#1 . ‘ " .. ' ‘ I ; ,1 4‘ ' .c _ 3“" L " '. .1.‘ ', V '- - .. ' '41.,“ , , 1 . {- r.‘ .35..) 4 ~ 5“ 1.. v 4", I-i f V . (I: ~. . 141'.“ fi;}&® A - . ‘, , {In _ ;. :1 '14." ' 'I. .’ 137:??4 u VHIW . "J ”‘5: £14.; J".{. ”4'.“ 120' ”"33 .' , 3 km? 4 , 2"”(H . wart-w . - - . 4% 1 ”NU 41 . . . 1'". . ‘ .wa';A.r,,. . w'..u‘.. “ ”lyu‘gfinmwifiw'fl. “'l'?‘ W M, r. . z}..v:g Ni '4. .334! - I ”I. s - 4‘4“ 4 I". z A". .1 a" mi“ ’4' a: \ bwmfihgpA ' 5.: ”.1. £5.25 ' ’11:: ‘.4\;,'.-‘~l.‘,)' ' ‘ -- ‘ ‘40., ‘3! a} if. ”£7 4 4 4 “I 4' "23‘ . CW.‘ 5:. . 1? 4'55 ‘ .. .4 J ‘ Q I" ' «‘4 .‘a- ”3 . film“. ' 4 .a F 4 {04'1” ' ? 34:15.3 UV 1 4 a” N" ' 29.11;." v ‘0 .353; A??? 535‘”; .5 . 45%.. ' 3...... 4 4:2!th '_.‘.. 1:1.5";fi.y:}. ' .- 41‘). - "-1 "a? t ‘ { i '7 '51- “.7133. .[ 1-7.1} .4" w; -. w (“aft .1 15,313“ x.‘ .3 . +5: :wv...'.—. . < \. Y I't‘. I I ' ".v.‘.‘ o 'a ”if!“ ‘. M t.’ r.'.r..-r fr}; 1' . ; -.-:- r312” fi, . ,I :43. a . "mm! ”Pu: " '1'! t. .,.'.~.I,I,',,:. t’v- 1.3.4} .. “Ii/14;: {m .ad‘un. w”; is; ' “'11. ’0», ?l}.3(.flw 99- ’eg 14"- 5kg; 3 54' FM" 1'”) n] .‘G‘. 4H,:,( 471,4! 9:“ :. I , mm- .., $0M w 1;, r" I ragga! ,4 3,1241. Cg A/ .4“ . ‘.’£},L' J51" ’5’: 1", 5/4? -r 5ng . J ;J n‘: , I Kt”..- ' 7-1;: , H.“ iim‘ffi‘ . 1.; ‘ , .: - ""1. ,‘h v 175."4‘J;3§ 'Tr (by WI. @1231. Jig-2'40. , .- f '1" .";'1"\.~ I 4J‘rIf-1‘!‘ I]! {:4qu , J. _ 'r’n ”'{fl'fi -. In" ,n" Y”, ’47: ‘3‘! rm}, 5 i 4.97;"; {hank "'u {5].}; z. .1... . I". vw ’3: £3215?" ~.;, 139‘ ' 9:2." fig era‘u'- 5-555”: fig. :51‘.»» -: LIBRARY Michigan Stat. University This is to certify that the dissertation entitled FOOD DEMAND ANALYSIS IN URBAN WEST JAVA, INDONESIA presented by Agus Pakpahan has been accepted towards fulfillment of the requirements for Ph . D degree in Mal Economics L. V. Manderscheid/ Major professor Date 030% /7Oé<() V t/ MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771 MSU ’ RETURNING MATERIALS: Place in book drop to NOV 1 9 2003 0c? LIBRARJES remove this checkout from ._;_. your record. FINES will be charged if book is returned after the date stamped below. fiiglee P y, '.~ We SEP 1 9 1999 FOOD DEMAND ANALYSIS IN URBAN WEST JAVA, INDONESIA BY Agus Pakpahan A DISSERIATICN Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR.OP PHILOSOPHY Department of Agricultural Economics 1988 .1 ‘r /‘ [.2 ABSTRACT FOOD DEMAND ANALYSIS IN URBAN WEST JAVA, INDONESIA BY Agus Pakpahan This study sought knowledge about how urban household food consumption behavior is influenced by changes in prices, expenditure, and household size and composition. This knowledge is important for food policy-makers. The Working and the Working-Theil-Suhm (WTS) model were used. Household welfare loss due to price increase and Engel's equivalence scales were also computed. Working's model showed that food expenditure elasticity of a single house- hold is lower than that of other household sizes. Clas- sifying food into ten commodity groups indicated that cereal, sugar, and tobacco are necessities; fish, meat and poultry, and eggs and milk are luxuries; and tuber, vegeta- bles, soybeans and nuts, and fruit are independent of income. Price is an important instrument for food demand policy. This research showed that increase in each commod- ity price significantly reduces the demand for that commodity. The examination of cross-price effects indicates that cassava is not a cereal substitute. The animal products system of commodities such as meat and poultry, fish, and eggs and milk are substitutes. ' In addition, examination of the effects of price increase showed that Agus Pakpahan urban consumer welfare is significantly determined by the price of cereal. The effect of price increase is distrib- uted disproportionately. Engel equivalence scales indicate that to be equally well off a larger household requires more income than does a smaller one. The magnitude of equivalence scales varied across regions in West Java. This research also showed ,that, ceteris paribus, reduction of household size will decrease demand for cereal and will increase demand for other commodities. Therefore, this research implies that family planning is a very important policy which may not only solve the problems of food-population imbalance but may also improve household nutrient intakes. Copyright by AGUS PAKPAHAN 1988 untuk saudara-saudaraku di indonesia yang masih bergelut untuk isi perut ii ACKNOWLEDGEMENT I am indebted to many individuals and institutions and wish to take this opportunity to extend my sincere apprecia- tion for the assistance and cooperation I received through- out my graduate program. I am indebted to the government of Indonesia for providing the opportunity and financial support to expand my intellectual capacity, without which the path of my academic development could have been quite different. To be more specific, I am indebted to the Center for Agra-Economic Research, the Agency for Agricultural Research and Develop- ment (A.A.R.D.) which sent me to Michigan State University to pursue the program--- special thanks here to Dr. Sjarifuddin Baharsjah, a former head of the Center, and Dr. Faisal Kasryno, the current head of the Center. The writer is also indebted to Mr. Budhojo Sukotjo, a former NAR-II project leader; Dr. Djoko Budianto, a current NAR-II project leader, Dr. Ibrahim. Manwan, chairperson. of the training committee of A.A.R.D.; and Winrock International especially to Dr. Ralph Retzlaff, Ms. Roberta Gottfried (former fellowships secretary) and Ms. Pat Whitehead for their assistance. iii A special thanks goes to Dr. Lester V. Manderscheid who supervised the thesis, providing intellectual stimulation and invaluable guidance. I am also grateful to Dr. Stan Thompson for his suggestions and constructive criticism. The study benefited also from the suggestions and insights of Dr. Robert Myers. I am appreciative of their individual contributions. Earlier in my graduate program, Dr. Lee M. James of the Department of Forestry, Dr. Daniel E. Chappelle of the Department of Resource Development, and Dr. Lawrence Libby, now at the University of Florida, performed success- ively as major professors, both in name and deed. Further- more, I am also grateful to Dr. Allan Schmid, Dr. John Hoehn, and Dr. Richard Bernsten who served as members of my guidance committee. Dr. Stanley Johnson of Iowa State University provided the data, and the Central Bureau of Statistics Indonesia gave permission to use them. Dr. Mohammad Wardhani has not only mailed the data from Ames to East Lansing, but also provided important information related to the data. I am deeply grateful to all of them. I have a special word of appreciation for Mr. Chris Wolf and Mrs. Margaret Beaver for computer assistance. I am thankful for their help. My special thanks also go to Mr. Kerr Soerj antono for helping me to recover my damaged files and for his continuing guidance in computer works. I am also indebted to Dr. P. Lovell who made this volume read- able. iv I am deeply indebted to my parents for their encourage- ment and support. I would like also to express my gratitude to my parents-in-law for their support. Finally and not the least of all, to my wife Ani, for her passion, encour- agement, and help; to my sons Angga and Andya, who are still too young to understand the real meaning of work, I offer my warmest gratitude. TABLE OF CONTENTS Chapter I 0 INTRODUCTION 0 O O O O O O O O O O O O O O O O O O O O O O O BaCRground O O O O O O O O O O O O O O O I O O O O O O O 0 Objectives of the Study .............. scope O O ...... O I O O O O O O O O O O O O O O O O O O I O O 0 Organization of the Dissertation ...... II. CONCEPTUAL FRAMEWORK ........... ...... The Allocation Models of Consumer Demand ............................. The Choice of Functional Forms ........ Working's Model ....................... Working's Model Including Substitution Effects . .......... ................... III. RESEARCH METHODS ...... .... .............. Classification of Commodities .......... Definition of Commodities ..... ..... .... Prices of Composite Commodities and Real Food Expenditure ............. Spatial Aggregation .................... Estimation of Engel Curves ............. Measuring Effects of Household Size and Region ............................ Estimation of Demand for Food: Demand System Approach ............... Data COO...0...0OOOOOOOOOOOOOOOOOOOOOOOO IV. FOOD CONSUMPTION PERFORMANCE IN WEST JAVA Expenditure, Food Expenditure and Food Share ................................. Allocation of Food Share ............... Distribution of Household According to Food Share ............................ Summary ................................ vi ~qc-Nra H on 11 17 20 25 25 26 28 30 31 33 35 38 4O 40 41 44 45 V. VI. VIII. EFFECTS OF EXPENDITURE, PRICE AND HOUSEHOLD SIZE ON DEMAND FOR FOOD ..... Engel Curves for Food Across Household Size ................................. Food Expenditure Elasticities Across Budget Share ......................... Effects of Household Composition on Food Consumption ..................... Expenditure Elasticities for Food Groups ....................... ..... ... Compensated and Uncompensated Own Price Elasticities ......................... Compensated and Uncompensated Cross- Price Elasticities ................... . Price Effects Across Commodity Shares.. smary O0..........OOOOOOOO0.00.0.0... WELFARE ANALYSIS OF THE HOUSEHOLD ..... Welfare Effects of Food Price Changes Comparisons of Households' Welfare .... swam OO0.0.0.0.0000...0.00.00.00.00. IMPLICATIONS OF RESEARCH FINDINGS FOR FOOD POLICY ......OOOOOOOOIOOOCOOO A Brief History of Food Price Policy in Indonesia ............................ Implications of Food Price Increase ... Implications of Changes in Household Size and Composition ................. smary OO0.0......OOOOOOOOO0.00.0.0... SUM! MD CONCLUSION ......OOOOOOOOOO BIBLIOGRAPHY APPENDICES vii 51 51 55 59 65 68 71 76 78 80 80 84 89 91 91 98 103 106 108 5.8. 5.9. 5.10. 5.11. LIST OF TABLES Dummy structure for measuring effects of household size and of region ............... Average urban household expenditure for 10 food groups in West Java Indonesia .............. Average budget share and standard deviation for 10 food groups of the urban household inweSt Java OOOOOOOOOOOOOOOOOO00.0.00... .0 Comparison of Working's coefficients, average budget shares, food expenditure elasticities, and sum of squared error for seven household sizes in urban regions in West Java ......... Effects of expenditure and household size on fOOd Share ......OOOOOOOOOOO0.00.0.00......O Effects of household size and region on food Share .........OOOOOOOOOOOOOOOO0.0.0.0000...O Food expenditure elasticities across budget Shares OOOO......OOIOOOOOOOOOOOOOOO00......O Effects of expenditure and household composition on food share according to a region in West Java ......OOOOOOOOOOOO......OOOOOOOOOOOOOOOO Parameter estimates for food groups under the Working framework when prices are assumed constant ......OOOOOOOOOOOOOOOI0.00.00.00.000 Household composition elasticities derived from a constant and a non constant price version Expenditure elasticities derived from WTS ....... Compensated own price elasticities .............. Uncompensated own price elasticities ............ Compensated price elasticities across commodity Shares 0............OOOOOOOOOOOOOOO......O... viii 35 42 43 54 56 58 60 60 64 67 70 70 77 A.8. Compensating variation for a 50 percent price increase with average food expense/week/ household = Rp 7288 .. ........... ....... ..... Values of compensating variations across commodity shares ............................ Equivalence scales for a household with respect to a household composed of two adults and no Children OOOOOOOOOOOOOOOOOOOOOOIO. ........... Effects of 50 percent price increase on changes in quantity consumed ........................ Comparisons of the average actual expenditures and the required expenditures based on Engel's equivalence scales .................. LIST OF APPENDICES Parameter estimates for food groups using WTS in West Java without imposing restrictions ..... Parameter estimates for food groups using WTS in West Java when homogeneity restriction was imposed 0.0IO.......OOOOOOOOOOOOOOOOOO......O Parameter estimates for food groups using WTS when symmetry was imposed ................... Parameter estimates of 10 food groups when block independence between food, sugar, and tobacco was imposed ........................ Parameter estimates for 10 food groups when real expenditure was expressed in per capita term Parameter estimates for food groups when log number of children and log number of adult w.r. incomorated ......OOOOOOIIOOOOOOOOOIOO Parameter estimates for food groups when household size was incorporated ........... ggnpgnsatgg own and cross price elasticities for food groups of urban household in West Java (without imposing homogeneity) ........... 83 84 88 99 105 127 128 129 130 131 132 133 134 A.9. Compensated own and cross price elasticities for food groups of urban household in West Java (imposing homogeneity) ................... A.lo. ggnpgnggtgg own and cross price elasticities for food groups of urban household in West Java (imposing symmetry) ....................... A.11. ngpgngatgd own and cross price elasticities for food groups of urban household in West Java (block independence between food, sugar, and tobacco).................................... A.12. Uncompensated own and cross price elasticities , for food groups of urban household in West Java (without imposing homogeneity) ........ A.l3. Unggmpgnggtgg own and cross price elasticities for food groups of urban household in west Java (imposing homogeneity) . . . . . . . . . . . . . . . A.l4. Uncompensated own and cross price elasticities for food groups of urban household in West Java (imposing symmetry) . A.15. nnggmpgnfigtgg own and cross price elasticities for food groups of urban household in West Java (assuming block independence between food sugar and tobacco) .. ................ B.l. Nutritive values of tropical root crops (per 1009 .dible portion) ......OOOOOOOOOOOOOO0...... C.1. Indonesia (Map) ................................ Ce2e "eat Java (Map) ......OOOOOOOOOOOOOOOOOO00...... 134 135 135 136 136 137 137 138 139 140 LIST OF FIGURES Plot between food share and logarithm of household expenditure Relationships between shares of cereal, vegetables, and tuber, and food share Relationships between shares of meat and poultry, fish, eggs and milk and food share Relationships between shares of tobacco, soybeans and nuts, fruit, and sugar, and food share Distribution of household samples according to food share xi 46 47 48 49 50 CHAPTER I INTRODUCTION Eastman When policy-makers want to design, to implement, or to evaluate a certain food policy, they might want to know the impact of that policy on food consumption, food production, structural changes in food sectors, and the ‘welfare of consumers and producers. In the case of Indonesia in general or West Java in particular, the government may want to know the impact of food. price increase, changes in household income, or changes in household size on quantity demanded of an individual or a group of food commodities. A real example can be taken from the case of household size reduction in West Java. In this province we observe that there was a nine percent household size reduction during the period from 1980 to 1985, or about 1.8 percent per year (C.B.S., 1987). Based on this fact policy-makers might want to know the impact of such reduction on demand for food. Of course, knowledge about the impact of policy changes is different from a policy itself or the creation of a policy. To make a new policy we need to know both value-free positi- vistic knowledge and knowledge about values of the subject imatter with which we are dealing. The latter, therefore, is ‘much more complicated (see Johnson, 1986). 1 2 Food demand studies in Indonesia are not new. Previous demand studies in Indonesia, however, were mostly focused on a single commodity and analysis thereof was mostly based on national data. For example, Timmer (1971a,b) estimated wheat flour consumption and rice consumption, respectively; and Timmer and Alderman (1979) estimated consumption parameters for rice and cassava. The most recent study conducted by Johnson gt 9]... (1986) , using demand system framework, estimated income, price, and household size elasticities for thirteen food groups. As a consequence there is a lack of knowledge about interrelationships among commodities and food consumption behavior across regions. Knowledge about food consumption behavior in particular, or consumption behavior in general, for each region is very important for policy-makers because each region in Indonesia is composed of both different cultural groups and natural endowments. Therefore, the parameters estimated based on national data are too restrictive to be applied to a specific community. Based on this reason, this research was intent on study of the food consumption behavior of urban consumers in West Java, where Sundanese is a majority group, using the demand system approach. W In general this research sought knowledge about the impact of changes in household income, prices, and household 3 size and composition on demand for food in urban West Java, Indonesia. To be more specific, this research sought the following knowledge: 1. Knowledge about the relationship between food effective demand and expenditure, given constant prices and household size for five urban regions in West Java. This is the estimation. of‘ Engel's curve ‘which is important for generating knowledge about the effect of income changes on food effective demand. One might view this as being a simple phenomenon and because it is simple that it is not important. However, Engel's estimates are crucial because : (i) estimation of this curve is easier than the estimation of price effects on demand: (ii) its importance and usefulness for food policy are obvious since we usually have a clearer idea about future income than about future prices; (iii) in some cases income is more important than price, particularly when price does not convey reliable infome- tion. 2. Knowledge about the relative welfare of different household sizes or of household compositions. This know- ledge is important because children not only give utility to their parents, but also create costs. Food costs are very obvious. This knowledge will, for example, be important for taxation policies. 3. Knowledge about the relationship- between food demand and food expenditure, prices and household character- 4 istics under various restrictions such as symmetry and homogeneity. This knowledge may show the behavior of consumer food demand under various circumstances. The estimates themselves are, further, important for food policy analysis. 4. Knowledge about the implications of the findings as a part of means to evaluate or to design food policy in . West Java. In this stage we tried to show the implica- tions or relations between our findings and some important policy objectives. 5.29.9.1. The scope of this study was limited to the study of consumer behavior where the household was treated as a unit of analysis. It is important to state explicitly that the household rather than the individual was treated as a unit of analysis because it will make us aware that the unit in this research is different from the unit in the theory of consumer choice, which is based on individual preferences and budget constraints. Choosing the household rather than the individual as a unit of sample is unavoidable since data are based on family or household units. In addition, choosing the household rather than the individual will be more appropriate with respect to the empirical world. Most food consumption decisions are made by and - within house- holds. Finally, choosing different consumption units will 5 result in different policy implications (see Atkinson, 1983; Atkinson and Stiglitz,1980:26). The scope of the analysis was also limited to examina- tion of the (quantitative) relationship between food consumption and household composition and size, income, and prices. In addition, equivalence scales, cost of children and welfare losses of consumers due to price increases have been computed. Finally, this research was neither an evaluation of nor research for designing a specific ,food policy to solve a specific food problem. This research was intended to generate knowledge which is important for food policy decision-making and not food policy itself. Food 'was defined for' 10 broad categories -of food: cereals, tuber, fish, meat and. poultry, eggs and. milk, vegetable, soybeans and nuts, fruits, sugar, and tobacco. These food groups compose about 87 percent of the total food budget in West Java. Sugar, though it composes only about 3 percent of the household food budget, possesses an important position in public policy agendas. Tobacco, on the other hand, composes a large part of the household budget. For example, the proportion of tobacco expenditure in West Java samples is about, 12 percent of total food expenditure. Finally, focusing analysis on food only is realistic especially when we realize that about 60 to 70 percent of total expenditure in developing countries' economy goes to food. 6 This research is also limited in geographic scope. It dealt only with the analysis of demand for food in the urban areas of the province of West Java. The main reasons why this research selected urban West Java as a geographic unit of analysis are: (i) we deal with specific decision makers, i.e., a governor, under whose authority may exist a specific food problem. (ii) Choosing a specific cultural background which is revealed in food habits will be appro- priate as a first approximation of homogeneous preferences. Aggregating across cultural groups will be too restrictive. (iii) Environment may contribute to different patterns of food habits, kinds and quantity of food available, and the role of the household in the economy. The latter is very important in that the role of ‘the rural household can simultaneously be as food consumer and food producer. The urban household, on the other hand, usually acts as food consumer only. Therefore, limiting the scope to the urban consumer makes the analysis simpler. (iv) Most previous food consumption studies in Indonesia were based on a national aggregate. The parameters estimated in those studies are too restrictive for a specific cultural setting. This research should not be viewed as a substi- tute to the national based studies but should be viewed as their complement. 7 W This dissertation was divided into eight chapters. Chapter I presented background information, the objective of the study, and the scope of the study. Conceptual framework and research methods were discussed in Chapter II and Chapter III, respectively. Chapter IV presented general descriptions of important food consumption perform- ances in West Java which are important for doing analysis in subsequent chapters. Chapter V and Chapter VI presented the estimates of demand parameters and welfare analysis of the household, respectively. Chapter VII showed implications of findings on food price policy and Chapter VIII consisted of the summary and conclusions of this study. Finally, biblio- graphy and the appendices were placed at the end of this volume. CHAPTER II CONCEPTUAL FRAMEWORK W The consumer is assumed to have a nice utility function, e.g., continuously differentiable and strictly quasi-concave, to represent consumer preferencesz. Our problem here was to model the behavior of the consumer which is assumed to maxi- mize his utility function subject to a linear budget con- straint. The solution to this problem was used as a frame- work for conducting the empirical estimation of demand para- meters. For complete treatment of the solution to the above problem, that is, a system of Marshallian or Hicksian demand functions see Varian (1984), Russell and Wilkinson (1978), Layard and Walters (1978) and Theil (1975). There are at least five different available methods for modeling such a system of demand equations (see Theil and 1 The framework is called an allocation model because it has a unique characteristic: the sum of the components equals the aggregate. Thus, if the consumer's budget during the analysis is assumed to be fixed and the consumer alloc- ates the total expenditure among various goods and services, then the summation of expenditures across goods and services in the budget must equal total expenditure (see Theil, 1975: Bewley, 1986). 2 See varian (1984) and Russell and Wilkinson (1978) for complete discussions of axiomatic structures of consumer preferences. 9 Clements, 1987; Deaton, 1986; Barten, 1977: Brown and Deaton, 1972). The first one is well known as a pragmatic approach. It is a method of estimation where the specification of demand functions is neither generated from demand theory nor are the restrictions generated by demand theory utilized3. Criticism to this approach is usually associated with its lack of theoretical plausibility. Double log equations which (are' very popular specified demand equations, for example, violate the adding-up constraint (Yoshihara, 1969). A second approach to demand specifications belongs to Stone's methods, that is, the specification of demand equations derived from direct utility function. The Klein- Rubin utility function is a well known utility function underlying the Linear Expenditure System of demand equations. This system of demand equations is derived based on the assumption of additivity of preferences. It fulfills theoretical demand restrictions such as homogeneity of degree zero in income and prices, symmetry, and additivity. However, this is not a flexible demand function because inferior goods are excluded (Johnson g3; 11., 1984:64). In addition, additivity of preferences according to Deaton (1974:346) "will lead to severe distortion of measurement". That is, additivity assumptions imply approximate linear relationships between own-price elasticities and income elasticities. 3 Demand theory provides restrictions such as additiv- ity, homogeneity of degree zero in prices and income, and symmetry of cross price effects. 10 Therefore, if the own-price elasticity of i increases, its income elasticity must increase as well. As a consequence, income elasticity of j must decrease. The third well known approach of demand specification is a system of demand equations derived from an indirect utility function. A system of demand equations is derived by using Roy's identity‘. Well known examples are the indirect translog demand system which is extensively used by Christen- sen, Jorgenson, and Lau (1975) and. the indirect addilog demand system (Houthakker, 1960). These demand systems are both consistent with demand theory and are flexible. However, translog models have some disadvantages such as (Theil,1980): (i) their parameters have no simple economic interpretation relative, for example, to the linear expenditure system, (ii) the number of parameters tends to increase about propor- tionally to the square of the number of goods, and (iii) they are nonlinear in parameters (Johnson g; 31., 1984). Further- more, Flood, Finke, and Theil (1984) showed that, judged based on the behavior of the estimates of income elasticities for Japanese and Swedish data, the translog model 'was inferior relative to the Working model. Income elasticities for food derived from the translog model indicated that the higher the income level, the higher the value of income 4 Let v(p,M) be an indirect utility function. Roy's 6v/6pi identity says that - -------- - xi“. 6v/6M 11 elasticities for food. Finally, the indirect addilog demand system yields income elasticities which are independent of the level of income, and the cross price elasticities are only affected by the commodity whose price is changing and not on the good whose quantity is responding (Johnson, _e_t al., 1984). Specification of demand equations based on a specified cost function is the fourth approach. The AIDS model invented by Deaton and Muellbauer (1980b) which is generated from the PIGLOG class of preferences is a well known example of demand equations derived from a specific cost function. The final approach is that of the specification of demand equations not based on a specific cost function, an indirect utility function, or a direct utility function. They are, however, based on a direct differentiation of a general form of Marshallian demand equations and then applying the results of utility' maximization subject to budget constraint. This method was invented by Theil (Theil, 1975), and is well known as a differential approach to model demand functions. The Rotterdam model is a familiar example. The_QhQise_Qf_£unctienal_Eerme The results of empirical demand research are largely determined by the correct specification of the algebraic functional forms used. This step is the most difficult part in the research process because there is no common agreement 12 among economists regarding the forms of function which are best suited to our purpose. In this respect "neither economic theory nor available empirical knowledge provide, in general, a sufficiently complete specification of the economic functional relationship so as to determine its precise algebraic form" (Lau, 1986:1516). (See also Kmenta, 1971:532). The field of empirical demand analysis provides a good example of not only how views about correct functional forms of demand equations vary among researchers, but also the views about whether we should base our specification on utility, indirect utility, cost function, or not base specification on them at all. Based upon these divergent points of view we found at least five different approaches to specifying demand functions as discussed above. The choice of functional form discussed here is the ex ante choice of the algebraic form of the function prior to actual estimation. Therefore, there are almost unlimited candidates for algebraic functional forms, including kinds and number of variables, the forms of the functions : linear or non-linear, number of equations, etc., available to the researcher. To narrow this possibility and to avoid making an arbitrary choice of the functional forms, we need to establish criteria. Lau (1986:1520) provided five criteria for determining algebraic functional form : (i) theoretical consistency, (ii) domain of applicability,_ (iii) flexi- bility, (iv) computational facility, and (v) factual conform- 13 ity. Therefore, we chose the functions which meet those criteria. Applying the above criteria reduces the field of choice. For example, the Linear Expenditure System is excluded from the field of choice because it is not flexible, e.g., inferior goods are excluded. The double log demand system is also excluded because it violates the additivity restriction, and is therefore, not consistent with the theory. However, making a choice among translog, AIDS, and Rotterdam models is quite difficult. The Rotterdam model is not derived from a utility, an indirect utility, or a cost function, but as argued by Theil (1980), why should we believe that true consumer preferences are correctly represented by translog or PIGLOG cost funct- ions ? Theil argued that we have no need to specify utility function or cost function to represent consumer preferences in the first place. What we need to do is utilize the results of consumer utility theory without regard to any specific utility or cost functions in order to represent consumer behavior toward price and income changes (Theil, 1980).5 The translog and the AIDS models are consistent with 5 See Theil (1975, 1980) and Theil and Clements (1987) for full discussion of the differential approach to consumer demand. 14 theory, and the Rotterdam model6 is also indirectly derived from utility theory. Furthermore, they are flexible. To choose among them, then, we need to rely on the fourth and fifth criteria presented above. The computational facility criterion is important particularly for cases in developing countries where sophis- ticated computer programs are usually not available. The translog model which is non-linear in parameters requires more complicated software and demands more computing costs. However, the AIDS model is also non-linear in parameters, but we can still specify its linear version by choosing its appropriate price index, e.g., Stone's price index. The Rotterdam model, in addition, can be both, depending on the assumptions about its marginal budget share and price effects. Based on this knowledge, there might be no clear cut argument for choosing one of the functional forms among competing alternatives because they may require the same degree of computational facility. Since these models are already in existence and have 5 Recent development was made by Mountain (1988) . Mountain (1988) found that the approximate compensated elasticity computed at a particular budget share and income elasticity derived from the Rotterdam model are not different from elasticities generated from other functional forms. The difference is that the Rotterdam model started with expenditure shares rather than with the underlying support function, e.g., an indirect utility or cost function. The discrete Rotterdam model, like the other flexible functional forms, at the individual consumer level is a valid linear approximation in variable space. The order of approximation is no lower than that for other flexible functional forms. 15 been used for awhile, the fifth criterion becomes crucial in the process of choice of functional forms if the candidates for functional forms cannot be excluded by the first four criteria. The problem here is deciding’ what kinds of indicator or performance are appropriate to tell us that, say, functional form F conforms to reality better than other functional forms. This is a problem of interpretation of reality or fact. In the framework of logical positivism (Johnson, 1986:43) "by interpreted we mean a language. in which abstract symbols are treated as standing for something regarded as part of the real world". Based on this view, an hypothesis which has passed the tests and has been accepted as a theory can be viewed as a part of the real world. Therefore, factual conformity must be subjected to a theory. In the field of demand analysis, theory says that effects of income on quantity demanded, given constant prices, can be used to classify commodities into luxuries, necessities and inferior goods. It is commonly accepted that food is a necessity based on empirical estimation of its income elasticities (Working, 1943: Laser, 1963: Theil and Suhm, 1981: Theil and Clements, 1987). This knowledge can be used as a criterion of factual conformity. That is, a functional form conforms better with reality if it consist- ently predicts that food is a necessity. Flood, Finke and Theil (1984), using- Japanese and Swedish data, tested the predictive performance of the 16 translog and working models based on their income elastici- ties. The translog model gave an unstable prediction. For Japanese data, food income elasticities increase as income level increases, but for Swedish data food income elastici- ties decrease as income level increases. Food income elasticities for Japanese data are unacceptable as judged by earlier findings. On the other hand, the Working model produced stable and acceptable results. For both sets of data the Working model resulted in decreased income elastici- ties for food as income level increased. Therefore, based on this criterion we excluded the translog model from our field of choice. The AIDS model is the last model available. In a constant price situation this model is identical with the Working model. Another alternative is the Working-Theil-Suhm (WTS)7 model which is also derived, as is the AIDS, from the Working model but was generated through a differential approach to consumer demand (Theil and Suhm, 1981) . This research used the WTS model because it more or less meets all of the criteria provided above. In addition, El-Eraky (1987) showed that the WTS model was superior relative to the AIDS model in estimating food demand parameters for Egypt. Since application of the WTS is still rare relative to the popular AIDS model, this also motivated the writer to use the WTS. 7 The name of Working-Theil-Suhm model is adopted from El-Eraky (1987). 17 The following sections will discuss the derivation and properties of these models. W Working (1943) discovered the general relationship between budget shares on commodities and total consumer expenditure. The most important findings are (i) " the proportion of total expenditure devoted to the different purposes tend to be about the same for families of the same total expenditure per person even though the families differ with respect to income, size and proportion of income saved:" and (ii) " the jproportion of 'total expenditure ‘that is devoted to food tends to decrease exactly in arithmetic progression as total expenditure increases in geometric progression" (Working, 1943: 45). The second result is very important as a basic foundation for both estimation of demand parameters and welfare analysis. The latter is associated with Engel's law, namely, the percentage of income spent on food is inversely related to the level of income. In addition Engel used food share as a common denominator for making welfare comparison, that is, two households have an ' equal welfare if they have an equal food share (see Deaton and Muellbauer, 1980a: Deaton, 1981: Deaton and Muellbauer, 1986). The algebraic form of Working's law for- food category can be expressed as: 18 (2.1) W1 = 31 + bi log M where "i and M are the share of food in total household expenditure and total household expenditure, respectivelya. To conform with the budget constraint we need to restrict ai and bi : (2.2) 21 a1 = 1 , Ebi = 0 Working's model has a typical form of relationship between its marginal budget share and its. budget share, namely9, (2.3) Bi = Wi + bi 6xi where Bi = Pi --- ,i.e., marginal budget share of good i. 6M Therefore, in Working's model the divergences between marginal budget shares and budget shares of its corresponding commodities are determined by how significant the behavioral response, bit is to changes in the total expenditure, given prices and other factors remain constant. Working's model also provides a specific form of income elasticities. The algebraic form of income elasticities derived from the Working model is: 8 Pi X1 th “1 a --;-- , where Pi is the i 's commodity price. 9 This relation can be derived as follows : Multiply (2.1) by M, one obtains : (i) wiM = a1 M + b1 M log M. This is equivalent to : (ii) pixi = a1 M + bi M log M. Take the derivative of (ii) with respect to M gives: (iii) 6(pixi)/6M - 81 + bi + bi 109 M = bi + Vi: 19 (2.4) E1 8 1 + (bi/Vi) The values of income elasticities can take E1 2 1 , 05 E1 5 1, and E1 < 0. The first condition indicates that good i is a luxury good for that class of expenditure level. The second condition implies that the good i is a necessity with respect to that income level, and the third condition implies that good i is inferior for that income level. - Another interesting case derived from Working's model is that the ratio between marginal budget share and budget share for the same commodity is equal to income elasticity for that commodity. Mathematically, it can be expressed : 81 53: 91!: (2.5) "' = [P1 " l / [""] W1 6M M Equation (2.5) implies that the status of a commodity i with respect to income is determined by both marginal budget share and budget share of that commodity. The value of Vi will always be greater than or equal to zero but Bi can be positive, negative or zero. In summary, four important properties of the Working model have been shown: (1) the Working model conforms to empirical evidence of food demand parameters, particularly income elasticities for food (Working, 1943: Leser, 1963: Flood, Finke and Theil, 1984: Theil and Suhm, 1981: Seale and Theil, 1986, 1987: El-Eraky, 1987): (ii) Working's model is consistent with theory (Deaton, 1986): (iii) Working's model is flexible (El-Eraky, 1987): (iv) the Working model 20 allows perfect nonlinear aggregation across consumers (Muellbauer, 1975: El-Eraky, 1987). H l' , H I J I J I' E l !i! ll BEE ! The most important elements in empirical demand analysis are the estimation of income elasticities, and compensated and uncompensated of own and cross price elasticities. Price changes have two effects: income effects and substitution effects. This section deals with the incorporation of price variables in the Working model based on a differential approach to consumer demand which is appropriate for analy- zing cross-sectional data. Most demand system studies appearing in the literature are based on time series data. For developing countries such as Indonesia, time series data are relatively scarce with respect to number of observations required for statistical estimation. In addition, time series data are also subject to criticism such as the substantial correlation existing between real income and the relative price of food, and the shifts over time in many factors not included in the equations (Crockett, 1960). Cross-sectional data are an alternative source of information. The utilization of cross-sectional data has some advantages such as (i) avoiding the problems of serial correlation and structural changes which_ are usually problematic in analyzing time series data, (ii) usually more 21 disaggregated to a particular region or to socioeconomic aspects of the population (Green, e3; a1", 1979). However, cross-sectional estimates are also subject to criticism especially when one wants to use such estimates for making economic forecasts. In addition, cross-sectional analysis is not appropriate for analysis of the consumption of durable goods because the time variable here is crucial. Another criticism of cross-sectional data is that there is a lack of variation of price variables in cross-sectional data. This is not entirely true. The existence of varia- tions in a consumer's reservation price is one possibility. Another possibility is the existence of search costs making uniform price impossible (Diamond,1978). A good example is gasoline price. Gasoline prices in the U.S.A. , even though this product is chemically homogeneous, are rarely identical from one gas station to another nearby competitor. There- fore, the existence of price variations across regions causes the existence of deviation between the observation and the mean of prices. The following is an attempt to construct a model of demand equations incorporating effects of price on quantity demanded based upon cross-sectional datalo. Let us define wihit as a budget share for a commodity i of a household h at the prices p1*,i.e., the geometric average of prices across households, and average geometric 1° See Teklu and Johnson (1987) and Johnson et a1. ( 1986) for estimates of price elasticities for food commod- ities in Indonesia generated from cross-sectional data. 22 . expenditure level Mh*. Then we have Working's model in the form : (2.6) W1h* - a1 + bi log Mh* Adding Vih - Vih to the left hand side of (2.6) and re- arranging terms we obtain: (2.7) wih = a1 + bi log Mh* + (wih - Wih*) where Vih is the observed budget share of good i for a household h. Differences between Vih and wih* are due to the differences between prices paid by each household ( Plh: p2h, .., Pnh) and the average price of each good (p*1). P*i is defined as the geometric mean of prices across households, i.e., A (2.8) log P*i = 1/H 3h log Pih , i= 1,2,..,n: h= 1,2,..,H, and Pih is the price of a commodity i paid by a household h. The results of the differential approach to consumer demand givesll: (2.9) widuoqxi) - Bump) [ duog 14) - 2k vamp) duos pp] + 8j V13(H:P) [M109 Pj) - 2k 3101.9) 6(109 Did] and assumes the parameters are constant. The existence of (Win - Vih*) can be interpreted as a result of price changes from p*1 to Pih when the real expenditure Mh. remains constant. Within the framework of a differential approach, constant real income means that: “- See Theil (1975:1980) for a complete discussion of the derivation of this equation. 23 (2.10) d(log M) - Ek wk d(log pk) = 0, or d(log u) - 2k wk d(logpk) Therefore, demand equations become : (2-11) W1 d(logxi) - Zj V1j [d(log Pj) - 8k 31 d(log Pk)] Recall also the result of total differential of budget share as: (2.12) dwi = “1 d log p1 + "i d log xi - “1 d log M Substituting the above results in this equation gives : (2.13) dwi - wi(d log p1 - Ej d log pj) + 23 Vij [d(log pj) - 3k 31 d(log Pkll dwi can be interpreted as (Vih - Vih*) and V1 is interpreted as Vih and then substituting (2.13) into (2.7) gives : (2.14.a) Yih = ai + b1 log Mh + Ej 'ij log Pjh/Pj* To incorporate household composition variables we add N, N1 and N2 variables in the right hand side of (2.14.a) as: (2.14.b) Yih = 31* + b1* log Mh + :3 'ij* log Pjh/Pj* + silogN (2.14.c) Yih s 81 + Pi log Mh + Ej *ij log Pjh/Pj* + 311 log N1 + 821 log N2 where Yih - w1h(1-log Pin/Pi* + 83 th log Pjh/Pj*)120 fij is an element of the nxn with rank (n-l) Slutsky matrix and log p*1 as defined in (2.8): and N, N1, N; are the size of household, the number of household members 5 10 years of age (children) and the number of household members > 10 years of 12 See Theil and Suhm (1981) for derivation of this expression. 24 age (adults), respectively. The homogeneity and symmetry constraints are also linear in their parameters. The adding-up restriction of WTS means that (2.15) 21 31 I 1, Ebi I E tij = 0. The homogeneity restriction is given by (2.16) Ej 'ij I 0 for i I 1,2, ...., n. and Slutsky symmetry is given by (2.17) tij = tji for all pairs (i,j) where if j. Price and income elasticity based on the above WTS model are (i) compensated price elasticity: e11 a ”ii/Vi uncompensated price elasticity: 6*11 = 811 - (V1+b1) (ii) expenditure elasticity : E1 I 1 + (bi/Vi) Comparing the WTS (2.14) to the AIDS (see Deaton and Muellbauer, 1980a,b) we may conclude that (i) the WTS and the AIDS will reduce to the Working model when prices are assumed constant, (ii) the substitution terms of the AIDS are much 'more complicated, involving’ double-subscripted, para- meters relative to the WTS which has only single subscripted parameters.i CHAPTER III RESEARCH METHODS :1 s.“ !' E: ii” The concept of commodities involves both goods and serv- ices. Arrow and Fisher (1974) gives precise properties attached to commodity, that is : place, time, and physical properties. Shubik ( 1987) added ownership as an important property of commodity. The implication of those properties for empirical work is crucial, since we will have indefinite numbers of commodities which are impossible to investigate empirically. In empirical work we need a small number of commodities, that is "it is almost a necessity to simplify matters artificially so as to reduce the number of variables which are to be handled” (Samuelson, 1963:144). In other words we need to summarize the information through grouping goods together when they display similar roles in consumer behavior (Simmons, 1974:61). The method of commodity classification in this research is as follows: (i) it is assumed that food is separable from other commodities such as housing, clothing, and so on, including leisure. It is justifiable to assume that cross price effects among highly aggregated goods vanish (Theil, 1975). (ii) Food is composed of 10 commodities such as cereal (CER), tuber (TUB), fish (FISH), meats and poultry 25 26 (MEP), eggs and milk (EGM), vegetable (VEG), soybean and nuts (SOYN), fruit (FRT), sugar (SUG), and tobacco (TOB). These ten commodities have been chosen not based on knowledge about elasticities of substitutions nor complementarity among commodity elements such as suggested by Hicks (1981) but based on our a priori knowledge about food needs and food habits among Sundanese. The ten food groups mentioned above are assumed to represent total food consumption of the household. This assumption is realistic because those ten food groups compose the major household food expenditure (87 percent of total food expenditure). In addition, the remaining food catego- ries are difficult to include because they have neither price nor quantity variables. The main purpose of this research was to analyze the behavior of household food consumption toward changes in total expenditure, prices tand.2household. size. Therefore, the categories of food ‘which do not contain prices, or’ price ‘variables cannot. be generated from ‘the available data, and are thus excluded from the analysis. Finally, the most important reason for excluding those kinds of food categories is that they' are not important in (current) food policy issues. Wines As a consequence of aggregation the term commodities as defined above is not self-explanatory. For example, the 27 meaning of cereal, tuber, or meat might not be directly understood. Therefore, it is necessary to clarify the term by providing elements of the aggregation. Definition of food groups1 Cereal (CER) Cereal includes all types of food and food products which are produced from rice, corn or wheat: glutinous rice, rice, corn, wheat flour, corn flour, and others. Tuber (TUB) Tuber is a category of food including cassava and its products, sweet potatoes, potatoes, sagu, 'talas' (taro), and others. Fish (FISH) Fish is a category of food including sea fish, fresh-water fish, salted and dried fish, canned fish, shrimp, crabs and oysters, and others. Meat and Poultry (MEP) Meat and poultry are categories of food including beef, lamb, pork, chicken, and others. Eggs and Milk (EGM) This category of food includes eggs, fresh milk, dried milk, condensed milk, and milk products. Vegetables (VEG) Vegetables include chinese spinach, 'kangkung' (swamp 1 The translation from Indonesian language to English follows Wall (1985). 28 cabbage), chinese cabbage, green beans, 'kacang panj ang' (yard long beans), tomatoes, carrots, cucumber, cassava leaves, egg plants, bean sprouts, shallot, garlic, chili, 'petai', 'genjer', 'jengkol' (stink beans), and others. Soya beans and Nuts (SOYN) This category of food includes: peanuts, mungbeans, red kidney beans, soya beans, cowpeas, tofu, tempe (soya bean cake), 'tauco' (soy paste), 'oncom' (fermented cake), and others. Fnfit(flfl) The fruit category includes oranges and tangerines, mangoes, apples, avocados, 'rambutan', 'dukuh', 'durian', 'salak' (snakeskin fruit), pineapples, bananas, papaya, 'jambu air' (rose apple), 'jambu biji' ( guava), 'belim- bing' (star fruit), 'sawo' (sapodilla plum), watermelon, and others. Sugar (SUG) Sugar includes palm sugar and granulated cane sugar. Tobacco (TOB) Tobacco includes clove cigarettes, cigarettes, and tobacco. a :; e. eliee: ; “UH... ;: ..,e {:2 "00‘s Ae‘ge - The commodities defined above are composite commodities. Price of commodity is defined by geometric average (Theil and Shum, 1981) and is expressed in natural logarithmic form. This is important with respect to practical usage of such 29 prices because the WTS and the AIDS models utilized variables defined in logarithmic form. Budget share of each individual cgmmgdity_gf a commodity group, for example, budget share of rice in the budget of cereal, is used as weight. The calculation procedure is as follows: /#———\LKRTVQMJ€ (3.1) log pih - skei uk/ui log(Mk/th) where pin is (composite) price of commodity i paid by (household h, say cereal paid by h, Mk is expenditure of household h on commodity k where k is in i, say rice or corn, M1 is household h expenditure on (composite) commodity i, and th is quantity of commodity k bought by household h. The last term of the right hand side of (3.1) equals price of commodity k per unit. Equation (3.1) implies that we permit households Egnpay differfifljihggiges for. 3.11.3... finemggmodity. Therefore, there must be an average price for each i which differs from prices paid by households. The_averagg%p£igeflgf commodity i iQLGalgnlated~a§ follows: 1 H (3.2) log p*i - --- 2 log(p1h), h = 1,2, ...., H. H h=1 where H is total household number. This expression is the same as (2.8) in Chapter II. Finally, we need to calculate the price‘iggxfifgg all. cgmmgdities. This is calculated by : 10 (3-3) lop 9* '121 win 1°9( Pih/P*1) 30 where wih = (pih xih)/Mh, that is the share of commodity i in the nth household food budget, uh. Real household food expenditure is household food expenditure deflated by (3.2): (3.4) log Mh = log(Mf/p*), where Mf is nominal food expendi- ture. Widen Spatial aggregation in this research is an aggregation of households into spatial units. Spatial unit is an administrative unit such as district or kota madya. The latter is like an urban administrative unit. Therefore, there are no rural household categories in this unit. The total number of districts and kota madya in West Java is 23. Furthermore, district/kota madyas are aggregated into larger spatial units called regions. The criteria to aggregate those districts/kotamadyas into regions are :(i) agro-ecological similarities, (ii) contiguity of the areas, and (iii) similarities in social customs. and 'traditions. Based on these criteria we have five regions: ( 1) Region A (North-West Coast): Pandeglang, Lebak, Tangerang, and Serang districts, (2) Region B (Priangan): Bogor, Sukabumi, Cianjur, Bandung districts and Kodya Bogor, Kodya Sukabumi, and Kodya Bandung, (3) Region C (East Priangan) : Sumedang, Garut, Tasikmalaya, Ciamis districts, (4) Region D (North-East Coast): Kuningan, Majalengka, Indramayu, Cirebon districts and kodya Cirebon, and (5) region E (North-Coast): Subang, 31 Purwakarta, Karawang and Bekasi districts. The regional unit is important for making regional comparisons such as house- hold equivalence scales or costs of children. W W Estimation of Engel curves was classified into two categories with respect to the aggregation of commodities: first, Engel curves for food as a single aggregated commo- dity: and second, Engel curves for each commodity of food. In both cases we assume prices are constant. Furthermore, for the purpose of computation of Engel equivalence scalesz, the numbers of children and adults per household are speci- fied as explanatory variables. We use Working's model in the following way: (3.1) whi I aih + bhi log MP + error where i refers to commodities, and h refers to household.) h in this research is defined for seven household sizes, Mh is total expenditure (expenditure on food and non-food commo- dities) and aih and bin are parameters. To conform with the budget constraint we need to restrict the parameters : (3.2) 21 a1 I 1 , Ebi = 0 and to avoid singularity of the variance-covariance matrix, 2 We will discuss the estimation procedures of equiv- alence scales and its results in Chapter VI. 32 we drop the non-food equation. Since the result of dropping the non-food equation is a single equation, (3.1) was estimated by OLS. Based on ( 3.1) we have seven equations, one for each household size. Error terms in (3.1) are assumed to have the following properties: (i) the mean of error terms is zero [E(e1)I0]: (ii) variance of error terms across observations is constant (homoskedasticity) [ E(ei’) I 0‘]: (iii) covariance of error terms is zero (nonauto- regression) [ E(eiej) I 0 for i f j] (iv) normality, i.e., 51 is normally distributed (Kmenta,1971:202). Equation (3.1) is used to estimate Engel coefficients for food in West Java. Engel coefficients for non-food can be recovered using the property of (3.2) . In addition, we are interested in comparing the results if we use expendi- ture of household in per capita terms: (3.3) "i I a1 + b1 log M/N + error Equation (3.1) was also modified by adding number of children, N1, i.e., number of household members with ages 5 10 years: and number of adults, N2, i.e., number of household members with ages > ten years, as additional explanatory variables. (3.4) W1 3 a1 + bi 109 M + 01 N1 + 62 N2 + error 33 Here we assume number of children and number of adults are independent and are exogenously determined from household decisions. Various alternative specifications of (3.4) following Deaton (1981) such as quadratic forms involving the interactions between demographic variables and household expenditures were also attempted.3 (3.4) was used to estimate Engel's parameters for West Java samples and for samples in each region. WWW Besides income and prices, food demand is also deter- mined by the size of the household. It is intuitively appealing that larger households consume more food than do smaller households, given other factors remain constant. Furthermore, size may also affect the expenditure's para- meter, e.g. , more percentage of additional income goes to food for a larger household. We use dummy variables to approximate the effects of size on household food consumption such as (3.8). (3.8) win I aih + bih logM + 3h dh Sh + Eh 9h Sh*logM + error where Sh (hIl,3,4,5,6, 27) is household size. The size of the household here is represented by dummy variables, that 3 Quadratic forms with or without interaction among variables were tried but unsuccessful because there was always singularity in the variance-covariance matrix. The singularity of the variance-covariance matrix~ is caused by the existence of perfect colinearity between log M and its square. 34 is, variables which take binary values: they have a value of 1 if they belong to a certain category of h and zero if otherwise. Table 3.1. below clarifies the problem. To avoid perfect collinearity among dummy variables and the intercept, one of them must be dropped. In this research we dropped household size I 2 by assigning zero if the samples belong to this category (region A in the case of estimating the effects of region). As a result, the estimated equation for household size I 2 will be in the form of: (3.9) W12 I aiz + biz 109M Equation (3.9) was used as a reference, namely, we compare all of the other equations to (3.9). This occurs, for example, when the coefficient of dummy variables of both intercept (d1) and slope (91) of‘ household. belonging to household size I 1 are significantly different from zero. We can write the estimated equation for household belonging to household size I 1 as: (3.10) wil - (a12+d1) + (b12+gl)logM We see that the behavior of household size I 1 is measured relative to the behavior of household size I 2. The parameters .of other' household size categories were also measured relative to household size I2. Therefore, household size I 2 is called a reference. The same procedure is used to measure effects of region on food consumption. 35 Table 3.1. Dummy structure for measuring effects of house- hold size and of region Variables Dummy Structure 31 82 S3 S4 85 86 37 H. Size: 1 1 0 0 0 0 O 0 2 O 0 0 0 0 0 O 3 0 0 1 0 O 0 0 4 0 0 O 1 0 0 O 5 0 0 0 O 1 0 O 6 0 0 0 O 0 1 0 2 7 0 0 0 0 O 0 1 Region: R1 R2 R3 R4 R5 A 0 0 0 0 0 B 0 1 0 0 O C 0 0 1 O O D 0 0 0 1 0 E 0 0 0 0 1 ,; ; u, _., . .-,.,. . .q. - ..,... e -u ,.. .1 , Food in this research was defined for 10 aggregated commodities (see previous section for the definition of food items). Furthermore, we assumed that food is separable from non-food commodities including leisure. The food demand system here is known as a conditional demand system. This research utilized the WTS model of demand system as developed in Chapter II. The functional form of the WTS demand system is : (3.11.a) Yih I a1 + bi 10g Mh + Ej 'ij log Pjh/Pj* + ‘ih (3.ll.b) yin - a1* + b1* log uh + mi 113* log pjh/pj* + 81 logN + 5*ih (3.11.c) yih - a1 + 51 log uh + 8j #13 109 Pjh/Pj* 36 + 811 109 N1 + 821 109 N2 + 21h where h is an index for a household (h I 1,2, .... , H), andi i I 1,2, ..., 9, and 61b is an error term. We drop the mall: commodity to avoid singularity due to the property of total; “I. w- Nfi ..W - o *W sum of elements equaling aggregate. The dropped equation canzigi ...1-7 ...-......M ...-...... ..W— I—u-‘A- ““ be recovereibLusing the homogeneity assumption. ( See Theil J I I “7‘“ “ I (1975, 1980), Bewley (1986)). The form of (3.11) is usually called seemingly unrelated regEEEEion (SUR) because the error ”terms in different Mug-I“ -.....M-w. ov-W‘ '--"H‘ equations are possibly mutually correlated (Kmenta, 1971: 4...- HM *u-m—“MW '“uwuwm'lw— a‘. ”.7. 518) . Equation (3.11) can be estimated by 01.8 for each commodity. The resulting parameters are unbiased and consistent. However, " by estimating each equation separ- ...,» mIM» ately and independently, we are disregarding the 1nformat1on about the mutual correlation of the disturbances, and the Mflwfir‘A effieiency of the estimators becomes questionable " (Kmenta, 1971:518). The best linear unbiased estimator of (3.11) is given by Aitken's generalized least squares for instances when the variance-covariance matrix is known, or the Two-Stage Aitken estimator when the variance-covariance matrix is unknown. The first stage in the latter procedure is to estimate the variance-covariance matrix from ordinary least square residual as suggested by Zellner (1962). The two stage Aitken is asymptotically equivalent to Aitken's generalized least square estimator and, therefore, to the maximum likelihood estimator (Kmenta, 1971:525). 37 Furthermore, Aitken's estimator will be identical to OLS in two special cases : (i) when the error terms of different equations are actually unrelated, or (ii) when each of the seemingly unrelated regressions involves exactly the same explanatory variables (Kmenta, 1971:521). This is the case in demand system research. Therefore, (ii) implies that demand parameters are invariant of whether OLS or Aitken's estimator is used. However, using a system approach such as SUR provides us an opportunity to apply symmetry restriction across demand equations and to estimate them under such a circumstance. In this research we used both OLS and SUR. The earlier method is used for checking of the latter method. For SUR estimation we used an algorithm available in the SAS computer program, the two stage Aitken estimator or Zellner method. The mechanics of estimating parameters using SUR are as follows: (1) Write equations for each commodity in a form of (3.11). Then, we will have ten equations. (ii) Drop one equation, e.g., an equation for tobacco, from the estimation. Then, we have nine equations in a system. The parameters in the dropped equation can be recovered by using homogeneity restriction. (iii) Apply an algorithm of SUR available in the SAS package to estimate the parameters. 38 Data A large household data set for Indonesia can be found in the National Economic Surveys (SUSENAS) conducted by the Central Bureau of Statistics. The 1980 data are called 1980 SURGASAR data. This set of data contains not only SUSENAS data but also includes other data which are usually generated from agricultural, animal husbandry, prices, and village statistics surveys. The data used in this research are data which are stored on magnetic tape. The data were obtained from Iowa State University under permission from the Central Bureau of Statistics of Indonesia. The sampling frame was started with the division of a region into rural and urban areas. Surveys of agriculture and animal husbandry therefore are only conducted in rural areas. In this sampling frame there are two kinds of sample units: (1) a village unit, and (ii) a household unit. A village unit was made based on the 1980 population census. This sampling unit is used to select the samples up to village level. Within this sampling unit, census blocks were selected. Finally, the household samples were selected from each census block (C.B.S., 1980a). Urban areas, except Jakarta, in all provinces were classified according to population size. A three stage sampling procedure was used. At the first stage, n villages were drawn. Furthermore, a block census was .drawn randomly from each village. Finally, about 5 to 10 households were 39 systematically drawn from each census block after they were classified according to their main source of earnings. The overall sampling fraction was about 1/500 to 1/1000 household (C.B.S., 1980a). This research analyzed the data of urban households in West Java from the 1980 SURGASAR data. The total number of households analyzed was 1905. CHAPTER IV FOOD CONSUMPTION PERFORMANCE IN WEST JAVA WWW There are numerous ways to measure household food consumption. One of the methods used by researchers measures household consumption based on the physical amount of food items actually consumed. Other researchers use household expenditure on various food items or use the proportion of expenditure spent on food items (food shares). Each method has its advantages and limitations. For example, by using food share we have a free unit of measurement, and therefore, we may compare food consumption across commodities. Quantity of food consumed, based on this method, can be computed from the share if we have expenditure and price data. The average weekly (total) household expenditure and household food expenditure in West Java in 1980 were Rp 14,199 and Rp 7,288, respectively; and the average food share was 51 percent. This means that the average urban household in West Java spent about one half of its total household expenditure on food. This is lower than the average food share for urban households in Indonesia (59.84 percent) in 1980 and the average food share in Java and Hadura in 1976 (60.23 percent) (C.B.S., 1978, 1983). Therefore, the average household in West Java spent less of its income for food than did the national average or the average of households in Java 40 41 and Hadura. According to Engel's law, the average urban household in West Java has a higher welfare level than the household in Java and Madura or in the nation. Fig. 4.1 (see figures at the end of this chapter) shows the relationship between food share of the urban household in West Java against a natural logarithm (log) of (total) real expenditure per week. The corresponding plots indicate that food share is negatively correlated to log of real expendi- ture. This result supports Engel's hypothesis about the relationship between food share and income or expenditurel. AW Tables 4.1 and 4.2 show the average urban household expenditure and the allocation of food expenditure on each food item considered in this research. The three largest food expenditures were for cereal, meat and poultry, and tobacco. These three food items composed 56 percent of food expenditure. Expenditure for cereal was the largest because it is a main foodstuff for most Indonesians. Expenditure for meat and poultry was large not because the households consume a large amount of meat and poultry but because the price of this item is high. In these tables we also observed that cereal had the lowest variability in both expenditure and 1 More discussions of the relationships between food share, and food commodity share and total food expenditure, food prices, and household structures can be found in the next chapters. 42 share in food budget. Table 4.1. Average urban household expenditure for 10 food groups in West Java Indonesia Food groups Mean Standard Coefficient of (Rp./Week) Deviation Variation (%)* Cereal 2513 1306 52 Tuber 240 237 99 Fish 797 810 102 Heat and Poultry 1568 1668 106 Eggs and Milk 767 906 118 Vegetables 513 424 82 Soybeans and Nuts 560 544 97 Fruit 567 768 135 Sugar 194 , 184 94 Tobacco 894 843 99 "((1) a) * Coefficient of variation (CV) was computed using the formula: CV a (SDx100)/mean It. will be interesting' to (observe the relations of commodity shares, e.g., share of cereal, with total food share.2 We expect that the higher the food share of the household, the higher the share of cereal and the lower the shares of meat and poultry, eggs and milk, and fish. Fig. 4.2 shows the relationships between food share and the shares of cereal, vegetables, and tuber. Our expectation was true for cereal, that is the higher the food budget share, the higher the share of cereal. This is sufficient to show that the poor spend more income for cereal, and the rich 2 Food share was used instead of income or expenditure because the author believes that food share gives a better measure of household welfare than does income, especially for cases in developing countries. Furthermore, using food share as a measure of welfare is also consistent with Engel's law. 43 do otherwise. The relationship between proportion of expenditure on tuber and vegetable and food share was not clear. Fig. 4.2 shows that tuber consumption is increasing as food share increases but at a rate much lower than the rate of cereal. Furthermore, consumption of vegetables seems to be decreasing at a very low rate as food share increases. Table 4.2. Average budget share and standard deviation for 10 food groups of the urban household in West Java Food groups Mean Standard Deviation Coefficient of (SD) Variation* (’3) Cereal .30 .1027 34.2 Tuber .02 .0190 95.0 Fish .10 .0602 60.2 Meat and Poultry .14 .0852 60.8 Eggs and Milk .09 .0657 73.0 Vegetables .06 .0294 49.0 Soybeans and Nuts .08 .0492 61.3 Fruit .06 .0408 68.0 Sugar .03 .0145 48.3 Tobacco .12 .0648 54.0 Total Food Expenses 1.00 * Coefficient of variation (CV) was computed using the formula : CV a (SDx100)/mean ' Fig. 4.3 clearly shows that as food share increases, the share of meat and poultry in household budget decreases. This is intuitively plausible because the poorer the household, the lower its purchasing power will be. Therefore, the poorer household buys less meat and poultry. This figure implies that low income households fulfill protein require- ments by consuming more fish, except for households who have 44 a sufficiently low income, i.e., food share greater than 75 percent. The pattern of eggs and milk in relation to food share does not appear to be linear. As food share decreases (income increases) the household spends more of its income on eggs and milk. However, after its food share drops to less than about 45 percent, the household spends less of its budget on eggs and milk. ‘ Finally, shares of tobacco, soybeans and nuts, fruit, and sugar with food share might be independent. Food share which is an approximation of“ household ‘welfare. does not determine the household expenditure pattern on tobacco, soybeans and nuts, fruit, and sugar (see Fig. 4.4). Wm In the above section we examined the allocation of food budget among its components. In this section we were interested in knowing the distribution of households in urban West Java according to food share. Even though this research was not a study about poverty, knowledge about distribution of households according to food share is important for food policy discussions. Fig. 4.5 shows that ,the distribution of households according to food share in urban West Java more or less approximates a normal distribution. About 34 percent of households spend about 55 percent of their income on food and about 62 percent of the household samples spend more than 45 half of their income on food. This situation indicates that a majority of households in West Java spend a great deal of their income for food. Summary Cereal, meat and. poultry, tobacco, and fish.‘compose about 66 percent of the food budget. The largest food expenditure is for cereal and the lowest food expenditure is for tuber. Food share and total expenditure (including non food expenditure) seem to have a negative relationship. Furthermore, relationships between food share and each share of food groups in the food budget show : (i) cereal has a positive relationship with food share. This means that the poorer the household the larger the proportion of cereal in the food budget. (ii) In contrast, meat and poultry shows a negative relationship with food share. The rich household spends more on meat and poultry than does the poor one. (iii) As food share decreases (welfare increases) a household spends an increasing portion of its budget on eggs and milk. However, after the food share reaches about 45 percent from the right direction, the expenditure for eggs and. milk declines. (iv) Fish seems to be a main source of animal protein for low income households. (v) The proportion of vegetable, tuber, tobacco, soybeans and nut, fruit, and sugar seem independent of ‘the levels of food. share (welfare). Finally, the distribution of households according to food share is approximating normal. 46 1.04 0.94 0.8- O’.7" 0 ° 0 . . .:.' 0.5— ° .3 0.5— ° ,3 0.4— Food Shore (%) 0.3— .4 0.2- ° 0.1‘ e .1 0‘0 I I T I ' I f l T I . I I i 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 Log Expenditure (Rp./month) Fig.4.1. Plot between food shore and log household expenditure in urban West Java, IndoneSIo 47 100s g; 80-l v 0 E L 60- U) >5 :‘.:.' E 401 cereal E 1 M/ O O 20-l : C ; f 3 _; ——: mfiou. : - : : 3 A ‘05.! 0 fl r T T r fik ff r t r f r If r f 1 0 10 20 ab 40 50 6'0 7'0 80 90 100 Food Shore (Percent) Fig. 4.2. Relationships between shores of cereals, vegetables, and tuber, and food shore 48 100- i; 80‘ V 0 B L 60- (D >~ .t’ 8 40~ E. E 'l O . o 20- ‘ :‘Mmmm ' ' \eggandmflk OIIfiI'IfiIIIrI'I'rTI‘fi 0 10 20 30 40 50 60. 70 80 90 100 Food Share (Percent) Fig. 4.3. Relationships between shares of meat and poultry, fish, eggs and milk, and food share 49 100- @804 v o 'l '6 £604 m >‘ J :2'.’ '840~ E .l E O 020- Nit—4w soyondnut -fI-a Nng—I'LC : f; t 0ITIIII'T'I'III'I 'fi 0 10 20 30 40 50 60 70 80 90 100 Food Share (Percent) Fig. 4.4. Relationships between shares of tobacco, soybeans and nuts, fruit, sugar, and food share 50 sass-z“. x22.“-Haas-“.33”.“as.Haas-.3333: “a.H3%“.H.“as.H.Ha.H.“as.“a.“as.“son-H22.“assess-Haas.“-H. .3“.stass-“2.“.“asagas-HI.“-Has-HI.“-H. gas-H.assuage-Hz ”222.“: z“. A”. 401 0" q A 0 O 0 3 2 1 ARV 205030: .6 tabEaz 1.0 0.0 0.2 0.4 0.6 0.8 Food Share Fig. 4.5. Distribution of household samples according to food share CHAPTER V EFFECTS OF EXPENDITURE, PRICE AND HOUSEHOLD SIZE ON DEMAND FOR FOOD Total food and food groups have been analyzed separate- ly. The main purpose of this separation is that we want to know both the consumption behavior of food as an aggregated good and the consumption behavior of food as more disaggre- gated commodities. The earlier knowledge is important particularly for the formation of macro policy which usually deals with aggregated variables. Knowledge about household consumption of food items such as cereal, meat, and so on, furthermore, is important for specific food policy. This chapter presents the results of estimating food demand parameters in urban West Java. We limited the analysis to estimation of the effects of expenditure, food prices and household composition on food demand. W122: This section presents the estimates of effects of income and household size on food consumption, given prices and other factors remain constant. Household size is considered an important factor because any increase or reduction of size, given other factors remain constant, will affect the household's effective demand for food. Food, in this 51 52 section, was treated as a.single good because we are inter- ested in knowing' about the Ibehavior' of’ households' food effective demand toward changes in income and household size. Table 5.1 (p.54) presents the results of the simple regression analysis. Expenditure elasticity of food, when samples were pooled and household (total) expenditure was used as an explanatory variable, was around 0.86. In addition, it became 0.78 when expenditure per capita was used as an explanatory variable. Both figures, however, indicated that food is necessity. Furthermore, food expenditure elasticities were not constant across household size. Table 5.1 shows that food expenditure elasticity for the household size of 1 was much lower than that of other household sizes. This implies that a group of single households values food less than other household sizes and will spend more of its additional income on non-food goods than will other house- hold sizes, provided they have the same amount of increase in income. This is intuitively plausible and conforms to empirical observations. It is possible because a single individual will have more freedom to spend income than will a married household. Married, and particularly married couples with children, given identical income levels, require more money for food since there are more members in the household in need of food. This means that household size may shape preferences and obviously shapes household, budget con- straints. 53 Table 5.2 (p.56) provides similar information to that in Table 5.1 except now we measured, using dummy variables, the effects of household size on food consumption behavior. Among dummy variables, there are only two estimates signifi- cantly different from zero, that is dummy intercept and slope for household size = 1. Thus, the intercepts and the slopes for all sizes of households, except for household size = 1, are the same as the intercept and the slope of household size = 2. This result seems to correspond to the results in Table 5.1. Equation (5.1) below denotes the effective demand behavior for food for household size a 1. (5.1) w =- 2.1144 - 0.1991 109 M Food expenditure elasticity associated with (5.1) is 0.5490. Now, let us consider the effects of household size together with the effects of region on household food consumption. Table 5.3 (p.58) presents the results of the effects of region on the intercepts and the slopes of Working's equation. As we found in Table 5.2, the results in Table 5.3 also show no changes in information regarding the effects of household size an intercept and marginal budget shares of food, except if we use a 10 percent level of significance. That is, a single household living in region D (Kuningan, Cirebon, Kodya Cirebon, Majalengka, and Indramayu) has an equation different from the reference household and reference region (region A). Working's equation for a household size a 1 who resides in region D is: 54 Table 5.1. Comparison of Working's coefficients, average budget shares, food expenditure elasticities, and sum of squared error for seven household sizes in urban regions in West Java Category bhi w Ehi SSE # of observa- tions Food: Pooled -.0708 .5132 .8618 47.6583 1905~ Caput -.1124 .5132 .7810 43.6796 1905 H. size: 1 -.1992 .4415 .5488 19.4664 65 2 -.1040 .4884 .7871 3.8457 198 3 -.0944 .5253 .8202 4.6034 310 4 -.0925 .5158 .8206 3.7710 281 5 -.1202 .5171 .7675 4.0912 316 6 -.1125 .5184 .7830 2.6968 234 7 or more -.1102 .5187 .7875 4.7114 501 Non-food .1124 .4868 1.2309 - 1905 (caput) Notes: bhi's is expenditure's coefficient and ahi is ignored w is average budget share for each household size Ehi is food expenditure elasticity SSE is sum square of error of each regression All coefficients are significantly different from zero at a 5 percent level of significance The non-food equation is obtained from homogeneity assump tion. 55 (5.2) W = 1.8215 - 0.1706 ln M The food expenditure elasticity associated with (5.2) is 0.61. ... _49-12' _ - l.s ' -‘ ._ os- 3.20? In Table 5.1 we saw food expenditure elasticities for different household sizes. By construction, the Working model also provides a relationship between expenditure elasticity and budget share for each corresponding commodity. The relationships are as follows: E1 = 1 + ( bi/Wi) As we observe, there is an inverse relationship between E1 and wi. Furthermore, we are also able to classify the commodi- ties based on the behavior of “1 and bi- If “1 > 0, hi < 0 and Vi >Ibil, then the good is a necessity: If "i > 0, hi < 0 and Vi ten years, (N2). 'Parameter estimates in Table 5.5 have been derived from equation (5.3): (5.3) wiR = aiR + biR log MR + c1 N1 + c2N2 + error, where R refers to region (R - A, B, C, D, E). Our interest here was to estimate income and household composition elasticities of food across regions. Income elasticities of food can be calculated using Working's formula as given above. Furthermore, household composition elasticities of food demand can be calculated using: Gwi N M c N (504) EN = ..... -- a ...... 5“ X1 P1 W1 where N can be N1 or N; as defined above, and c can also be c1 or c; as in (5.3). Applying the above formula and taking the values of wi such as "i = 0.5132 (see Table 5.1), and taking N1 = 3, we obtain, for example, a children elasticity of demand for food in West Java of 0.14, and adult elasticity of food demand of 0.07. The latter was evaluated at N2 I 2. Therefore, if the Table 5.4. Food expenditure elasticities across food shares 60 Food Share (W) Expenditure Elasticities (E1) 000000000 00.00.... \omslmuusuww -0.12 0.44 0.63 0.72 0.77 0.81 0.84 0.86 0.87 Notes: E1 was calculated based on the Caput equation in Table 5.1 Table 5.5. Effects of expenditure and household composition on food share according to a region in West Java Region Intercept Log M N1 N2 West Java 1.4414 -.1101 .0254 .0176 (.0522) (.0061) (.0026) (.0022) A 1.6634 -.l412 .0391 .0272 (.1470) (.0169) (.0074) (.0056) B 1.4458 -.1101 .0253 .0162 (.0582) (.0068) (.0028) (.0023) C 1.6323 -.1228 .0044 .0076 (.3584) (.0432) (.0189) (.0076) D 1.1417 -.0836 .0308 .0203 (.1084) (.0129) (.0059) (.0052) E 1.2597 -.0888 .0161 .0207 (.1378) (.0164) (.0060) (.0046) Notes: Number in the brackets is the value of the standard deviation N1 and N2 are household members with age 5 10 years and age > 10 years, respectively. 61 number of children is doubled, quantity demanded for food will increase by 14 percent, given other factors remain constant. In addition, closer examination of Table 5.5 indicates that the response of food share with respect to changes in number of children is greater than the response of food share with respect to changes in numbers of adults, except for cases in regions C and E. A similar finding was reported by Deaton and Muellbauer ( 1986) for Indonesia using 1978 SUSENAS data. This is reasonable for developing country cases because most expenditure for children in developing countries goes to food. Applying the income elasticity formula derived from Working's model obtained income elasticities for food: 0.78, 0.72, 0.78, 0.76, 0.83, and 0.82 for West Java (aggregate), regions A, B, C, D, and E, respectively. Differences in income elasticities across regions are not large. Now, examine the household consumption behavior where food was disaggregated into ten groups of food. Table 5.6 (p.64) presents the results showing the effects of expendi- ture, number of children, and number of adults on the effective demand for each food item, given prices and other factors remain constant. Food expenditure, number of children and number of adults reveal different effects on different kinds of food. The response of most commodity shares with respect to changes in household composition is 62 negative, with the exception of cereal. Comparing Table 5.6 and Table 5.5 indicates that when food categories are lumped as food, the effects of number of adults and number of children on food demand are always positive. Disaggregating food into 10 commodities, however, shows that relationships between food demand and household size are negative except for cereal. The positive relationship between household size and food consumption indicates that the positive effect of household size on cereal consumption outweighs its negative effects on consumption for other food groups. Therefore, breaking food into more specific groups gives more knowledge about effects of changes in expenditure and household size toward changes in food consumption. A positive sign of household size effects on the equation of demand for cereal indicates that an additional household member will increase demand for cereal, given other factors remain constant. At the same time, demand for meat and poultry, fish, eggs and milk decreases as household size increases. Therefore, given a fixed income, increase in household size will increase demand for cereal, but will decrease demand for other items, mainly luxury foods such as meat, fish, eggs and milk which are more expensive. This behavior is reasonable because at a given fixed income, an additional member creates cost to the household. The most obvious cost is expenditure for necessities (cereal in this case) and because income is fixed, then there must be a 63 reduction in spending for other goods, namely, more expensive food such as meat and poultry, fish, and eggs and milk. Table 5.7 (p.66) indicates household composition elasticities. The number of children elasticity of demand (based on a constant price version) for cereal is 0.19 which means that demand for cereal will increase by 19 percent if the number of children is doubled, given total expenditure, number of adults and other factors remain constant. The number of adults elasticity of demand for cereal is quite high: doubling the number of adults will increase demand for cereal by 43 percent. Stated another way, reducing the number of adults, for example, from four to two will reduce the household demand for cereal by almost one half relative to initial consumption. The demand for eggs and milk, and meat and poultry will, on the other hand, decrease by 29 percent if the number of adults doubles. Doubling the number of children, moreover, will reduce demand for meat by 12 percent. Negative effects of changes of number of adults on the demand for luxury foods are obvious. Furthermore, given a fixed income, a household will also spend less on tobacco if its size increases because food is more important to serve the needs of all of the members of the household (Bojer, 1977). 64 Table 5.6. Parameter estimates for food groups under the Working framework when prices are assumed constant Share Intercept Log Mf Log N1 Log N2 CER 1.6309* -.1658* .0586* .1314* [.1019] [.0127] [.0099] [ .0114] TUB .0593 -.0038 -.0034 .0003 [.0319] [.0037] [.0029] [ .0033] FISH -.1364 .0272* -.0078 -.0097 [.0886] [.0103] [.0081] [ .0093] EGM -.2698* .0437* -.0128 -.0261* [.1026] [.0120] [.0094] [ .0108] MEP -.6813* .0951* -.0174 -.0406* [.1176] [.0137] [.0137] [ .0124] VEG .0848+ -.0012 -.0053 -.0055 [.0469] [.0054] [.0042] [ .0049] SOYN .2132* -.0163 -.0006 -.0083 [.0733] [.0086] [.0067] [ .0077] FRT -.0519 -.0144 -.0075 -.0115* [.0655] [.0076] [.0059] [ .0068] T03 .1312 .0053 -.0061 -.0426* [.1021] [.0119] [.0093] [ .0107] Notes: * parameter significance at a = 5% N1 might contain zero values. SAS program treats them as missing values and those are excluded from the com- putation. Therefore, the parameters under the log N1 column should be read cautiously. Table 5.7 shows number of children and of adult elasticities for food items which were derived from both a fixed price model and prices included in the model. The comparisons of household composition elasticities of demand across models show that the WTS with and without price variables in the model gives almost the same magnitude of elasticity (see Table 5.6 and Table A.6 to identify the parameters which are significantly‘ different from zero). This means that the effect of household size is independent 65 from price variables. WWW Estimation of the WTS model based on assumptions of homogeneity and symmetry, and with no restrictions on parameters has been tried. Table 5.8 (p.67) below presents the results. The main hypothesis here is that fish, meat and poultry, and eggs and milk are luxury foods and the remaining of food groups are necessities. Since what we have estimated are food conditional demand functions, expenditure elasticities are not directly obtained from. the results. The following' procedure 'was ‘used. to calculate expenditure elasticities. We define that xi = f(Mf), where Hg is total food expenditure. Mf is assumed to be a function of total expenditure, Mf - g(M). Then, by chain rule we obtain expenditure elasticity, E1: 6 log xi 6x1 M 6 x1 Mf 6Mf M (5.4) E1 - ---------- - ----- a ------- - ----- 6 log M 6 M xi 6 Mf xi 6 M Mf This can be simplified as : Bi I 31' - Ef, where Ei'is conditional expenditure elasticity of category i and Bf is food expenditure elasticity. We obtained Ef from Table 5.1 using the caput equation with E: - 0.78. 66 Table 5.7. Household composition elasticities derived from a constant and a non constant price version constant price Price included Commodity version1 in estimation N1 N2 N1 N2 Cereal 0.19 0.43 0.19 0.42 Tuber -0.17 0.02 -0.22 0.05 Fish -0.07 -0.10 -0.11 -0.16 Meat and -0.12 -0.29 -0.12 -0.28 poultry Eggs and -0.14 -0.29 -0.12 -0.21 milk Vegetables -0.08 —0.09 -0.08 -0.08 Soybeans -0.00 -0.10 -0.02 0.01 and nuts Fruit -0.12 -0.19 -0.13 -0.20 Tobacco -0.05 -0.35 -0.02 -0.28 Notes: 1 Computed based on Table 5.6 2 Computed based on Table A.6 Entries in Table 5.8 were calculated based on Table A.1, Table A.2 and Table A.3. Zero expenditure elasticities do not mean that there are no effects of expenditure on the consumption of such food categories, but at a 10 percent significance level the parameters are not significantly different from zero. Based on the above results we can see that the expenditure elasticities obtained are invariant to the imposition of homogeneity and symmetry restrictions. In addition, as one usually expects, we find that meat and poultry, fish, eggs and milk are luxury foods for households in West Java. The expenditure elasticities for fish, and meat and poultry are around 1.03 and 1.23, respectivelyu This research also 67 indicates that cereal, sugar, and tobacco are necessities with expenditure elasticities of about .64, .72, and .64, respectively; Tuber, ‘vegetables, soybeans. and. nuts, and fruit, however, cannot. be clearly’ classified. since. their expenditure elasticities are not significantly different from zero. The latter corresponds to Fig. 4.2 and Fig. 4.4. In these figures we see that the shares of tuber, vegetables, soybeans and nuts, and fruit are independent of the food share. The last column in Table 5.8 reports results from Johnson gt g1; (1986). Table 5.8. Expenditure elasticities derived from WTS Commodity E1 E2 E3 E4 Cereal 0.64 0.64 0.61 0.235 Tuber 0.00 0.00 0.00 Fish 1.03 1.02 1.02 Meat and poultry 1.23 1.22 1.22 1.48b Eggs and milk 0.96 1.00 1.11 Vegetables 0.00 0.00 0.00 Soybeans and nuts 0.00 0.00 0.00 Fruit 0.00 0.00 0.00 Sugar 0.72 0.72 0.84 0.48c Tobacco 0.64 0.64 0.76 Notes: E1 - No restrictions on the parameters have been imposed, estimated by OLS E2 = Homogeneity was imposed, estimated by SUR E3 - Symmetry was imposed, SUR E4 8 Estimates are taken from Johnson et_al. (1986) where a is for rice, b for animal products, and c for sweet- ener Zero is used where expenditure coefficients are not significantly different from zero at a I 10 %. 68 .u.-,~. -. ... , .u.-,;. ;. 04. - ; . .; The main hypothesis here is well known as the law of demand. If the good is normal, (income elasticity for that good is positive but less than one), then increase in price of that good will reduce quantity demanded, given income and other factors remain constant. This hypothesis was tested by the examination of each uncompensated own price elasticity for each food group because that hypothesis deals with the Marshallian demand equations. All entries in Table 5.10 have a negative sign, therefore, our estimates support our hypothesis stated above. In addition, all own price elasti- cities are significantly different from zero at a - 5 percent. This is not the case for cross-price elasticities (see the following section). It implies that own price changes are more important than cross-price changes in affecting changes in demand for each commodity. Table 5.9 (p.70) presents compensated own price elasti- cities which measure the effects of price changes when the consumer is compensated. The Hicksian price effects will be identical to the Marshallianuprice effects if income effect ..fi-vi“ ...- v—m V’- 1--- \ of price changes of the good being considered is zero. WM ...—afi- 1..” -\\ Comparison of Table 5.9 and Table 5.10 (p.70) shows that the rows of tuber, vegetables, soybeans and nuts, and fruit in Table 5.9 and Table 5.10, respectively, contain the same magnitude of price elasticities. This means that changes in 69 price of those commodities do not change the real income of the consumer. The effects of price changes for those commodities are all attributed to substitution effects of price changes. The column entries in Table 5.9 and Table 5.10 denote the compensated and uncompensated price elasticities under no restriction, symmetry, homogeneity, and block independence, respectively. We did the imposition of such restrictions on the price parameters because we are interested in knowing whether own-price coefficients are sensitive to restriction. The results of such imposition show (read Table 5.9 and Table 5.10 according to a column) that there are no big changes in magnitudes (except for sugar) and signs of both compensated and uncompensated own-price elasticities. As suggested by theory all compensated price elasticities have a negative sign which means that given a utility level, price increase will always be followed by a reduction in the quantity demanded. The figures in Table 5.9 were computed directly from the WTS estimates provided in Appendix A. 70 Table 5.9. Compensated own price elasticities Commodity e1 e2 e3 e4 Cereal -0.54 -0.49 -0.50 -0.50 Tuber -1.24 -1.23 -1.25 -1.24 Fish -0.39 -0.37 -0.38 -0.38 Meat and -0.48 -0.45 -0.46 -0.48 poultry Eggs and -0.42 -0.45 -0.41 -0.42 milk Vegetables -0.66 -0.72 -0.66 -0.68 Soybeans -1.02 -0.99 -0.99 -1.02 and nuts Fruit -0.31 -0.28 -0.24 -0.26 Sugar -0.39 -0.05 -0.63 -0.06 Tobacco -0.28 -0.32 -0.29 -0.26 Notes : e1 = Parameter restrictions were not imposed, OLS e2 = Symmetry was imposed, SUR e3 = Homogeneity was imposed, SUR e4 = Block independence (food, sugar, tobacco) was imposed, SUR Table 5.10. Uncompensated own price elasticities Commodity e*1 e*2 e*3 e*4 Cereal -0.73 -0.67 -0.69 -0.69 Tuber -1.24 -l.23 -l.25 -1.24 Fish -0.49 -0.47 -0.48 -0.48 Meat and -0.65 -0.55 -0.63 -0.65 poultry Eggs and -0.51 -0.54 -0.50 -0.50 milk Vegetables -0.66 -0.72 —0.66 -0.68 Soybeans -1.02 -0.99 -0.99 -1.02 and nuts Fruit -0.31 -0.28 -0.24 -0.26 Sugar -0.41 -0.07 -0.65 -0.08 Tobacco -0.36 -0.39 -0.37 -0.34 8*1 = Parameter restrictions were not imposed, OLS e*2 = Symmetry was imposed, SUR e*3 = Homogeneity was imposed, SUR e*4 a Block independence ( food, sugar, tobacco) was imposed 71 eIIe;!:. -._ ... , “lg-F. -._ .-:-- - .e ' -; Another measure of effects of price changes derived from a demand system approach is compensated (uncompensated) gross price elasticity. As discussed in Chapter II, this elastici- ty measures the effects of changes of the jth price on changes of demand for a commodity i. Therefore, knowledge about cross price elasticities is important for analyzing the impact of changing the price of one commodity on demand for other' commodities. For' example, we estimate changes of quantity demanded of meat due to changes in fish price. Since this involves cross-equation effects of price changes, we face more difficulties both in specification, estimation and in testing of the demand parameters. However, a major advantage of using a demand system approach is modeling those interactions within a unit of a system. The compensated and uncompensated price elasticities for 10 food groups in West Java can be read in Tables A.8 - A.15. We also include own price elasticities in those tables which we thought important for making comparisons between direct and indirect effects of price changes. In this section we are interested in examining the following hypo- theses: (i) Tuber and cereal are substitutes: (ii) Fish, eggs and milk, and meat are substitutes: (iii) Vegetables, and soybeans and nuts are substi- tutes: 72 We are mainly interested in substitutes because only for substitutes do price policies have interesting applications. This assertion is a logical consequence of constructing consumer preferences which also has important empirical relevance. If consumer preferences are represented by a Leontief utility function, for example, changes in relative price will have no effects on demand. Changes in price, of course, will affect demand through changes in real income. Therefore, changes in relative price will affect demand as long as a structure of preferences allows for substitution. Asserting cross-price elasticities is an inductive approach to infer commodities' relations. Hypothesis (1): tuber and cereal are substitutes It is commonly believed that cereal and tuber are close substitutes in the preferences of consumers. In fact Timmer and Alderman (1979) show that cassava and rice for Indonesian consumers are substitutes. Table A.8 shows compensated own and cross price elastic- ities for urban households in West Java when demand restrict- ions are not imposed. We see that tuber and cereal are independent goods because at a 10 percent significance level we cannot reject our H0: we accept that they are independent. This finding is also consistent when homogeneity restriction is imposed. Therefore, based on this result we conclude that cereal and tuber are neither complements nor substitutes. 73 This research shows different consumer behavior toward price changes from Timmer and Alderman's results as mentioned above. It is possible that tuber for the West Java community is likely to be independent from other goods, particularly for urban households. Tuber for urban households is not the main staple. A low tuber budget share of the household food budget (2 percent) and high own price elasticity (-1.24) also indicate that tuber in this research is not a price inelastic commodity. A low tuber budget share is also responsible for high own-price elasticity. Evaluating own-price elasticity of tuber if, for example, the household spent 10 percent of its food budget on tuber, tuber became price inelastic (see -Table 5.11, p.77). The main point here is that the demand functions do not reveal that cereal, mostly rice, has closed substitutes. This knowledge is very important for food policy since it indicates that increasing the price of cereal, for example, will not increase demand for tuberl. Hypothesis (ii): Fish, eggs and milk, and meat are sub- stitutes 1 This finding is based on cross-sectional data. An interesting question arises: will the cereal-tuber relation- ship change in the long run or will cross-sectional estimates shift over time ? The answer may be yes or may be no because it largely depends on changes in the structure of consumer preferences. According to Crockett (1960:293) the demand parameter estimates from time series data are usually lower than those from cross-sectional data. 74 Animal protein food sources include fish, milk and eggs, and meat and poultry. It is intuitively reasonable to hypothesize that they are substitutes. Therefore, the cross effects of price changes should have a positive sign. Examination of the tables in Appendix A shows correspon- dence to our hypothesis, with the exception of fish. Fish is not a substitute for meat and poultry and vice versa. However, if we trace back our cross-price elasticity of fish with respect to changes of demand for meat, we find a positive sign but it is not significantly different from zero at a 10 percent significance level (see, e.g., Table A.1). Therefore, we may say that eggs and milk, fish, and meat and poultry are substitutes. This finding is important for further research in the animal products system of commodi- ties. Based on this finding, for example, meat and poultry, fish, and eggs and milk can be analyzed separately from other food commodities without harming our price effects esti- mates. This is important for detailed analysis of the animal product system of commodities when time and financial resources are limited. Hypothesis (iii): vegetables, soybeans and nuts are substi- tutes In hypothesis (iii) we hypothesized that vegetables and soybeans and nuts are substitutes. The reasons for proposing such an hypothesis are mainly based on empirical observations 75 and personal experiences. In this research we found that soybeans and nuts, and vegetables do not reveal a relationship such as hypothesized above. Soybeans and. nuts can Ibe 'viewed. as independent commodities such as tuber and cereal. Changes in the price of soybeans and nuts only affect demand for meat and poultry and sugar. With sugar it has complementary relationships and with meat and poultry it has substitute relationships. Vegetables, on the other hand, have a relationship with fish (complementary), meat and poultry (substitutes), eggs and milk (substitutes), and with tobacco (substitutes). We will ignore the relationship between tobacco and vegetables because such a relationship is intuitively difficult to explain. The reasons why vegetables and, soybeans and nuts are independent are not clear intuitively. Statistically they are independent because we cannot reject H0 at the 10 percent significance level. However, examining signs only (see e.g., Table A.1), we found that they have positive signs in their cross-effect of price changes. Therefore, they are poten- tially substitutes. The examination of cross-effects of price changes for other groups of commodities can be done by the reader. The point is that by using a demand system approach we get knowledge about the relationship of one commodity to another based on consumer preferences. 76 Was 1211 In this section we are interested in examining the own price effects across commodity shares, Vi- This knowledge is important because such elasticities will show the responses of different income status of households toward price changes. For example, households which spend a large amount of their budget on cereal will have a different response toward cereal price changes from the households who spend only. a small amount of their income on the same commodity. The WTS model assumes the price coefficient is constant. Then varying commodity share, vi, we obtain cross and own price elasticity. Table 5.11 shows that poor households which are usually characterized by a high percentage of cereal consumption are not responsive to cereal price changes relative to richer households which are usually characterized by a low percent- age of cereal consumption. 0n the other hand, richer households which are characterized by consuming a high percentage of luxury foods such as fish, meat and poultry, and eggs and milk will not change their behavior very much if prices of such luxury commodities change. These findings are important for identifying the consequences of price control in the food market. This research, for example, suggests that price variables are not a good instrument for helping the poor if we wish to increase the consumption of necessi- ties by the poor. Income subsidy will be better for the poor 77 who need more necessities because they are more responsive to increa- income changes than to price changes. In addition, sing the price of luxury foods, for example, will not hurt the rich very much and nor will the economy be affected. Therefore, the policy related to increasing price of luxury foods and transferring this revenue to subsidize the poor might be a plausible policy. We will continue the discussion about the effects of price changes in the next chapter. Table 5.11. Compensated Price elasticities across commodity shares Commodity Commodity Share (vi) (1) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cereal -1.62 -0.81 -0.54 -0.40 -0.32 -0.26 -0.23 -0.20 Tuber -0.24 -0.12 -0.08 -0.06 -0.05 -0.04 -0.04 -0.03 Fish -0.39 -0.20 -0.13 -0.10 -0.08 -0.07 -0.06 -0.05 Meat and Poul. -0.66 -0.33 -0.22 -0.16 -0.13 -0.11 -0.09 -0.08 Eggs and Milk -0.37 -0.18 -0.12 -0.09 -0.08 -0.06 -0.05 -0.04 Vegetables -0.40 -0.20 -0.13 -0.10 -0.08 -0.06 -0.06 -0.05 Soy. nuts -0.82 -0.41 -0.27 -0.20 -0.16 -0.14 -0.12 -0.10 Fruit -0.14 -0.07 -0.05 -0.04 -0.03 -0.02 -0.02 -0.02 Sugar -0.12 -0.06 -0.04 -0.03 -0.02 -0.02 -0.02 -0.01 Tobacco -0.34 -0.17 -0.12 -0.08 -0.06 -0.06 -0.05 -0.04 Notes : The computation of elasticities was based on Table A.1. 78 Summer! Food expenditure elasticities across household sizes are the same as for household size of two except for the one person household (Table 5.1) . Food consumption behavior across five regions in west Java is also not different from that in region A, except for region D (Table 5.3). Further- more, food expenditure elasticities are varied across food share (Table 5.4). Household size and composition play an important role in household food consumption. Both Working and the WTS model show consistent effect of changes in household size on demand for food. Increasing household size will increase consump- tion of cereal and will decrease consumption of meat and poultry, eggs and milk, fish, and so on, ceteris paribus. Therefore, success in family planning programs will shift demand for kinds of food, from necessities to luxuries (Table 5.7) . As has been expected, cereal is a necessity and meat and poultry, fish, and eggs and milk are luxuries. Tuber, vege- tables, soybeans and nuts, and fruit are independent from total expenditure at a ten percent level of significance. This finding is also supported by Figures 4.2 to 4.4. All uncompensated own price elasticities have a negative sign. Therefore, none of the food groups in this research is a Giffen good. Evaluated at the mean value of each commodity share, tuber, and soybeans and nuts are price elastic 79 commodities. However, if the own price elasticities are evaluated at 10 percent commodity share, only cereal is price elastic (see Table 5.11). Cereal is the most important commodity in the household food budget. The compensated and uncompensated own price elasticities of cereal evaluated at its mean commodity share (30 percent) are -0.54 and -0.73, respectively. Income effect of price changes of cereal is around 1.9 percent for a 10 percent change in cereal price (see Table 5.9 and Table 5.10). There is evidence of substitution among food commodities particularly between animal products. The most important finding' is that the data do not show' that. cereal is. a substitute for tuber for the urban West. Java consumer. Therefore, our result contradicts that of Timmer and Alderman (1979). One possible reason is that we used urban West Java data but Timmer and Alderman used national data. CHAPTER VI WELFARE ANALYSIS OF THE HOUSEHOLD The objectives are to present the estimates of welfare losses of households due to price increase, and to show the estimates of household Engel's equivalence scales using results presented in the previous chapter. Each section begins with a short review of the conceptual framework and procedures of computation of both welfare measures. WW Compensating variation (CV) and. equivalent ‘variation (EV) are two paths of changes of welfare which are usually used in welfare analysisl. CV of moving from situation 1 to situation 2 is defined as (Just e3; a1_,_, 1982:85) "the amount of income which must be taken away from a consumer (possibly negative) after a price and/or income change to restore the consumer's original welfare level": and EV "is the amount of income that must be given to a consumer (again possibly negative) in lieu of price and income changes to leave the consumer as well off as with the change”. 1 The Marshallian Consumer's surplus is another welfare measure. This research did not use this measure because surplus values derived from the Marshallian consumer's surplus are not unique, i.e., path dependent (see Silberberg, 1978). 80 81 Hicksian demand functions can be derived from cost function using Shephard's lemma. Cost function is defined by : (6 1) c(u.p) = Epiinp.u) = M Therefore, CV is represented by: (6.2) cv = c(p2,u1)-c(p1,u1) where p1 and p2 is a vector of prices at initial and for prospective situations, respectively. .Following McKenzie (1983), the equation (6.2) can be approximated by a Taylor series expansion. A Taylor series expansion around the initial value: of the cost function gives: 6 c(pl.u1) (6.3) C(pz.u1) = C(pl.u1) + 21 ---------- dpi 6Di 1 62 c(p1,u1) + - 21 Ej ----------- dpi dpj + R 2 5 Pi 6 Pj where R is the remainder, that is the terms higher than second order. The expression in (6.3) is equivalent to : (6.4) c(p2.u1) ~ c dpi 6xi(pl.u1) + 1/2 21 Ej ---------- dpi dpj 6pj where R is ignored. By moving c(p1,u1) to the left hand side, we obtain CV in the following manner: . 1 1 6x1 amt 29$ 85 «up 8599 899 58599 8899 98 889 98 29:99 358 «8.3 22:: b .353 9828753.. 9:83. «35:. a)... use: 5 «53 9:3 83:88“. 985 .89 sou-sauce soulless ...< 03.5 128 n-59.u 5998.9 589.9 "N59.“ nan-9.9 59.98.. 5599.9 55959.9 3999.9 5998.9 55999.9 5899.9 399.9- 5586- 9559.9 9999.9 9999.9 «N596 n~9.9- 5959.9 98.9 9599.9- ~9~9.9- .3549 85 5989.9 "~98.” 59599.9 8.9 598.9 358.9 38.9 58.. 59599.9 398.9 5989.9 3559.9 8.9 9959.9- 986 5959.9- 5896 58.9 9599.9 89.9 989.9- 5986 98.9- 86 89 59599.9 55959.9 3999.9 588.9 5589.9 558.9 59999.9 398.9 38.9 55959.9 59999.9 59599.. 989.9- 9396- 939.9- 986 359.9 598.9 98.9. 86 n86. 98.9 9996- E96 55. 3999.9 3959.9 598.9 398.9 3999.9 5399.9 59959.9 2.58.. 59999.9 n59~9.9 2,999.9 59999.9 986 38.9 8.9 359.9- 986 ~58.9 «559.9 8.9 8.9 359.9 9599.9 8.9 :59 8999.9 5859.9 38.9 88.9 59999.9 58. .9 3999.9 58.9 3999.9 5359.9 55999.9 5536.9 59.9 §59.9- 58.9- 8.9 ~396- 9986 9959.9 958.9 98.9. 99.96 98.9- 99956 99> R599.”— n9o~94 5989.9 595 59.9 5929.9 5998.9 3.39.9 5998.9 5989.9 5993.9 59999.9 3959.9 9559.9- 958.9- 9959.9 8.9 8.9 N536. 293.9 939.9 58.9 539.9 9559.9 38.9- in 5999.9 5399.9 3959.9 2.39.9 3959.9 5558.9 3959.9 329.9 5959.9 "~93.” 59959.9 55999.9 3~96 98.9 989.9- -¢9.9 ~936- 86- 98.9- 586 .3596 359.9 9999.9 89.9- mm: 3559..“— 559~94 59599.9 85 59.9 5359.9 5998.9 5539.9 398.9 59599.9 358.9 55999.9 5359.9 n39.9- 9959-9- ~86- 98.9 98.9 359.9 958.9 593.9- 95996 5986 9-96 999.. .9- :92 59999.9 5998.9 28.9 538.9 598.9 59599.9 55999.9 538.9 5539.9 358.9 5589.9 55959.9 ....—.996- 8.9 ~586- ~86 989.9- 386 98.9 98.9 mN9-9- 83.9 9599.9- 8.36 35 55959.9 5939.9 3~59.9 55959.9 5989.9 398.9 3-9J 59959.9 3~59J 5939.9 «92.9.. 3955.9 2.36 35 .9 1.8.9. 59596- 399.9 8.9- .3596 5-9.9- 399.9- 58.9. 9999.9. ~5n96 «mu 8599 9899 55.99 25899 9939 £9999 99.99 29:99 599 5999 .22.: .35.: >9 $8.559 .889. n! 5318982. :2! .18... m5: 2.» 9:3 .998 989 .55 nous-Coo 5325...; .~.< 03.5 129 3&8.“ 830.0 an P0.” 38.0. 300.0- £06 gm 898.0 "~84 8000.. 330.0 8.0 530.0. 8.0 MNM0.0 he... 3N8.“ "300.0 808.0 8.80.. "300.0 8.0. ~86 350.0- 080.0 950.0 :8 8.0 88.0 38.0 3000.0 38.0 "ammo.” 88.0. 900.0 ~86 990.0. 8.0- 309.0 0w> H38; 38.” 88.0 330.0 330.0 H800.” ~86 080.0 8.0 0000.0 A5.30.0 pow—.0. 5m "38.” 8000.0 38.0 830.. 230.0 8000.“ £304 008.0 ~86- 53.0 33.0 2.30.0. 300.0 o§.0- ....w: "8.. 8000.0 808.. 830.0 2000.0 308.0 3500.. 308.. 000.0 8.0 8.0 80.0 030.0 N580. ..N~0.0 2.3070. :0: R98.“— 8984 "8.0 38.0 38.0 "“8.“ 8'00.“ 3.4.00.0 80.0.0 008.0. 300.0. 38.0 ”8.0. 8.0 8.0 33.0. 200.0. 33.0 8... "“8.“ 3000.0 830.. 82.0.0 82.0.0 330.. 2000.0 339.0 SPO; none—.0 38.0 8.0. 500.0 3.5.0 58.0. «5.0.0 33.0. 08.0 N36. 8.0. 500... «mu 800 2:00 800 0E0 .898 E0 .6230 SKO .530 #2:: J.E.: 3 5850 .889: no: >503!- coi .993 too» .8» v.3 .0 .3258 532.2... mi 03-» 130 CNS.”— 2000.” 2:050.“— 0—M0.0- 050.0. 000N.0 8» n8.“— n0p00.0 2.58.0 2.8.0- 0N00.0- 8.3.0 30 3000.“. "008.0 .580.“ "008.0 H0000.“— nm08.0 H080.“ nmNNO.“_ "0000.0 "09.4.0.0 0090.0- 008.0 N0..0.0 58.0 080.0- 836 N000.0- NON0.0 0.80.0- «000.0 5.... 3000.0 8.0 "0000.0 "008.”. n00_.0.0 258.0 8000.0 "008.0 2000.0 H090.“ 0000.0 8.0. 300.0 2.8.0 003.0 30.0 0000.0 008.0 08.0 «08.0 250 3000.0 n08.“— 884 n0N8J 380.0 "008.0 "0000.“. n00_.0.0 R0004 “N000.” 080.0- 508.0 5000.0- 508.0 020.0 058.0 0000.0. 83.0 08.0. 300.0 00> H0500.”— n-NSJ ENS-u .0084 ”003.0 308.“. n05004 "800.0 8000.. 3.0504 0090.0 00.0.0 0000.0 0000.0- 0N0.0 008.0 0000.0 0500.0 $0.0 0000.0- .30 2.0.0.0 "030.”. "003.. "058.0 H003.“ n0N 90.0 803.0 n0NO0J 308.0 H0500.“ 080.0 A33.0 0000.0- mm00.0- N500.0- 008.0 080.0 0000.0- 0500.0 m000.0- mm: "0500.0 n0- 3; ENE; "008.0 8090.0 308.0 8500.0 3N3.” 2000.0 "0050.0 530.0- 286 008.0 005.0 m000.0 0000.0- N8.0 003.0 00N0.0 003.0- .5: 830.“. 2h8.0 38.0 "98.0 95000.. "0.4.8.0 880.0 0008.“. HPNSJ 3090.0 5900.0- 86 0.00.0- Np8.0 300.0 N086 00N0.0. N980- n80.0. 300.0 35 n0Np0J $090.0 ”080.0 308.0 8NNOJ nan—0.0 n0Np0d nNNmod 2090.0 "002..“— 00P0.0- 030.0- 0000.0 28.0- 00N0.0 00N0.0- 0000.0- 89.0. Np50.0. N000.0 emu 850 0800 5500 2800 00:0 .8000 00:40 20:00 3.30 009.0 402.2.— . «052— .5 52:50 .88.! a! 8083 u:- ...§o .08. 5233 85388:. x83 :2... 259.0 080 0— 0c .3133 532...... .04 3.1» 131 n0N—0d 803.0 380.0 80—0.... "0.20.0 308.0 "Na—0.0 :08.“— 3804 8000.. 8000.0 8000.0 0000.0- 0MSO 0200 N030 0000.0 0Np0.0 0NN0.0- 003.0 0N00.0 NNN00- 00N0.0- 00.3.0 8» 030.0 H008.”— BPSJ "00.8.0 3N8.” H098.” Eng-u nNNSJ "030.”. 30004 830.0 F20.“— N—000 0.. 90.0. 0N80 008.0. 080.0 000.0 0800 N080 080.0. 0050 N300. N030 30 83.0 Chm-.0.“ "008.0 308.0 Cg.“ 308.0 8000.0 308.0 "0000.0 H003.” 3000.0 H0000.”— 0§.0- 00N0.0- 0200- N080 090.0 0080 «N000- 0080 008.0- 00N0.0 003.0 005.0. 30 Ego-0 H030.“— 900004 SJ 308.0 "008.0 "003.0 308.0 8000.0 n00N0.0 8000.0 3000.0 030.0 88.0 008.0 008.0. 0000.0 0.8.0 00000 38.0 0000.0 5000 N800 E00 250 308.0 nun—0.0 3008.0 H008; 808.0 n88.” .0000.“— 308.” H0000.“ 80.0.0 3000.0 H003.” 008.0 n08.0- 0§.0- “8.0 00.0- N080 F030 0080 N300- 050.0 0000.0 003.0 00> "02.0.0 303.0 8000.0 82.0.0 "090.0 "008.0 503.0 "008.0 H0000.”— n0Nn0J 3000.0 "0000.0 008.0- 0080- «0.00 02.00 0000.0 N030- th00 030.0 300.0 00.0 00N0.0 093.0- Em 30.0.. 3000.0 2.0.0.0 290.. 3090.0 $08.0 $03.0 2.9.0.“. 80004 H0900.“ .020.” 2000.0 ..NN00 050.0 n30.0- 00000 030.0- «08.0. 2.000. :80- pNN0.0 00N0.0- 0300 0000.0- aw: 82.0.0 50.004 3000.0 n0— 5.0 82.0.. "008.0 H090.”— nN08J 3500.0 9050.0 R80.“ 2000.. 030.0- 30 «0'00- s§0 0300 02.00 080.0 0n00.0- 0N80 00.00 0000.0 FONN.0- :20 8g; H008.“ 2.80.0 20.8.0 snug-0 n98; R30.“ 3N8; 880.0 "008.0 "080.0 "0090.0 38.0- 0080. 2.000. 38.0 030.0- 0—80 0000.0 38.0 00N0.0- 0080. 080.0 0n~00 3» .30—0.. H0900.” "0000.0 "N08.” 390.0 308.. ES.” 3— «0.0 3000.0 "003.0 803.0 "0.00.0 pmN00 0NN0.0 N300 88.0- 0000.0 880 N30 N080. 0000.0- 0N00.0- 0009.0. 0000; «mu 8000 0:000 5000 3800 002.0 .8000 09.00 :0 _ 000 3:0 «300 05>...— . cm;— 5 003-50 E: 33.88 c. 38293 8.. 83.2.93 .83 =2 8.... 3.5... .82 2 8.. .8138 3.82... 3. .3: 32.8 8010.”. H8.“ 88.“ 88.. "~84 n98.” 38.0 H8.” 88.“ "8.” P8.“— 90.0 500.0. 08.0- «90.0. 08.0 8.0- 080.0 8.0 580.0 008.0 800 52.0.0- 080. 08.0 80 358.0 880.. 58.0 38.0 303.. 82.0.0 "08.“. RPSJ 3&8; 380.0 508.0 880.0 33.0.” 53.0. 80. N80. :30. 8.0 800 '80 0 8.0 0000.0- «~30 030.0 30.0. SE 358.” 2.8.0 "8.8.0 R8.“ n88.“ 32.0.“ "08.“ "020.“. E84 380.“ 388.0 800.“— 32.0.0 80 £80. 80 8.0 80. ~80 2.8.0 3'00 080 08.0 «~80 2.000. 8.0 28 R8.“— 84 8.0.0 88.. :08.“ 38.0 8M8; 3N8.”— 884 =8.“— n~kp0d C80.“ 3.30.” 080. N80. 80. «8.0. 8.0 030.0. 38.0 ~m~00 8.0 3.000. mn—00 98.0 350.0 0w> 2 D 30.0.0 "08.. 800.0 3000.0 890.. 2.30.0 P8.” 32.0.. 8. F0.“— 84 3000.0 82.0.0 509.. .330. 0:00. 80. 2.00 390.0 390.0 8000. «500 ~n_.00 N80 0050 0300 $0. 5w nun-.0.“ 303.0 8.. 3:0.“— nggd 8:0.“ H8.“ :80; "MM—.0.” 32.0.0 303.0 390.0 8%...“— 830. 020.0. 8.0 «2.0.0. 300 E00. N80. 0mm00. 0080. 0300 2.8.0. 8.0 3000. am: 8.“ :8.” "053.0 8.” 82.0.0 Eng.“— 3084 89.0.. 83.. 380.. "030.0 830.0 :00...“ N300. 505.0. 2.80 00.00. 8.0 8.0 33.0 38.0 300.0. ..800 005.0 3000 2.20. an: 8.0 38.0 "2.0.“. 38.0 28.0 830.“ 3.8.“ H08.“ 808.” 38.0 "8'0.“ 38.. 308.0 2.80 300.0- 80- ~80. 8.0 2.000. 88.0 300.0 8.0 0800. «08.0. 0800 38.0 8» 3304 22.04 3.30.. 22.0.. 32.0.“ ES.” 58.“ 390.. 820.“ H0034 "030.0 830.. ammN—J 05~w0 £80 0580 08.0 80. “~00 2.80 00.30 {.80- 300.0- “:0- 22.0. «~00; «mu pg: 3.82.. 9.9.0 :30 2840 0280 :08 «Uta aw u :0 850 3350 32.... . cw»... 3 .0850 032889.. 2! :28 .3 Si: no. u:- co..3....o 3 .385: no. :23 .995 no... 3 .3138 so»!!! .o.¢ .33 133 880.. 238.. 350.. 58.. 380.. 3.8.. "38.. ”NR... 008.0 003.0- 05.0 008.0- 080.0 ~08.0 ~§0 008.0 880.. $90.. 3000.. 308.. «080.. 208.. 8000.. 808.. 00.0.0- ~h~06- ~500- 0086 003.0 N086 0300- 008.0 "00.00.. 803.. 808.. use. 3000.. ”008.. "02.0.. 308.. 500.0 «08.0 008.0 38.0- 08.0 98.0 35.0 38.0 "0000.. 890.. "0000.. "008.. 8000.. 58.. $000.. 808.. 3.0- 008.0- 38.? 38.0 0000.0 3.0 003.0 008.0 2.20.. 303.. £00.. 890.. 39.0.. 808.. 803.. 308.. 00~0.0- 008.0- 336 005.0 0000.0 003.0- 0m~00 090.0 "020.. "300.8000. "00.00.. 830.. ”008.. £90.. 830.. 500.0- 003.0 0000.0- 003.0 0000.0- {8.0- n000.0- 008.0- 8000.. "00.3.. "00.00.. n38. "Nu—0.. 2.08.. nun-0.. 2.08.. 0000.0- 0p8.0 020.0- 38.0- 086 «5.0.0 080.0 800.0- §.. "008.. "300.. "58.. n38. "2.8.. £30.. n0~8.. ~08.0- 0.08.0- 0—00.0- «~86 0.00.0- 1.8.0 0000.0 58.0 890.. 3.00.. ”0000.. $03.. 30.0.. 2.08.. "0590.. 330.. 5009.0 090.0 080.0 386- 003.0 58.0 030.0 0.08.0- 830. "058.. 380.. "890.. 080.0- 00p0.0 0000.0- 3.0 03-. 50000.. n0-0.. 830.. £00.. 0000.0- m-0.0 080.0 pn~0.0 E... 830.. «003.. "00.00.. 2.030.. 080.0 008.0 050.0 0000.0 :50 $30.. 890.. "0000.. 308.. 0000.0- m0p0.0 050.0 0~k0.0 00> 8000.. n83. 303.. 3000.. 0000.0 900.0 h0~0.0 0~m...0- 5m 8000.. 3.00.. 330.. "009.. ~0~0.0 0~n0.0- '00—.0 ~0—b.0- 0m: 3000.. 850.. 8000.. "0900.. ~86 N306 0000.0 $3.? :0: 830.. 308.. "030.. n0p~0.. 00N0.0- $8.0- 0 003.0 3.. 3.000.. 803.. "02.0.. n-0_... 38.0. 008.0- 0009.0- 0.00... «mu mum... 9.08 2.000 500 om>00 3w00 03.00 30.000 8000 0800 05.2... .002: b .3260 03-80305 3 02. 32.252 :2... .995 080 .8. “Ella 0o 83...»: goal-...... .s.< 035 134 Table A.8. ngpgn§§§§§ own and cross price elasticities for food groups of urban household in West Java (without imposing homogeneity) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER -.54 .11 TUB -1.24 .21 FISH -.39 .15 MEP .14 -.48 -.27 .27 EGM 1.11 .18 .27 -.42 .22 VEG -.13 .32 .11 -.66 .17 SOYN -1.02 FRT -.31 SUG .49 .20 .28 .13 - .32 T03 .10 - .28 Table A.9. ngpgnggtgg own and cross price elasticities for food groups of urban household in WEst Java (imposing homo- geneity) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER -.50 .11 .48 TUB -1.25 .22 .24 FISH -.38 .14 MEP .14 -.46 .25 .30 36” -041 .22 -091 VEG .13 .31 -.65 “.72 “.16 SOYN -.99 -.11 FRT .11 .27 -.24 -.69 SUG .20 .26 .14 -.36 -.63 TOBACCO .10 .31 .14 -.29 135 Table A.10. ggmpgnsaggd own and cross price elasticities for food groups of urban household in West Java (imposing symmetry) (SUR) CODEOditY CER TUB FISH MEP EGM VEG SOYN FRT SUG CER “.49 .07 TUB -1023 021 033 024 FISH “.37 .13 .06 MEP “.45 .09 .16 .29 EC" “.45 .07 .08 .04 VEG “.72 SCYN “ .99 “.11 FRT “.28 SUG “.05 Table A.11. ggmpgnggggg own and cross price elasticities for food groups of urban household in West Java (block independe- nce between food, sugar, and tobacco) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER -.50 .11 TUB “1.24 .21 FISH “.38 .14 MEP .14 -.48 -.28 .27 EGM 1.08 .18 .26 -.42 .20 V36 .13 .32 -.11 -.68 SOYN -1.02 FRT .27 -.26 SUG .00 TOBACCO -.26 136 Table A.12. W own and cross price elasticities for food groups of urban household in West Java (without imposing homogeneity) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER “.73 .07 TUB “1.24 .21 FISH “.49 .05 MEP' .12 “.65 “.34 .17 EC" 082 008 013 -051 .16 V36 “.13 .32 .12 “.66 .17 SOYN “1.02 FRT “.31 SUG .27 .12 .18 .06 “ .38 T03 .04 “.36 Table 13.13. W own and cross price elasticities for food groups of urban household in West Java (imposing homogeneity) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER “.69 .00 .28 TUB -1.25 .22 FISH . “.48 .04 MEP 011 -063 018 .20 EC" .31 “.50 .16 “.94 VEG .13 .31 -.66 SOYN -.99 FRT .11 .27 -.24 -.69 SUG .13 .17 .08 -.41 -.64 TOBACCO .03 .25 ' .08 -.37 Table A. 14 . for symmetry) ( SUR) 137 W own and cross price elasticities food groups of urban household in West Java (imposing Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG CER -.67 .05 TUB -1.23 .21 .33 .24 FISH -.47 .04 .03 MEP -.55 .02 .06 .25 EGM -.54 .00 .01 .01 VEG -.72 SOYN -.99 -.14 FRT -.28 SUG -.07 Table A.15. W own and cross price elasticities for food groups of urban household in West Java (assuming block independence food sugar and tobacco) Commodity CER TUB FISH MEP EGM VEG SOYN FRT SUG TOB CER TUB “.69 “1.24 .21 FISH HEP EGH -.48 .11 .78 .08 VEG SOYN .13 FRT SUG TOBACCO .00 .04 “.65 “.35 .17 .12 “.50 .14 .32 “.68 -1002 .27 “.26 -003 ‘034 138 wwmn .mtvcmcu “ notaom 3.: S n. .3 Ed 23 8a 3 S S. a... o... ...S “2 3.333 e E 3.0 8.... o 2 v ...N an an I: o.~ 3:. 2H s: .5335 and Ed 38 o S on can a..." as . 3.... 2.. 55:. Ego n . 35 o 3 on on 3; so a...» an 22 :50 a 3. .66 N _ .o .... 8:. S S 3” 9: 2. ...2. an 25.2 ...: .5 Ed was o _ .o o a” 3 an 2 ..2 3. 9.2. 2.. 25> Z 3 pod _ ..o on .2. he an a ...: ca 9:. man 3229. an 3 Mod 25 . 83 «on ad on on E." 3 he. .3. so...» can a 3 86 is an 2N a... on a ...8 3 «.2. N2. 223 a . . 3339. 32.6 H...” 3. so .65 35 o .2 .3 an 3 «.2 3 “.3 3... .520 a: 2.: a... 9.: us. a: as. «E as. us. a u .8 ..x :05 . 4 _ d .d 3 I. .d m m m m. m. mm, m m. w m m. m on. m ... m m. .o m .... a m. ...... .. .. p v m... m. m. n. m m. u. m ..m" D m m 3 p v 1 s o .... m. . A p Acowutoa mea_um aces Lmav macsu poor FMUWQOLB co mmapm> m>wowtuzz .H.m m_amk 139 .Pmmcoccs .H.u .mcu nal‘h caution: .u« :0. .ou . 33.2..- an Chen. 6...: .hn cone—e. ces-:2: .On kiln.— elofi .n— cassen enough .0. 3.33. .on sous-h cocci-:1 .on «wand-bro» allnue-.£oued..n. 113.. .n «:22: .3323» .3 . ac:- ccuchZax ..n ...-:3... a)... .3 an: .o coon—om «.2315 ..n 33: nal: .2 scan. 1):. .0. can: 3:23.... .n 3.9:..— .ucI-nau .nn has“ enough 0.2. ...u dun-non :0 .a :3: ...-all.” .n =3: :23...” .nn m us:- eu-rvcoh :2. .nu .n roux-I1- .c £004 ...-.5»: 8:06 .- . . u eugHQOCQuouel insane 8 .32 $3285 <5; . 140 .. . . ....,/ $5.... . g 2::- 33.31! ll...- -- ‘fiwcnn..=» .. VI... , a. Lem... / 1W (“Il‘l‘lflx‘lIWTIE+I w M1 L: Li (,Hl‘ ‘11} M 3 1293 00080 1468