets market flow provisions of marketing orders long run growth or decline in prices and quantities trade (DVONUIuwai-i I will discuss each in turn. 6.5.1 ANNUAL VERSUS PERENNIAL PRODUCTION It is not surprising that this variable helps to explain market instability. Perennial crops and products of larger livestock are characterized by decisions which are difficult to reverse except over long time periods. Consequently adjustment to changing market conditions and to mistakes takes a long time. For example the high fixed costs of investment in the planting of fruit trees results in fruit production for a long period to come at low variable costs. This form of asset fixity is more likely to be a problem for perennial crops than annual crops. The regressions indicate that perennial commodities have about 53% greater price instability and 67% greater quantity instability than annual crops, when allowance is made for other variables. This production characteristic is clearly a very important source of market instabililty. 116 6.5.2 PROCESSING This parameter is also a proxy for contractual institutional arrangements as the processing industries frequently rely on contractual exchange mechanisms to ensure reliable supply. The available data on the extent of contracting (e.g., Mighell and Jones, 1963; Mighell and Hoofnagle, 1972: Lang, 1977) have insufficient commodity coverage for this analysis. These regressions indicate that commodities which are processed have a greater degree of quantity instability but possibly less price instability than commodities which predominantly go to fresh markets. The reduced price instability may not be surprising given that most contracts are price contracts rather than being quantity contracts (McLaughlin, 1983). Moreover, it is possible that the data do not always reflect the full extent of the price instability experienced by producers where the contracting arrangements are through producer owned cooperatives. In these cases, part of the producers returns from a crop may be in the form of a dividend payment (Staatz, 1984). The dividend portion of the total price received may be more variable than the nominal price paid on or near delivery. Hence there may be instances when there is some under-reporting of the instability of prices. The apparent price stabilizing role of contracting arrangements has not apparently led to greater stability of quantities. It is possible that processing firms which 117 process relatively unstable commodities have sought out contracting arrangements to reduce their instability but that these commodities remain relatively unstable, though perhaps less stable than otherwise, but the data can not show this. It might be noted that tart cherries is one of the most unstable markets and practically all production is processed. It should be noted that the greater quantity instability can not be attributed purely to the fact that perennial crops tend to be produced under contract: for this factor has already been accounted for by the ANN variable described above. 6.5.3 PRICE SUPPORTS This study provides some evidence (though not significant) to suggest that government price support programs have been associated with more stable prices to producers but little different production stablility than other comparable commodities. It is possible that this is because of the nature of the production characteristics of supported commodities, but there is quite wide variety in this respect, for example, between tobacco, corn and milk. The fact that price is more stable for these commodities and production stability is little different from other commodities, suggests that these institutional arrangements may be relatively successful in their price stablization objectives but that this is not being transferred through to 118 improving coordination with resulting reductions in area or production instability. This aggregated study can not give firm conclusions in this respect but these results are suggestive of useful lines of inquiry concerning the differential impact of price supports on price and quantity instability. 6.5.4 GROSS VALUE OF PRODUCTION It might be expected that the industries with greater gross value would be able to use the ‘voice’ of the political process to obtain government programs that are effective in reducing instability (Hirshman, 1970). Whilst price supports and marketing orders are explicitly considered in this study, government participation in the food system is not limited to these institutions. This study is supportive of this hypothesis. Commodities with high gross value of production have both lower quantity and price instability (taking into account other variables) than commodities which are less economically important. However there are alternative hypotheses that could be advanced to explain the relationship between GVP and stability. For example industries with high value probably have a lower cost per unit for the acquistion of information which aids in coordination. The optimum amount of information to gather under such conditions will be greater and less costly per unit for high value industries than other industries. This reason is related to the first in that a large part of 119 government provided market information and statistical services are concentrated on the higher valued industries. The cut back in statistical collection for many ‘minor’ industries in the early 19803, is evidence of this relationship. 6.5.5 FUTURES MARKETS Tomek and Robinson (1981, p266) ask the question "does trading in futures contracts increase the magnitude of the variance of annual cash prices?" They suggest that "futures markets may, in some instances, help stabilize production by providing relatively stable forward prices that can be assured by hedging. In addition,... available evidence suggests futures prices tend to have smaller annual variances than cash prices. The influence of futures prices on annual variability of cash prices, if any, would seem to be in the direction of reducing them." (See also Cox, 1976; and Powers, 1970). The evidence presented here from the cross-sectional analysis is in agreement with Tomek and Robinson’s suggestion concerning production stabilization. Both areas harvested and production show either less, or at least no greater, instability for commodities which have established futures markets than those which do not. However the evidence of our study is that these markets exhibit greater price instability than other markets. There is, of course, a popular conception that the existence of futures markets has a destabilizing influence on prices, which is at 13m odds with the inference drawn by academic economists. What reasons can be advanced then to explain the apparently greater price instability in these markets? It is possible that futures markets have been developed for commodities that are less price stable than others. It seems likely that the increased use of futures markets for grains in the 1970s has been encouraged by an increase in price variability. The existence of a sufficient degree of price variability seems to be a prerequisite for an effectively operating futures market. Thus the fact that these institutions have evolved for particular industries with unstable prices is a possibility. There are of course other factors which are important for an effectively operating futures market. These include the technical feasibility of writing contract terms that are satisfactory to both buyers and sellers, and the market organization of the commodity. This argument, however, does not explain why production may be more stable in these commodities. Another possible explanation can be advanced that relies on the interaction of futures and storage markets. It is likely that the optimum amount of storage in a market for buyers and sellers is different when there is a futures market than when there is not. For example, grain buyers may be able to satisfy their precautionary needs for adequate supplies through participation in the futures market rather than holding their own stocks. In such a case the total amount of inventory may average lower where a futures market is in existence than otherwise. Lower average inventories make the cash market more susceptible to fluctuations in supply or demand. Thus spot prices could well be more unstable. A third possible explanation derives from the possible effect of the futures marketing institution on the demand curve of buyers. If buyers are able to lock in a price for themselves then they may be less responsive to changes in the spot price, i.e., their demand becomes more inelastic. This would explain both greater price instability and greater stability in the quantity demanded but not necessarily quantity produced. The greater spot price instability of commodities with futures markets is an important finding of this study and warrants further investigation. There is some empirical work in this area (e.g., Powers, 1970: and Cox, 1976) but the question is still an open one. 6.5.6 MARKETING ORDERS Marketing orders are government supervised marketing arrangements for certain fruit, vegetables and speciality crops, which have as their purpose the aim of fostering orderly marketing. Some 48 federal orders are currently in operation. The provisions of the orders differ among commodities and among geographical areas. However they all IE m authorize certain restrictions on the qualities and/or quantities of products that can be marketed. The restrictions vary among orders and may include packaging standards, minimum requirements for grade and size, limitations on quantities shipped during certain periods within the marketing season, limitations on quantities going to the fresh market and, in some cases limitations on total marketings. All but three of the current orders include quality standards. In addition, most orders include provisions which may be described as market support activities. These include standardization of containers, levies for research and sometimes for advertising. About half the marketing orders have various types of quantity controls. These represent the strongest form of regulation available from orders as they may be used to affect prices. These are of two types: volume management provisions and market flow regulations. The volume management provisions are of three types: producer allotments, market allocation provisions and reserve pools. The market flow provisions may be handler prorates or shipping holidays. The market flow provisions are aimed at distributing the seasons’ production over the crop year to avoid seasonal gluts and shortages. In principle all of the production is sold. On the other hand the volume management provisions attempt to increase price by reducing the quantity sold on the primary market. (See Heifner et al, 1981; Jesse, 1979; Jesse and Johnson, 1981). Ls] 13 The diversity of these provisons are summarized in this study with three dummy variables. The first for commodities with any type of order, second those with volume management provisions and third those with market flow regulations. As it is an explicit aim of these orders, it might be expected that commodities with marketing orders would experience less instability than other commodities. The evidence of this study does not support such a view. Industries with orders tend to have greater price and quantity instability, though not significantly so. This may be interpreted to imply that these orders are not being effective, or it could be that the relatively unstable industries are more likely to demand marketing order institutions than other industries. No unambigous statement can be made about the direction of causality. However it is clear that industries with marketing orders are no more stable than those without them. However those orders which have market flow provisions are somewhat different. These industries appear to exhibit greater price instability but less production and area instability than other industries with marketing orders. It is not immediately clear why this should be the case. It is reasonable to suppose that the market flow provisons permit greater intraseasonal price stability, which gives clearer market signals to producers about how much to plant. However 1 4 fl) this does not seem to be translated into greater annual price stability. 6.5.7 LONG RUN CHANGES IN PRICES AND QUANTITIES Hamm (1981) has suggested that certain commodities which exhibit marked instability will not survive among the constellation of available agricultural goods. He argues that marketing institutions are less able to cope with unstable commodities (or varieties) and that these will experience declining demand from marketing and processing institutions despite consumer level acceptance. Examples of such commodities might be apricots and asparagus. One empirical test of this hypothesis would be if there was a negative relationship between price and quantity instability and growth in productipn. In fact, the data do not support this hypothesis. Quantity instability is significantly and positively related to production growth. The relationship between price instability and production growth is also positive but it is not at all significant. A better test of this hypothesis is provided in the next chapter where the relationship between increased instability (rather than the level of instability) and production growth is investigated. The data do provide some evidence that growth industries and those with the greatest declines in real prices experience greater production instabiliity. It does seem intuitively likely that, in periods of growth or rapid technological change, decisions about optimum investment 135 strategies are likely to be more difficult to make. It is not surprising, in turn, that coordination is more difficult under these circumstances than under more stable conditions. 6.5.8 TRADE It is often argued that many international agricultural markets are residual markets which remain after other countries have insulated their own agricultural sectors. Hence the equilibrating and stabilizing role of the market is left to the residual, and sometimes thin, international market, which must absorb most of the instability that would otherwise be spread more evenly. Thus those commodities which enter international trade are more likely to be unstable. The data provide some support for this hypothesis. Internationally traded commodities do give evidence of being less stable than other commodities. This is especially true of prices of export commodities. The evidence is not so strong for import commodities. However a number of these are tropical fruits which are only grown in quite small quantities in the US (e.g., bananas in Hawaii), and where a substantial proportion of world production enters world trade. 6.5.9 SIMULTANEOUS MODEL In this section I will present some results from a simultaneous specification of a model of market instability. This may increase the understanding of the factors 136 influencing instability above that available from the reduced form equations discussed above. The proposed model incorporates some of the above discussion, but the choice of which independent variables to include in each equation is a little arbitrary. However, a choice needs to be made for this analysis. The proposed formulation is as follows: SDDA=f(SDDPD, ANN, GVP, MDQ, PROC) SDDQ=f(SDDA, SDDPD, ANN) SDDPD=f DVQL'14)=-DVQL(8) DWQL(1S)=—DVQL(7) DYQL<:5)=—DVQL(6) DVQL<17>=~DVQL(S) DVQL(18)=—DVRL(4) DVQL<19)=-DVQL(3) DVQL(E@)=—DVQL(E> DVQL(EI)=-DVQL(1) DX=.@5 flLX=10 CV=.:5 RHE=® SDX=CV§SOR<(1+RHO>/(1—RHD))*MUX X(3)=® QQNDOMIZE TIMER FOR K=1 TO 1®0 ELM=® SUMX=® SUMSG= 0 SLMSOX=® :0? I =1 TO 3% GOSUB 3003 J 36 189 545 546 550 573 380 59% 6&3 610 63% 53% 35 336 637 51.1.17) 641 543 643 ,— . .- LIN-+5 -: EH6 -..”- Iv. *4m ERQQ 301% 503% 3030 304% 325m 2068 190 SUMX=SUMX+X(I) SUMSQX=SUMSDX+(X(I))AE I: i=1 SOTO ewe M==SGR VK=VH+VRR(K) SC=SC+CVX NEX' H SMX=MX/1@m SSX=SX/1@® SCV=SC/1@® SSK=SK/1@@ SVK=VH/1@@ :RINT SMX SSX SCV SSH SVH CV RHO R=RND(1) A=1+ R/Dx 3=INT(Q) C(I)=DVQL(E)+(R-DX*(E-1))*(DVQL(B+1)-DVQL(B))/DX X(I)=(SDX*C(I)+MUX)+RHU*X(I—l) IF X(I><@ THEN X(I>=0 RETURN APPENDIX E A COMMENT ON MYERS AND RUNGE’S ARTICLE In a recent article, Myers and Runge describe a method to decompose instability in the US corn market among supply and demand components. They reach the surprising conclusion that demand factors are far more important than supply factors in explaining recent market instability. MR find that their results are quite robust under likely ranges for supply and demand elasticities. However, the results are not so robust on further examination. In particular, if the price series is deflated, which MR do not do, then the principal conclusions of the decomposition are radically reversed, and supply effects predominate. Table E.1 compares MR’s results under ranges of elasticity assumptions with those when the price series is deflated by the CPI. Only two of nine entries have demand effects predominant compared to nine of nine using MR’s nominal prices. 191 19 ['13 Table E.1 Ratio of Variance in the Demand Intercept to Variance in the Supply Intercept (DSR) for Corn under a range of Elasticity Assumptions: 1971-72 through 1982-83 SUPPLY DEMAND ELASTICITY ELASTICITY -.3 -.7 -1.1 NOMINAL DEFLATED NOMINAL DEFLATED NOMINAL DEFLATED .2 1.62 .37 3.03 .67 5.26 1.67 .4 1.45 .24 2.72 .43 4.73 1.07 .6 1.13 .16 2.12 .29 3.68 .73 An entry of 1.00 would indicate that the variances of the supply and demand intercepts were equal over the period under the elasticity assumptions shown. Initially it is not obvious why deflating the price series makes so much difference to the conclusions reached, and it may not be obvious whether it is better to deflate or not. To elucidate this point it is useful to examine the expression which generates the figures in Table 5.1. The authors use a static partial equilibrium model with linear supply and demand functions: Qtd = at + th (demand) Qts = ct + dPt (supply) Qtd = Qts (equilibrium) where Qtd and Qts are quantities demanded and supplied and Pt is the (farm) price received; at and ct are net supply and demand intercepts which incorporate exogenous demand and 193 supply shifters; b and d are constant slope parameters which are calculated from prior estimates of elasticities at the means: b = ed * mean(0)/mean(P) < 0 d = e5 * mean(0)/mean(P) > 0 then solving for at and ct and taking variances gives: var(a) = var(Q) + b2 ' var(P) - 2b * cov(P,0) var(c) = var(Q) + d2 i var(P) - 2d I cov(P,Q) and DSR = var(a)/var(c> Thus the value of DSR is dependent upon, among other parameters, the covariance of price and quantity. In the present case, production increases steadily throughout the period whilst inflation ensures a similar growth to the nominal price series. Consequentally the covariance is highly positive. To accommodate such a high covariance it is necessary to have significant shifts in the demand curve. However this shift is mainly in one direction (ie outward). It is difficult to attribute this shift to ‘instability’ in supply and demand. It would be more accurate to attribute it to trend factors. It is noteworthy that deflating the series leads to a negative value for the covariance term, with the consequent result that supply factors predominate over demand factors when ‘instability’ is decomposed. Thus this analysis shows that there is a danger of confusing trend factors with instability and that when they are confused then anomolous results are possible. At a minimum I" ‘ x 194 the analyst should deflate price series and, for relatively long periods such as the present case, then detrending of the series should also be considered. What do these considerations mean for the substance of MR’s analysis? It is useful to analyze price and quantity instability analogously to MR’s Table 3 but with deflated and detrended prices. This is done in Table E.2 below. It would appear now that a good case can be made for the argument that demand factors have been of increasing importance in explaining the increased instability in the corn market. However the case must now be made on the basis of an increase in the estimates of the farm level demand elasticity. Assumption of fixed elasticities is insufficient to produce this conclusion alone. This result is consistent with the work of Tweeten (1983) and other researchers who conclude that the growth in export demand has led both to increased instability and to an increase in the elasticity of demand. It is noteworthy that even with the higher elasticity estimate, quantity instability must be attributed mainly to supply effects. 195 Table E.2 Decomposition of corn price and quantity into supply, demand and interaction components. Elasticity assumption Supply .4 .4 .4 Demand ..03 -I3 -101 Time Period 1962-70 1971-82 1971-82 Decomposition x x x Variance of price 100 100 100 Supply effect 488 81 18 Demand effect 228 18 42 Interaction -616 1 41 Variance of quantity 100 100 100 Supply effect 28 72 212 Demand effect 24 29 66 Interaction 48 -1 -177 DSR .47 .22 2.35 APPENDIX F DERIVED DATA FOR THE ANALYSIS OF THE STUDY This appendix provides the data set for the analysis of the study. The data in this appendix have been calculated from published data (mainly from various issues of Agricultural Statistics) as described in the text. The following tables list the derived data, from which the tables are compiled and the regression analysis is made. The meaning of the variables codes that head each column are given in Appendix B; and the meaning of the commodity codes are found in Appendix A. 196 197 Table F.1 DERIVED DATA FOR ANALYSIS OF THE STUDY COHMOD TYPE VARDPD VARDQ VARDY VARDA COVZAV CVTPD CVTP CV70 CASE-N0 Bt67L84344543552317638056a4857‘6‘7500794758797937‘83‘0013663226007913353441I‘7B‘3237333I2550722704IWA°55597 I58256965238920856“”56‘8460I48697823833652‘20520377028‘OI397I1373‘1625296247I7629478°6048279064887° 0555267 13 112 1111 212 111 2111121222144111 12312125323231 111211 112311 111 211111311 111 ....................... . ................ ..0.....0... . ........0........ ...................... 331641503423W92474282476315957901M255526 92816622.17329.625255069968 1811753.9699589050576727111636516485865 862623273817 11439119296156589191 560558 01001848 46236 713189984218 3968850 6441812381792135396606265497608 123122232232222221322311232111 124222321633434144 33333 43322231 33233222123 22353343222233222122221 231 ............................ .. .. . . ......0.....O. ...... ........O...... ........................ 7 5655576I765 567382I295596687‘9‘0IBB768‘I‘058 7305 4. 6925736296 3724I6‘2‘.345534I76345661223961 ‘678I97I2 3 328594I6939527219878729557583‘756‘2II51876 5952 9 I76538273‘ I2‘253289 9845526326088857I26I5 562672714 IQII III I I III 2I I I 222222I‘22233 22 2 332I2II 222222I2I I II2‘25322 IIzzIII2I2I2 32 166 286 265 260 ................................. 0000 t .00 . . u I .0 .0000 . u u . . 4 . . sooooooooo . 4 . . . . . . . .oooooooo . .oooooo . . 93 .............24..74081.Iw....358846437.........3649979772........23......80 9 4‘ I I2235.03‘.. 6. I.‘ 2 5 53.1455.I 31 8042 I 6 29 .. . .... .I .. .. .... 1.. . 2 14 I . I. . 2. . 13 . 3 . . . .............................. 0000000000000..00.....O..OOOO.........000000000..........00000000..000000.. 112277793437202 19142385922359917.............95..99387.15....935506941.........0548480696........82......19 5652I9HQI°23515I 253‘55I2I308I35I2 5 23693 Ia 69178432‘ 6444276 91 ‘8 66 16 12 11526 Sn... 1 51 11531 2 2 3 ................... ............ .00 . 0..OOOO.........000000000..........OOOOOOOO..OOOOOO.. 9540777849277500 1 SE .1.............17..33947.92....505926218..._.....3930463801........04......95 9053552°I55‘3080I53552I3I“5306"2 73 39555 ‘9 68 497980 ‘38202317 66 5° 1 1 31 1 115 2% 22 1 1 01 1 21 12 239839536065700585351I89092753°I00079552246966884992789735933‘0325599 2 373443‘230 04061139055669438285124835 5639209507506769I639‘5I2I8I°2564455263304I675763237365959755333010175 3‘ 3 557727 31352331957809927873531023 I7 23II25 I III 2 21‘ I 6I58336‘952u9 I3 1294119739N3I4 22‘821 3 2256II 9‘6588‘I ‘31 1 1 111 .............................................. ..0. .0....... ............0............................... 58628004895W7176730828822460354199111168778W85612.12901.69889824373187972152.582 286644230489154035886754531 05537535766 0372917493689226338550867361265 95763 06117 19647033321343237222 534 528782047706862272671361298 2231 1 2 711 224 131 7675433496561 33 13464 368315 222232112 1 324947823 128544 67564 : 444 .... I PRTULFPTTCSBPHACETPRPOFFTNPPPFPPFHFMRFCST"LGVOUBTNPPRSXNPNMLCLCNSYTBGWNFRGKMHKKLCBEEECBVTLGPRHSMNMPNSPMRSPBF GAARMPPRESACCEEBCCUCAPC0LOCBOUGUBIPEGREHHEGILVTRAIWEPTLWOULIEAAEUBOSS CBBEMLCPTVGGBPVIOAOURFPCRLROLAVVVAOABL EG'BLSSABE DP 5 T ST SSTC CSLGLTGNCCTTFOASCDP PSCF PSAFP'HP" 0 $50 DRRTHCUDAAPPPPPPBCCCCRN A 12345678901234567390123456739 901236567890123456789 1234557390123456789012345673901234567890123455789m12345m70 1111111111222222222 33333333334644.46664 5555555556566666556777777777733885838889999999999 00000 00 111111111 198 Table F.1 (cont'd) VARDPD1 VAR001 VARDPD2 VARDO2 VARDPD3 VARDO3 GVP MEANDO HEANDPD cmumo CASE-N0 .1 1..2.1 11 .1...1 ....1..31..1...23...2111312211...12..42. .22211 1..22..122142..2. .31 12123..34....1 21 243441 1 . . . . . . . .. . .. ......... . .. ..... . .. ...... . ..... .. . . ..... 311675345028551390948972455409971114141590549554439334773510408677537253925583428334131253711352749996781280 .................................................................................................... .............................................. lo. I ...-0.44 ...-......-IOOQOI-O~o~ao...-....-64.4.04...- 0406595406251315011610925740534134 18741437481994.46179070373244953547744832.39.1581209683474852240168824620 13512212118764047639 2436411782081 633494 2132881 14275251714825632779850680 76 94834406790536771 1 26348164 11111 71553113 11 31 3 5 1 44 11 5308873643284 82 676 7710 37 31 1 1 31915436 81 1284 1 I 22 I IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII O‘OOIOIGIOIOQOIIIIIOII04040-I 161415058982259550127694555218184 4417163133765489754410570 864150.79622967562895.7417996803838617030470678469 776917643182017917H555 3 8702344 275 866411660269549244417122661 03721631 6303 5203436572262518792294221154 516 54 111 9 514454351 2 287625 239316 333511 1 226621 7 2 01153352 531 I I 52 I 2112 3 I I .......................................... ouch-03.4400.-104.-no.0...-uvnn°oo°ounonlopuc...-.44-4-Ooouuuu. 23282018964707766938373240502689810020MW98W29463.50943.06974515.27054308296.47.10W1356335859719237651945974 72267464063680165652425431510356076694 53 71583 23239 70215804 73666393024 01 52 9773683920841237692481348 1 211 213 4 2 316 9 353121 5223463244968 11 16622 6963081 555485133 1 55 243450 144354 41683 67 4 412 II 4141 21124! 4- 4| I ......................................... a o... coo-...uno cu...4o-.....Oagono.........auuouoooooo0° 1864053542906828089342735871535681053570785445297.7207124.093%08.4566 506337275.18837019166999147.4......5.. 633418782224564615353821 679676311216709282672798 8936W05 488 34 5957 911232 8401499 0 8770019 2 8 2 112 211 1 2 1 3 1 4242228296300 1 1223 21 47322 2111 1 1225 4 5 42175 5 4 4| 4| 2 ‘2 4I4l4l ‘ 4' ............................. o u. 0 0:0 40:00 a no :ocoonuuqu-o 4.4 u..-u--o°.°ooooonoo 74923084 01940852975481954481229719608W99981041350.11565.6.178679.61909218436.51.14813002339066647.9......4.. 761663 460571023061314480921528522427 7585493 477 36839 2 62673 63587268517 54 35 900223256991639 9 1 311 3 312 322 8 26 11 426032249910 22 123 1 252121 2 111 249251244 5 I4! 21 I: I I ........................................ OO......0.......O......O... .....O. .0. . .000 .000000 .00 538103987364874363537428268124984 29604990..70290W.8973855.042807. 25735252334591.95.222819542009...1. .6.. 318323326537467428853322220446748 08489482 45765 7062875 275443 0232u585 9876 80 637233 72724 7 9 26172 121 2 113 41 1 1417 2 61121648 100 26 2283114 581416 3 10 4711 3 16 4 2 I ‘2 {IQ-2341 2 I I uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu OOo-ao-oo-ooouoOct-c.604-uo-vco-aco-00.04-44.4-.....ooo.oooooo.oo 02801754786676902047566258731936176218920..710130.30401.8. 83600.86017994172.97.75.049946646032...2......6.. 50142444206153 33723625179626042056711902 804502 56117 8 23815 69351280535 58 62 369134997749 4 2 1682 1 2 411 226 2911 fl 21 125335148 489176 1343 1 503 2 112 1 3 1 31 1 7344 3 3 I II PR1UAFPTTCSBPVACETPRROFFTNPPPFPPFNFHRPCSMMLGVOVBTNPPRSXNPWMLCLCNSYTBGWWFRGWHHKKLCQEEECBYTLGPRHSINHPNSPIRSPBF GAARMPPRESACCEEBCCHCAPCOLOCBOUGUBIPEGREH EGILVTRAI EPTLWO LIEAAEUBOSS BBE LCPTVG BPYIOAOWRPPCRLROAAYYYAOMBL EGHBLSSABE DP S T ST 5570 CSLGLTGNCCTTFOASCOP PSCF PSAFPHHPN 0 SSDDRRTHC OAAPPPPPPBCCCCR A 1234567890123456789012345678901234567890123456789 1234567890123456789012345678901234567890123456789W123N5W7N 1111111111222222222233333333334444444444 5555555556666666666777777777788888888889999999999 000 0 O IIIIIIIII 199 (cont'd) Table F.1 DFUTURE DSUPPORT DVOLNAN ONKYFLOU COMNDO ANNUAL PROCESS DIHPORT OEXPORT DMORDER CASE-NO 0000000000000000000000000000000000000000000000000000000 L0000000L0LL0LLV.0001.00000L L00LLL0LL000000000000000000 000000000000000000000 L00000000000000000000000000000000000000000000LLOLLLLL000LLLLL000000LLL00000000000000000 0000000000000000.0000000L1.0000000.... 0001 L00L00000000000000000000000900000000000000000000.00000000000000000 1 11111 cl “‘1 00000000000000001.000000000000000000001.001.000000. .0 .0 000001 L0L000000000O0000000000000000000000L0000000000000 00000000000001 .001 .0000 .0. 001 .0000 .......... 0. .0 .0 .0 .00000L L0 L0000000000000000000000000000 ..... 01 .00000000000 ‘ 1‘1 ‘1““l“" 11 ‘0. ‘ 1 ‘CIIICICI 000000000000000000000000000000000000000000000000000001.0000000 .0 .0 .0 .. .00000000000L.0.L0L00000.0000000000000 ‘ .I ‘ ‘0'“ 1" 1 000000000000000000000001 .0000L 00001 .000000000. . .0. 000001 001 L0000.000000.00..0.00.L0000000L000000..0000000.00 ‘11‘ .0... 1 .I " 1 ‘l .. ... .. .. ..... . ..... 0 00000 000000000000000. .0000 ...... 00089091906100930413060001999009046548059569813895999991.9.....95.... ..99....9W09841560901.00000007 .0 ....... 0. .0......................00000000000000000000.....00000........0. .001.0 L00........00000000000.....000 ‘IIIOIQICICICI ‘1' ‘l‘l‘dl‘l‘il‘Cl‘Cl‘l‘l‘l‘lll‘l‘Ql‘l‘l‘ 1‘1‘1 ‘l‘1I‘CI‘l1CI (ICI "I‘ICICICICICI ‘1‘CICI pRT APPTTCSBPUACETPRROFFTNPPPFPPFHFHRPCST“LGVO'BTNPPRSXNPNHLCLCNSYTBG NFRG "HKKL BEEECBYTLGPRHSHNHPNSPNRSPBF GAAWMPPRESACCEEBCCUCAPCOLOCBOUGUBXPEGREHHEGILVTRLIWEPTLWOULIE‘AEWBOSSWC885mLCPTVRGBPYIOAOORPPCRLROL‘YYY‘OABL EGUBLSSABE OP 5 T ST 5570 CSLGLTGNCCTTFO‘SCDP PSCF PSAFPUIP O SSODRRTHCUO“9PPPPPBCCCCRU ‘ 123456789012345578901234567390123‘56789 111111111122222222223333333333 75 76 78901234567890123456789 12345“?! 77788888888889999999999 00000 00 1111 ““‘ BIBLIOGRAPHY W .Adams, F. Gerard and Sonia A. Klein (1978), eds Stabilizing World Commodity Markets: AnalysisI Practice and Policy, Lexington Books, D.C. Heath and Co., Lexington, Mass. Agricultural Statistics (various issues), USDA. Allais, M. (1953), "Le Comportement de l’Homme Rationel devant 1e Risque, Critique des Postulates et Axiomes de l’Ecoles Americaine," Econometrica, 21 p502-546. Anderson, J.R., P.B.R. Hazell and P.L. Scandizzo (1977), "Considerations in Designing Stabilization Schemes," American Journal of Agricultural Economics, 59(5) Dec p980— 991. Antle, John M. (1983), "Incorporating Risk in Production Analysis," American Journal of Agricultural Economics, 65(5) Dec p1099-1106. Arrow, K. J. (1971), Essays in the Theory of Risk Bearing. Markham Publishing Co., Chicago. Bain, Joe S. (1959), Indusrial 0r anization, John Wiley and Sons, New York. Bieri, Jurg and Andrew Schmitz (1974), "Market Intermediaries and Price Instability: Some Welfare Implications," American Journal of A ricultural Economics, 56(2) May p280-285. Biswanger, H. P. (1979), ‘Risk and Uncertainty in Agricultural Development: An Overview’ in J. A. Roumassett, J. Boussard, and I. Singh (eds), Risk Uncertainty and Agricultural Development, New York: Agricultural Development Council, 1979, p383-398. Blandford, David and Susan E. Offutt (1981), An Evaluation of Alternative Indidcators of Commodity Instability, Staff Paper 81-19, Cornell Department of Agricultural Economics, Ithaca, New York. 200 801 Blandford, David and Nancy Schwartz (1982), World Wheat Price Variability: Changes in its Magnitude and Source, Staff Paper 82-33, Cornell Department of Agricultural Economics, Ithaca, New York. Blandford, David and Nancy Schwartz (1983), "Is the Variability of World Wheat Prices Increasing?" Food Polic , vol 8, P305-312. Bohrnstedt, G.W. and A.S. Goldberger (1969), "On the Exact Covariance of Products of Random Variables,” Journal of American Statistical Association, 64 p1439-42. Bond, Gary, Chris Vlastuin, Peter Crowley and Peter Demura (1984), “Why World Commodity Markets will Remain Unstable," Quarterly Review of the Rural Economy, 6(2) May p172-177. Bosworth, Barry P. and Robert 2. Lawrence (1982), Commodity Prices and the New Inflation, Brookings Institution, Washington DC. Burmeister, Edwin (1978), Is Price Stabilization Theoretically Desireable? in F. Gerard Adams and Sonia A. Klein (eds) Stabilizing World Commodit Markets, Lexington Books, D.C. Heath and Co. 1978 Burt, Oscar R. and R. M. Finley (1968) "Statistical Analysis of Identities in Random Variables.“ American Journal of Agricultural Economics, 50 p734-744. Butler, L. J. (1979), Some Aspects of Agricultural Price Instabilit , Ph.D. Dissertation, Michigan State University, East Lansing. Chambers, Robert G. (1983), “Risk in Agricultural Production: Discussion,” American Journal of Agricultural Economics, 65(5) Dec p1114-1115. Chiang, Alpha C. (1974), Fundamentals of Mathematical Economics, McGraw Hill Book Co., New York. Cochrane, Willard W. (1953), "Professor Schultz Discovers the Weather,“ Journgl of Farm Economics, 35 p280-83. Coppock, J.D. (1962), International Economic Instability, McGraw-Hill, New York. Cox, Charles C. (1976), "Futures Trading and Market Information," Journal of Political Econom , 84 p1215-1237. Currie, J.M., J.A. Murphy and A. Schmitz (1971),“The Concept of Economic Surplus and Its Use in Economic Analysis,” Economic Journal, vol 81, Dec. 30 m Cyert, R.M. and J.G. March (1963), A Behavioral Thgpry of the Firm, Prentice-Hall, Englewood Cliffs, New Jersey. Debreu, G. (1959), Theory of Value, Wiley and Sons, New York. Economic Report to the President 1984, United States Government Printing Office, Washington. Economic Research Service, USDA (1982), Instability of Farm Income: Report of an Interagency Task Force. Economic Research Service, USDA (1984), Background for 1985 Farm Legislation: (various commodities), Agriculture Information Bulletins. Edwards, Clark (1984), "Wheat Price: Past and Future Levels and Volatility," Agricultural Economic Research, 36(3) p28- 31. Ellsberg, Daniel (1981), ”Risk, Ambiguity and the Savage Axioms,” Qpartgrly Journal of Economics, p643-664. Fama, Eugene (1970), “Efficient Capital Markets: A Review of Theory and Empirical Work," Journal of Finance, XXV(2) May p383-417. Ferris, John (1984), "Implications of Crop Yield Variability to the Supply and Price Outlook for Feed Grains, Wheat and Soybeans in 1984 and 1985," MSU A ricultural Model, Spring, Firch, Robert S. (1964), "Stability of Farm Income in a Stabilizing Economy," Journal of Farm Economics, vol 46(2), May, p323-340. Firch, Robert S. (1977), "Sources of Commodity Market Instability in US Agriculture," American Journal of Agricultural Economics, 59(1) p162-169. Fleisher, Beverly (1985), A New Measure of Attitude Towards Risk, Ph.D. Dissertation, Michigan State University, East Lansing. Frankel, Jeffery A. (1984), "Commodity Prices and Money: Lessons from International Finance," American Journal of Agricultural Economics, 66(5) Dec p560-566. French, Ben C., Niniv Tamimi and Carole Frank Nuckton (1978), Marketing Order Program Alternatives: Use and Importance in California. 1949-75, University of California. Galbraith, J.K. (1967), The New Industrial State, Houghton Mifflin, Boston. Gadson, Kenneth E., J. Michael Price, Larry E. Salathe (1982), Food and Agricutural Policy Simulator (FAPSIM): Structural Eguations and Variable Definitions, Staff Report, National Economic Division, Economic Research Service, USDA. Gardner, B. Delworth (1977), "Discussion: Sources of Commodity Market Instability," American Journal of Agricultural Economics, 59 p188-191. Gardner, Bruce (1977), "Instability in US Agriculture- Discussion," American Journal of A ricultural Economics, 59 p185-187. George, P.S. and G.A. King (1971), Consumer Demand for Food Commodities in the United States with Pro ections for 1980, Giannini Foundation Monograph No.26, March. Goldberger, Arthur S. (1970), "On the Statistical Analysis of Identities: Comment," American Journal of Agricultupal Economics 52 p154-155. Goodman, Leo A. (1960), "On the Exact Variance of Products“ Journal of American Statistical Association, 55 p708-713. Halloran, John M. (1983), Ph.D. Dissertation, Michigan State University, East Lansing. Hamm, Larry (1981), Food Distributor Erocurgment Practices: Their Implications fgr Foog System Structure and Coordination, Ph.D. Dissertation, Michigan State University, East Lansing. Hammig, Michael D. and Ron C. Mittelhammer (1980), "An Imperfectly Competitive Market Model of the U.S. Lettuce Industry", Western Journal of A ricultural Economics, 5(1) July p1-12. Hannan, E.J. (1960), Time Series Analysis, Methuen and Co. Harrington, David and Clark Edwards (1984a), United States Agricultural Capacity and World Food Balances, ERS, USDA. Harrington, David and Clark Edwards (1984b), “The Future Productive Capacity of U.S. Agriculture: Economic, Technological, Resource, and Institutional Determinants," American Journal of Agricultural Economics, 66(5) Dec p585- 591. Hartman, David G. (1974), "The Egg Cycle and the Ability of Recursive Models to Explain It,” American Journal of Agricultural Economics, 56(2) May p254-262. 304 Hazell, Peter B.R. (1982), Instability in Indign Fgodgrain Production, Research Report No. 30. International Food Policy Research Institute, Washington D.C. Hazell, Peter B.R. (1984), ”Instability in Indian and United States Cereal Production," American Journal of Agricultural Economics, 66(3) Aug p302-311. Heady, Earl O. (1952), Economics of Agricultural Production and Resource Use, Prentice Hall, New Jersey. Helmberger, Peter, Gerald Campbell and William Dobson (1981), "Organization and Performance of Agricultural Markets," in A Surve of A ricultural Economic Literature, ed. by Lee Martin, University of Minnesota Press, Minneapolis, vol 3 p503-653. Heifner, R. (1981), A Review of Federal Marketing Orders for FruitsI VegetablesI and Specialty Crops: Economic Efficiency and Welfare Im lications, Agricultural Economic Report No.417, Agricultural Marketing Service, USDA. Heiner, R. A. (1983), "The Origin of Predictable Behavior," American Economic Review, 73(7) Sep p560-595. Hirchman, Albert 0. (1970), Exit Voice and Lo alt , Harvard University Press, Cambridge. Jesse, Edward V. (1979), Social Welfare Implications of Federal Marketing Orders for Fruit and Vegetables, Technical Bulletin No.1608, Economics, Statistics and Cooperatives Service, USDA. Jesse, Edward V. and Aaron C. Johnson (1981), Effggtiveness of Federal Masrketing Orders for Fruits and Ve stables, Agricultural Economic Report No.471, Economics and Statistics Service, USDA. Johnson, D. Gale (1947), Forward Prices for Agriculture, University of Chicago Press, Chicago, Illinois. Johnson, D. Gale (1975), "World Agriculture, Commodity Policy, and Price Variability," American Journal 9; Agricultural Economics, 57 p823-88. Johnson, D. Gale (1979), ”Policies for Achieving Farm Income Stability," paper presented at the 17th annual meeting of the Brazilian Agricultural Economics Association, at Brasilia, August 1. Johnson, Glenn L. and C. Leroy Ouance (1972), The Over- production Trgp in U.S. Agriculture, Resources for the Future, Inc., Johns Hopkins University Press, Baltimore. re 05 Just, Richard E. (1975), "Risk Aversion under Profit Maximization," American Journal of A ricultural Economics, 57(1) Feb p347-352. Kauffman, Daniel E. (1983), A Pork Contract Market: An Investigation into Attitudes about gnd Possibilities of such a Market for Slau hter Ho s, Ph.D. Dissertation, Michigan State University, East Lansing. Knight, Frank (1921), Risk Uncertaint and Profit, Houghton Mifflen Co. New York. Lang, Mahlon G. (1977), Collective Bargaining Between Producers and Handlers of Fruit gpd Vegetables for Processing: Setting LawsI Alternative Rules and Selected Conseguences, Ph.D. Dissertation, Michigan State University, East Lansing. Leamer, Edward E. (1983), "Let’s Take the Con out of Econometrics," American Economic Review, vol 37. Leibenstein, Harvey (1979), "A Branch of Economics is Missing: Micro-Micro Theory," Journal of Economic Literature, June, p477-502. Lin, William (1977) "Measuring Aggregate Supply Response under Instability,“ American Journgl of Agricultural Economics, 59(5) Dec p903-907. Lipsey, R.G. and K. Lancaster, (1956), "The General Theory of Second Best,” Review of Economic Studies, No. 63 p11-32. Lowell, Michael (1960), "Manufactures’ Inventories, Sales Expectations, and the Acceleration Principle," Econometrica, 29 p293-314. Luce, R.D. and H. Raiffa (1957), Games and Decisions, J. Wiley and Sons, New York. McCloskey, Donald N. (1983), "The Rhetoric of Economics," Journal of Economic Literature, vol 21, p481-517. Machina, M. (1983), "The Economic Theory of Individual Behavior Toward Risk: Theory, Evidence and New Directions." Technical Report 433, Institute for Mathematical Studies in the Social Sciences, Stanford University, Oct. McLaughlin, Edward W. (1983), Buying and Selling Practices in the Fruit and Vegetable Industry: Implications for Vertical Coordination, Ph.D. Dissertation, Michigan State University, East Lansing. awe, Manetsch, Thomas J. and Gerald L. Park (1984), System Analysis and Simulation with Applications to Economic and Social Systems (Parts I and II), Department of Electrical Engineering and System Science, Michigan State University, East Lansing. Mangum, Fred A. (1983), "Trade Policy Issues and Analysis," mimeograph paper presented to Michigan Farm Bureau, Lansing, Jan 25. Massell, B.F. (1969), "Price Stabilization and Welfare," Quarterly Journal of Economics, 38, May p285-297. Massell, B.F. (1970), "Some Welfare Implications of International Price Stabilization." Journal of Political Economy. 78 p404-417. Mehra, Shakuntla (1981), Instability in Indian Agriculture in the Context of the New Technology, Research Report No. 25. International Food Policy Research Institute, Washington D.C. Mighell, Ronald L. and William S. Hoofnagle (1972), Contract Production and Vertical Integration in FarmingI 1960 and 1970, ERS-479, Economic Research Service, USDA. Mighell, Ronald L. and L.A. Jones (1963), Vertical Coordination in Agriculture, Agricultural Economics Report No.19, USDA. Miller, Thomas A, Jerry A. Staples, Robert M. House and Charles V. Moore (1984), Implications of Increasing World Grain Market Fluctuations on US Agriculture, Agricultural Economic Report, Economic Research Service, USDA. Muth, John F. (1960) "Rational Expectations and the Theory of Price Movements", Econometrica, 29 p315-335. Myers, Robert J. and C. Ford Runge (1985) “The Relative Contribution of Supply and Demand to Instability in the U.S. Corn Market," North Central Journal of Agricultural Economics, 7(1) p70-77. Newbery, D.M.G. and J.E. Stiglitz (1981), The Theory of Commodity Price Stabilization: A Study in the Economics of Risk, Clarendon Press, Oxford. Newbery, D.M.G. and J.E. Stiglitz (1982), ”The Choice of Techniques and the Optimality of Market Equilibrium with Rational Expectations," Journal of Political Economy, 90(21), p223-246. Nerlove, Marc (1972), "Lags in Economic Behavior," Econometrica, 40 p221-251. Nerlove, Marc (1979), "The Dynamics of Supply: Retrospective and Prospect," American Journal of Agricultural Economics, 61 p874-888. Offutt, Susan E. (1982), Structure of the United States Corn Market and the Impact of Export Instability, Ph.D. Dissertation, Cornell University, New York. Offutt, Susan E. and David Blandford (1983), A Review of Empirical Technigues for the Analysis of Commodity Instability A.E.Res 83-7 Department of Agricultural Economics, Cornell University, Ithaca, New York. Oi W. (1961), "The Desirability of Price Instability under Perfect Competition," Econometrica, 29 p58-64. Olson, Mancur (1965), The Logic_pf Collgptive Action, Harvard University Press. Parker, Russell C. and John M. Connor (1979), Estimates of Consumer Loss Due to Monopoly in the U.S. Food Manufacturing Industries," American Journal of A ricultural Economics, 61 p626—39. Parkes, Richard W. (1978), "Inflation and Relative Variability," Journal of Political Econom , 86(1) p79-95. Pasour, E.C. Jr (1980), "A Semi Strong Evaluation of the Efficiency of the Hog Futures Market: Comment," American Journal of Agricultural Economics, 62, Aug, p581-583. Pindyck, R.S. and D.L. Rubinfeld (1976), Econometric Models and Economic Forecasts, McGraw-Hill Book Co., New York. Piggott, R.R. (1978), "Decomposition of the Variance of Gross Revenue into Demand and Supply Components." Australian Journal of Agricultural Economics, 22 p145-178. Powers, Mark J. (1970), "Does Futures Trading Reduce Price Fluctations in Cash Markets?" Americgn Econgmic Review, 60 p460-464 e Pratt, J.W. (1964), "Risk Aversion in the Small and in the Large." Econometrica, p122-136. Radner, R. (1968), “Competitive Equilibrium under Uncertainty," Econometrica, Jan. Richardson J.W. et al. (1983), "Economic Impacts of Increased Price Variability: A Case Study with Rice," paper presented at NC-169 workshop, West Lafayette, Indiana, Aug 3. EMS Robinson, W.L. (1975), “Unstable Farm Prices: Economic Consequences and Policy Options,“ American Journal of Agricultural Economics, 57, Dec, p769-777. Rothschild, M. and J. Stiglitz (1970), "Increasing Risk I: A Definition," Journal of Economic Theory, p225-243. Rothschild, M. (1973), "Models of Market Organization with Imperfect Information,” Journal of Political Economy, p1283- 1308. Rothschild, M. (1974), "Searching for the Lowest Price When the Distribution of Prices is Unknown," Journal of Political Economy, p689-711. Sackrin, S.M. (1957), "Measuring the Relative Influence of Acreage and Yield Changes on Crop Production," Agricultural Economic Research, vol 1X(4) October, p136-139. Samuelson, P.A. (1965), “Proof that Properly Anticipated Prices Fluctuate Randomly," Industrial Management Review, 61(1) p41-49. Samuelson, P.A. (1972), “The Consumer Does Benefit from Feasible Price Stability," Ouarterl Journal of Economics, 86 p476-493. Sandmo, A. (1971), "On the Theory of the Competitive Firm under Price Uncertainty," American Economic Review, p65-73. Savage, L.J. (1954), The Foundations of Statistics, John Wiley and Sons, New York. Scherer, F.M. (1970), Industrial Market Structure and Economic Performance, Rand Mcnally Publishing Company, Chicago. Schuh, G. Edward (1974), "The Exchange Rate and US Agriculture," Americgn Journgl of Agriculturgl Economicg, 56(1). Schuh, G. Edward (1975), "The Exchange Rate and US Agriculture: A Reply,” American Journal 9f Agricultural Economics, 57(4). Schuh, G. Edward (1976), "The New Macroeconomics of Agriculture,” American Journal of Agricultural Economics, 58(5) p802-811. Shaffer, James D. (1980), "Food System Organization and Performance: Towards a Conceptual Framework," American Journal of Agricultural Economics, May p310-318. 309 Sharpe, William F. (1964), "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk", Journal of Finance, 19 p425-442. Skinner, B.F. (1974), About Behaviorism, Alfred Knopf, New York. Smith, Gordon W. (1978), "Commodity Instability and Market Failure: A Survey of Issues" in Stabilizing World Commodity Markets, F. Gerard Adams and Sonia A. Klein (eds), Lexington Books, D.C. Heath and Co. 1978. Staatz, John M. (1984), A Theoretical Perspective on the Behavior pf Farmers’ Coo eratives, Ph.D. Dissertation, Michigan State University, East Lansing. Talpaz, Hovav (1974), "Multifrequency Cobweb Model: Decomposition of the Hog Cycle,” American Journal of Agricultural Economics, 56(1) Feb p38-49. Tisdell, C. (1963), "Uncertainty, Instability, Expected Profit," Econometrica, 31 p243-247. Tisdell, C. (1978), "Extension of 01’s Instability Theorem," Journal of Economic Theor , 17 p130-33. Tomek, W.G. and K.L. Robinson (1981), Agricultural Product Prices, Cornell University Press, Ithaca, New York. Turnovsky, Stephen (1974), Price Expectations and the Welfare Gains from Price Stabilization, Amepican Jourpal of Agricultural Economics 56 Nov p706-716. Turnovsky, Stephen (1976), “The Distribution of Welfare Gains from Price Stabilization: The Case of Multiplicative Disturbances,” International Egonomic Review, 17 p133-148. Turnovsky, Stephen (1978), “The Distribution of Welfare Gains from Price Stabilizaton: A Survey of Some Theoretical Issues." in Stabilizing Wgrlg Commodit Markets, F. Gerard Adams and Sonia A. Klein (eds) p 119-147, Lexington Books D.C. Heath and Co. Tweeten, Luther (1979), Foundations of Fgrm Policy. 2nd ed., University of Nebraska Press, Lincoln. Tweeten, Luther (1981), Puzzles for Agricultural Economists in the 1980's,” Journal of the Northeastern Agricultural Economic Council, vol X(2), October p1-8. Tweeten, Luther (1983), "Economic Instability in Agriculture: The Contributions of Prices, Government Programs and Exports,” American Journal of Agricultural Economics, 65(5) Dec p922-931. Tweeten, Luther and Delton Gerhoff (1977), "Instability in Supplies and Utilization of US Wheat,” Oklahoma Current Farm Economics, vol 50. von Neumann, John and Oscar Morgenstern (1944), Theory of Games and Economic Behavior, Princeton University Press, Princeton, NJ. Vellutini, Roberto de A.S. (1982), Structural Linkages between Agriculture and thg Rest of the Economy: An gyglugpion of Sggtoral and Macromodels, Staff Paper No.82- 23, Cornell Department of Agricultural Economics, Ithaca, New York. Waugh, F. V. (1944), ”Does the Consumer Benefit from Price Instability?" Ouarterl Journal of Economics, 58 p604-614. Williamson, Oliver E. (1979), "Transaction-Cost Economics: The Governance of Contractual Relations," Journal of Law and Economics, vol XXll, October, p233-261. Williamson, Oliver E. (1981), "The Modern Corporation: Origins, Evaluation, Attributes,“ Journal of Economic Literature, Dec p1537-1568. Willig, Robert D. (1976), "Consumer Surplus Without Apology," American Economic Review, 66(4) Sept, p589-597. M IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII mummzumumwmunrmmmmummm 3 1 37 1464