GEORG-AUGUST UNIVERSITAET GOETTINGEN

AN ALMOST IDEAL DEMAND

SYSTEM FOR FOOD
BASED ON CROSS SECTION DATA: RURAL
AND URBAN EAST JAVA, INDONESIA

Suharno

Göttingen, im September 2010

AN ALMOST IDEAL DEMAND SYSTEM FOR FOOD
BASED ON CROSS SECTION DATA:
RURAL AND URBAN EAST JAVA, INDONESIA

Dissertation
zur Erlangung des Doktorgrades
der Fakultät für Agrarwissenschaften
der Georg-August-Universität Göttingen

vorgelegt von
SUHARNO
geboren in Rembang Indonesien

Göttingen, im Mai 2002

D7
1. Referent : Prof. Dr. M. Leserer
2. Korreferent: Prof. Dr. A. Spiller
Tag der mündlichen Prüfung: 04.07.2002
 ABSTRACT

1. This is a micro-data based study of demand for food in the framework of a static, utility
maximizing, and partial model that enables the provision of knowledge on the
interrelatedness among the competing commodity groups in a complete demand system.
2. The dynamics which took place in the economy of contemporary Indonesia has created
an urgent need for policy makers and scholars of food and agriculture sector of this
country to have a knowledge on the spending behavior of the households in their
response on changing consumption determinants like income, relative prices, the
introduction of new brands in manufactured foods, an intensifying advertisement,
changing mode of retailing, etc., as well as the changes in the demography of
households themselves. The need is reinforced, as Indonesia after enjoying two decades
of economic booming was hit by a devastating economic crisis that broke out in July
1997, the ramification of which prevails until the time of study. The consequences of
this crisis are manifold. Economically speaking, the crisis has (i) forced Indonesia to
approach a market system that among others, liberalizes the previously intervened food
market, (ii) set the purchasing power of the average Indonesian back to the level of ten
years before (iii) also changed the prices relatively. Politically, the Indonesian
government is now facing an era of decentralization. These factors in combination
might change the consumption structure of different household groups in Indonesia.
Additionally, it places an urgent need to conduct a study also with local specific
perspective of consumption behavior.
3. Until today, the existing knowledge is deficient, because previous studies are limited to
the estimation of single equation model based on an aggregated data. Due to the
importance of the household as the decisive unit in consumption, and due to an
increasing accessibility of micro data, this study used a dis- aggregate micro data set
from the province of East Java, Indonesia.
4. Given that background, the objective of this study is firstly to find demand parameters
for food groups under investigation, based on which one can analyze the effects of
expenditure and price changes on demand of eleven food groups for different income
groups in the province of East Java, Indonesia. Secondly, to demonstrates the use of the
 study results for real policy questions about the food and agricultural sector. Thirdly, to
evaluate the specific welfare effects of selected price policies for different income
groups.
5. The brief exposition of the republic of Indonesia in a historical perspective indicates
that Indonesia is an economy with heavy state intervention in the past and departing
from this basic model is a matter of political pragmatism. Changing the economic
structure reduced the role of agricultural sector in terms of GDP contribution, but it is
still important for food provision and employment. Increasing income per capita per
year in the country reduced slightly percentage of expenditure on food. Rice
expenditure has a high share of total food expenditure in all household groups.
Therefore, food policy in Indonesia has dominantly centered on rice.
6. This study employed the cross sectional household consumption/expenditure micro data
set from the so called SUSENAS (the National Socio -Economic Survey), for the
periods 1990, 1993, 1996 and 1999 representing the province of East Java, Indonesia.
The consumption and income module of the SUSENAS survey covers all household
expenditures during a week of enumeration with full specification of commodities.
Listed in the questionnaires are 231 consumption items, for which data on quantities
and values were gathered. The data set of each survey periods is collected from 5692
households (1990), 7638 households (1993), 8015 Households (1996), and 8552
households (1999) in urban and rural areas. The central Bureau of Statistics applied the
three-stage stratified sampling for the SUSENAS. For food consumption the survey
reference period was one week prior to the enumeration of data.
7. The theoretical framework of this study is the neo-classical consumer economics.
Theory and the related methods are presented in order to justify the model used in this
study. Some theoretical, empirical and pragmatical considerations have brought us to
the decision to use the linearized approximation of an almost ideal demand system
(LA/AIDS) model. It satisfies the axioms of choice, aggregates perfectly over
consumers, has a functional form, which is consistent with household budget data, and
simple to estimate and test the true restrictions of demand theory. It also combines the
best of theoretical features of both Rotterdam and translog models. When Stone‘s index
is used in the model it is termed as a linear approximation of almost ideal demand
 system (LA/AIDS). The use of the concept of compensating variation suggests that
results of demand estimation contribute well to the analysis of policy. Compensating
variation is the compensating payment (amount of money) that leaves the consumer as
well of as before the economic change. It may be positive or negative. It is positive, if
the economic change makes consumer worse off, and negative, if the economic change
brings betterment to the consumer.
8. Because compensating variation is money metric, its expression is dependent on an
absolute expression in term of country‘s currency unit. This is less comparable. To
avoid this, one can transform it in a relative term by using for example, price index,
which is metric independent. Based on that, Fischer Ideal Price Index was used to
approximate the welfare change. Fischer Ideal Price Index is a geometric means of
p1
Laspeyres- (PL) price index , PL =  wi0 ), and the Paasche (PP) price index, Pp =
i P0

 p 
1 /   w1i 0  . It is expressed algebraically as = PL .PP . It represents a changing
 i P1 

purchasing power as an approximation of welfare change.

9. The estimated equations for the LA/AIDS are summarized in table 6.2 to 6.9. For all the
periods of surveys, covering urban and rural areas, there are 88 equations for the
LA/AIDS. Eighty equations out of these 88 were estimated directly using SAS program
the 6.12 edition, by applying the iterative seemingly unrelated regression (ITSUR)
estimation procedure. The parameter estimates for the rest of 8 equations were
recovered by using adding-up principle. In these models, the variation of budget shares
of eleven food groups in the study areas are determined by (the own- and cross) prices,
income level which is approximated by the weekly household‘s total expenditure on
food, the income group of the households, and the household size accommodating the
rest of demographical characteristics of the households. In total, 220 parameters in each
of equation are resulted directly or indirectly from this estimation. Table 6.10
summarizes the estimation performance by presenting the number of statistically
significant estimates out of 170 parameters in each equation that directly estimated in
this study. As a matter of statistics, the worse performance of the estimation is
represented by the one that give 55 per cent statistically significant estimates (table
code 6.2: Urban90). The best estimation performance is exhibited by the one that
 brought 78 per cent statistically significant estimates (table code 6.3: Rural90). The
facts, that more than the half of parameter estimates in each equation system is
statistically significant may be the basis to claim, that the model specification is
appropriate. Also, direct observation on the results of estimation indicates that majority
of parameter estimates are large relative to their standard errors. These deliver some
degree of confidence to say that the estimates are reliable. These in all suggest that our
hypothesis, as explicitly expressed in the LA/AIDS model, is supported by the data.
That is to say, that food demands in the study areas are responsive to prices, total food
expenditure level, income groups and the household size as measured from survey data.
10. The asymptotic likelihood ratio test on demand restrictions indicates that the result of
the test is consistent with the previous common findings by other authors. That is, the
homogeneity and symmetry restrictions were in most cases violated by the data.
However, it does not necessarily mean, that the theory is wrong; it may be rather the
case, that the data and model combined do not support the theory either because of data
property, and/or model specification.
11. The signs of the AIDS parameters deliver information on the nature of the demand for
food commodities. So, by inspection one can infer, those with negative expenditure
parameters β i  0 are income inelastic, and those with positive parameters, β i  0 are
income elastic. Observation on the AIDS estimates indicated that rice is in all cases
income inelastic. Other commodities exhibited a mix performance depending on the
areas and survey periods. Fish, meat, tobaccos and betel, and prepared food exhibited a
generality of being income elastic. Other findings that support the intuition is that all
food groups showed a negative own price elasticities. Most of commodity groups under
investigation, with exception of Eggs and Milks, are own price inelastic. The fact, that
the compensated own price elasticities are different clearly from those of the ordinary
own price elasticities indicated that there is a demand effects in each of price change of
the commodities groups being analyzed. Other food groups are responsive on the
change of rice price. The reverse is not the case. In general, cross price relationship
among the food groups are less influential. The inclusion of household size in the
AIDS model for food is justified by the fact, that most of the parameter estimates
representing household size were statistically significant. So for the majority of food
 items it holds that an additional of household‘s member will cause some household
expenditure to increase and others to decline to balance the household size variable. As
the number of household member increases, households reduce their consumption of
tobacco, fruits and vegetables, prepared foods, and some time, fish and meat. These
reductions are made in order to increase the consumption of other categories with
positive household size elasticities mainly rice, non-rice staple, and edible oil. The
increase of household size definitely associated with the decline in the food quality
consumed by the households. The consumption of cheap carbohydrate-rich food is
mainly the strategy taken by households having a large membership.
12. The estimated demand parameters provide a complete and consistent framework for
evaluating impacts of any government policy. The combination of direct rice- and
indirect tobacco pricing policies has been used in this study to demonstrate the
usefulness of the results of this study. The price of rice has an important impact on
private household‘s spending pattern, because of its important influence on the
household‘s budget. The policy exercise conducted in this study suggest, that
liberalizing the market of rice will make households of all income groups better-off, and
a combination of it with a tobacco-taxing (indirect pricing) will increase government
revenue without harming so much the poor households.
 ZUSAMMENFASSUNG .

1. Diese Arbeit ist eine Nachfragestudie, die auf den Mikro-Daten des Verbrauches für
Lebensmittel und im Rahmen einer statisch, Nutzenmaximierend, und partielle Modell
bearbeitet ist. Die Studie ermöglicht eine Bereitstellung von Information über das
Zusammenhang zwischen den konkurrierenden Warengruppen in einer vollständigen
Nachfrage System.
2. Derzeitige Wirtschaftentwicklung, die unter anderen wegen der im Juli 1997
ausgebrochene Krise ausgeprägt ist, hat eine dringende Notwendigkeit für die Politik
und Wissenschaftler der Lebensmittel und Landwirtschaft dieses Landes um eine
Information auf die Konsumsverhalten der Haushalte in ihrer Reaktionen auf die
Änderung der Verbrauch determinierenden Faktoren wie Einkommen, Preisverhältnis,
Einführung neuer Marken in Lebensmittelprodukte, Intensivierung der Anzeigen,
Änderung im Modus des Einzelhandels, usw., sowie die Änderungen in
demographische Faktor der Haushalte. Der Bedarf nach dieser Informationen sind um
so großer, weil es nach im Juli 1997 ausgebrochene Wirtschaftkrise ein tief greifende
Strukturwandel gibt, die vielfältige Folge mitgebracht hat. Diese Folge sind unter
anderen: (i) Indonesien ist daran gezwungen, die Wirtschaft, einschließlicher
Lebensmittelmarkt sich an einem Markt System zu orientieren; (ii) die durchschnittliche
Kaufkraft des Volkes ist zu der Ebene der vor zehn Jahre zurück gegangen; (iii) Der
Preisverhältnis verändert sich. (iv) Politisch gesehen, steht die indonesischen Regierung
derzeit vor einer Ära der Dezentralisierung. Diese Faktoren konnte es dazu führen, der
sich Struktur der verschiedenen Haushaltsgruppen in Indonesien zu ändern. Dazu ist es
Notwendig, eine Studie mit den lokalen spezifische Sicht des Verbrauchs verhaltens
durchzuführen.
3. Zu den Zeitpunkt ist die existierende Information unzulänglich, weil die vorherigen
Studien wenn überhaupt da sind, lediglich nur auf Einzel Gleichung schätzende Modell
begrenzt sind, und sie sind meisten basiert auf einen argregierten Datei. Auf Grund der
Wichtigkeit des Haushalts als die entscheidende Einheit in Verbrauch und auf Grund
einer wachsenden Erreichbarkeit von Makrodaten, hat diese Studie
 einen disaggregierten Haushalt Mikrodatensatz von der Provinz Ost Java, Indonesien
benutzt.
4. Die Studie hat folgende Ziele: Erstens, Nachfragenparameter für die untersuchten
Lebensmittelgruppen zu finden, damit man die Wirkung einer Preisänderungen auf die
Nachfrage der Lebensmittelgruppen für verschiedene Einkommengruppen in der
Provinz Ost Java, Indonesien analysieren kann. Zweitens, um zu zeigen, wie man
die Studienergebnisse für real politische Grundsatzfragen um die Lebensmittel und die
Landwirtschaft nutzen kann. Drittens, um die spezifischen Wohlfahrtwirkungen der
ausgewählten Preispolitik für verschiedene Einkommengruppen zu bewerten.
5. Ein historisch perspektive Überblick über die Republik von Indonesien zeigt an, dass
Indonesien eine Wirtschaft mit schwerer staatlicher Einmischung in der Vergangenheit
ist, und eine Änderungen von diesem grundlegenden Modell eine Sache des
politischen Pragmatismus ist. Die ändernde Wirtschaftstruktur des Landes hat
dazu zuführen, das die Rolle der Landwirtschaft im Brutto Inland Produkt (BIP)
Beitrags verringert ist, obwohl diese noch wichtig ist für
die Lebensmittelsevorkehrung und Anstellung. Steigende pro Kopfseinkommen pro
Jahr auf dem Land hat nur geringe Minderung des Verbrauches auf Nahrungsmitteln zu
Folge. Reiskonsum hat einen hohen Anteil der gesamter Nahrungsmittelausgaben in
allen Haushaltgruppen. Daher hat sich Nahrungsmittelpolitik in Indonesien
vorherrschend noch auf Reis konzentriert.
6. Dieses Studie hat den disaggregierten Mikrodatensatz von Haushaltsausgaben
bearbeitet. Dieser Datensatz ist von so genannten SUSENAS (die nationalen Sozial—
Wirtschaftliche Datenerhebung), für die Perioden 1990 1993, 1996 und 1999
von Ost Java Provinz Indonesien eingestellt. Der Ausgaben und der Einkommenmodul
von dem SUSENAS Verhebung bedecken alle Haushaltausgaben in einer Woche der
Aufzählung mit voller Spezifikation von Waren. Aufgeführt in den Umfragen sind 231
Verbrauchwaren, die Daten auf Quantitäten und Werte gesammelt wurden. Der
Datensatz für jede Verhebungsperiode ist von 5692 Haushalten (1990), 7638
Haushalten (1993), 8015 Haushalte (1996), und 8552 Haushalte (1999) in städtischen
und ländlichen Gebieten gesammelt. Die zentrale Behörde der Statistik hat die
 dreistufige stratifizierte Probe für den SUSENAS angewandt. Für Verbrauchsdaten der
Nahrungmittels war das Zeitreferenz eine Woche vor der Aufzählung von Daten.
7. Die theoretische Grundlage dieser Studie ist die Neonklassische Verbraucherwirtschaft.
Theorie und die verwandten Methoden sind präsentiert, um das in dieser
Studie gebrauchten Modell zu rechtfertigen. Wir haben aufgrund einige theoretisch,
empirisch und pragmatische Berücksichtigungen die Entscheidung getroffen, die
linearen Annäherung von der nahezu idealen Nachfragensystem ( (LA/AIDS) Modell
zu benutzen. Es befriedigt die Axiome der Wahl, argregiert perfekt über die
Verbrauchern, hat eine praktische Form, die verträglich mit Haushalthaushaltdaten ist,
ist einfach zu schätzen, und kann prüfen die wahren Einschränkungen der
Nachfragentheorie. Es kombiniert auch den Beste von theoretischen Eigenschaften von
sowohl Rotterdam als auch Translog Modelle. Wenn man der Preisindex von Stone im
Modell anwendet, ist das Modell als eine Lineare Annäherung der Nahezu idealer
Nachfrage System (LA/AIDS) genannt. Der Gebrauch des Compensating Variation
(CV) Konzeptes schlägt vor, dass die Ergebnisse der Nachfragenschätzung gut zur
politische Analyse beitragen kann. Das CV ist die Entschädigungszahlung (Betrag
des Geld) der den Verbraucher ebenso wohl als vor der wirtschaftlichen Änderung
verlässt. Es mag positiv oder negativ sein. Es ist positiv, wenn die wirtschaftliche
Änderung dem Verbraucher schlechter drauf macht, und Negativ, wenn die
wirtschaftliche Änderung dem Verbraucher Verbesserung bringt.
8. Da das CV Geld metrisch ist, ist sein Ausdruck abhängig auf einem absoluten Wert der
Währung des Landes. Dies ist weniger vergleichbar. Um dies zu vermeiden, kann es in
einem relativen Begriff durch Gebrauch zum Beispiel, eines Preisindexes, umgestalten
werden. Dadurch ist es metrisch unabhängig. Auf diesen Grund, wurde Fischer Idealer
Preisindex in dieser Studie benutzt, der Wohlfahrtsänderung anzunähern. Fischer
Idealer Preisindex ist ein geometrisches Mittel des Laspeyres- (PL) Preisindex, PL =
p1  p 
 wi0 , und der Paasche (PP) Preisindex Pp = 1 /   w1i 0  . Es ist algebraisch als
i P0  i P1 

PL .PP .ausgedrückt. Es vertritt eine Änderungskaufkraft, die als eine Annäherung der
Wohlfahrtsänderung gilt.
9. Die geschätzten Gleichungen für das LA/AIDS sind in Tabelle 6. 2 zu 6. 9
zusammengefasst. Für die ganzen Perioden von der Verhebungen, die städtische und
ländliche Gebiete bedecken, gibt es 88 Gleichungen für das LA/AIDS. Achtzig
Gleichungen aus diesen 88 wurden direkt durch das SAS Program (die 6,12 Ausgabe)
geschätzt, durch die Verwendung der iterativen scheinbar nicht verwandten Regression
(ITSUR) Schätzungsverfahren. Die Parameterschätzungen für den Rest von 8
Gleichungen wurden von Gebrauch der Prinzip summierung (add up principle)
wiedererlangt. In diesen Modellen wird die Veränderung der Budgetanteilen von elf
Nahrungsmittelsgruppen in den Studiegebieten von den folgenden Faktoren bestimmt:
Preise (das eigene- und kreuzt Preis), Einkommensnivue, die vom totalen Ausgaben der
wöchentlichen Budget auf Nahrungsmitteln angenähert werden, die Einkommengruppe
von den Haushalten, und der Haushaltgröße, die den Rest des demographische
Merkmale vertritt. Insgesamt sind 220 Parameter in jeder Gleichung, die direkt oder
indirekt von dieser Schätzung resultiert. Tabelle 6.10 fasst die Schätzungsleistung durch
die Vorlage der Anzahl der statistisch signifikante Schätzungen von 170 Parametern der
einzelnen Gleichungen zusammen, die direkt in dieser Studie geschätzt wurden.
Statistik gesehen, wird die schlechter Leistung der Schätzung von einer vertreten, die
55 Prozent statistisch signifikante Schätzungen gibt (Tabelle 6.2: Urban90). Die beste
Schätzungsleistung wird von einer vertreten, die 78 Prozent statistisch signifikante
Schätzungen gibt (Tabelle 6.3: Rural90). Die Tatsachen, dass mehr als die Hälfte von
Parameterschätzungen in jedem Gleichungssystem statistisch signifikant sind, gibt
einen Grund zu beanspruchen, dass die Modellspezifikation passend ist. Auch direkte
Beobachtung auf den Ergebnissen der Schätzung zeigt an, dass Mehrheit von
Parameterschätzungen großer sind, im Vergleich mit ihren Standard Fehlern. Die liefern
ein gewisses Maß an Vertrauen zu sagen, daß die Schätzungen zuverlässig sind. Diese
in allen vorschlagen, daß unsere Hypothese, wie ausdrücklich in der LA/AIDS Modell,
von der Daten unterstuzt wird. Das ist zu sagen, dass die Nachfrage nach
Nahrungsmittel in den Studiensgebieten ansprechend ist zu Preisen, totale Ausgaben für
Nahrungsmitteln, Einkommengruppen und die Haushaltgröße.
10. Die asymptotische Likelihood Ratio Test auf die Nachfrage Ristriktionen zeigt an, dass
das Ergebnis der Prüfung im Einklang mit der früheren algemeinen Ergebnisse von
 anderen Autoren steht. Das ist, der Homogenität und der Symmetrie Restriktionen in
den meisten Fällen von der Daten übertreten worden sind. Es bedeutet aber nicht
unbedingt, dass die Theorie falsch ist. Es kann der Fall sein, dass die Daten und Modell
nicht die Theorie unterstutzen kann entweder wegen der Dateneigenschaft, und/oder
Modell Spezifikation.
11. Die Zeichen von den AIDS Parametern liefern Informationen über die Eigenschaften
der Nachfrage nach Nahrungmitell. Man kann durch Besichtigung folgern, dass Waren
mit negativen Verbrauchparameter ( β i  0 a) Einkommen unelastisch sind, und

diejenige, die mit positiven Parametern ( β i  0 , Einkommen elastisch sind.

Beobachtung auf den AIDS Schätzungen hat angezeigt, dass Reis in alle Fälle
einkommen unelastisch ist. Andere Waren haben eine Mischungsleistung ausgestellt,
die von den Gebieten und Verhebungsperioden abhängen. Fisch, Fleisch, Tabake und
Betel, und vorbereitete Speise haben eine Allgemeinheit ausgestellt, einkommen
elastisch zu sein. Andere Ergebnisse, die im Einklang mit der Intuition haben, sind die
Ergibnisse die angezeigt haben, dass alle Nahrungsmittelgruppen eine negative
Preiselastizitäten besitzen. Meisten von der untersuchten Waregruppe, mit Ausnahme
von Eiern und Milch, sind Eigenpreis unelastisch. Die Tatsache, dass die entschädigten
eigenen Preiselastizitäten (compensated ownprice elasticity) deutlich verschieden sind
von denen der gewöhnlichen eigenen Preiseselastizitäten hat angezeigt, dass es
Nachfragenwirkungen in jeder Preisesänderung der Warengruppe gibt. Andere
Warengruppen sind ansprechend (responsive) auf der Änderung des Reisespreises. Das
Gegenteil ist nicht der Fall. In Allgemein ist kreuze Preisbeziehung unter den
Speisengruppen weniger einflussreich ist. Die Einbeziehung der Haushaltsgröße in den
ganzen AIDS Model fuer Nahrungsmittel wird gerechtfertigt von der Tatsache, dass die
meisten Parameterschätzungen, die Haushaltgröße vertreten, statistisch bedeutsam
(significant) waren. Deswegen, ist es fest gestelt, das die Ausgaben fuer die Mehrheit
der Nahrungsmittel von der Anzahl von Haushaltsmitglieder beeinflusst werden. Ein
zusätzliches Haushaltmitglied kann verursachen, dass einige Haushaltausgaben steigen
fuer das eine oder mindern für das anderen, um auszugleichen. Als die Anzahl von
Haushaltmitglied zunimmt dan verringern der Verbrauch des Tabaks, Früchte und
Gemüse, vorbereitete Speisen, und Fisch und Fleisch. Diese Reduktionen sind gemacht,
 um der Verbrauch von anderen Nahrungskategorien mit positiven elastizitäten,
hauptsächlich Reis, Nicht-Reisstoffen, und essbares Öl. Die Zunahme der
Haushaltsgröße ist mit der Abnahme derjenigen Speisenqualität verbunden. Der
Verbrauch der billiger Kohlenhydrats-reicher Speise ist hauptsächlich eine Strategie,
die von Haushalten mit große Mitgliedszahl genommen wird.
12. Die geschätzten Nachfragenparameter versorgen einen vollständigen und
gleichmäßigen Rahmen für Bewertenschläge irgendeiner Regierungspolitik. Die
Kombination des direkten Reises- und indirekte Tabak Preispolitikes ist in diesem
Studie benutzt worden, die Nützlichkeit der Ergebnisse dieses Studie vorzuführen. Der
Preis des Reises hat einen wichtigen Auswirkung auf das Ausgabenmuster von
privatem Haushalt; weil Reis ein wichtigen Einflusses auf dem Haushaltausgaben hat.
Die Preisberechnung, die in diesem Studie geleitet wird, schlägt vor, dass die
Liberalisierung des Reismarktes eine Wohlfahrtsverbesserung an aller
Einkommensgruppen macht. Wenn der Verbrauch von Tabak besteuert wird (indirekter
Preisberechnung), wird dann Regierungseinkünfte steigen, ohne das Schaden von so
viel armen Haushalten.
 AN ALMOST IDEAL DEMAND SYSTEM FOR FOOD
BASED ON CROSS SECTION DATA:
RURAL AND URBAN EAST JAVA, INDONESIA

TABLE OF CONTENTS Page

CHAPTER I. INTRODUCTION .................................................................................. 1

1.1 Structural Adjustment and Demand Analysis .................................................... 1
1.2 Need for a New Study on Household Reaction ................................................. 3
1.3 Cross-sectional Demand Analysis .................................................................... 4
1.4 Food Demand Study and Policy Analysis .......................................................... 5
1.5 The Need of Evaluation of Alternative Price Policies
on Different Income Groups .............................................................................. 6
1.6 The Scope ........................................................................................................... 8
1.7 Objectives of the Study ...................................................................................... 8
1.8 Organisation of the Study .................................................................................. 9

CHAPTER II. REVIEW OF ECONOMIC SETTING ................................................. 10

2.1 General Setting ................................................................................................... 10
2.2 The Economic System ....................................................................................... 11
2.3 The First Two Decades ...................................................................................... 12
2.4 1966 – 1980: Period of Recovery ...................................................................... 13
2.5 1980 – 1996 Stabilisation and Growth ............................................................... 13
2.6 The Debt ............................................................................................................. 14
2.7 Economic Structure ............................................................................................ 16
2.8 Food and Agricultural Policy ............................................................................. 17
2.9 Consumption Pattern .......................................................................................... 21
2.10Food Security and Distribution .......................................................................... 25

CHAPTER III. THE DATA SET: SUSENAS DATA ................................................. 28

3.1 SUSENAS Data ................................................................................................. 28
3.2 Data Collection Methods ................................................................................... 30
 3.3 Sampling Methods ............................................................................................. 31
3.4 The Concepts and Definitions ............................................................................ 32
3.4.1 Household ............................................................................................. 32
3.4.2 Special Household .................................................................................. 32
3.4.3 Household Member ................................................................................ 33
3.4.4 Consumption Expenditure ...................................................................... 33
3.4.5 Household Income ................................................................................. 33

CHAPTER IV. THEORETICAL FRAMEWORK ...................................................... 34

4.1 Neo-classical Theory of Consumer Behaviour .................................................. 35
4.1.1 Basic Concepts ....................................................................................... 36
a. Commodity Set .................................................................................. 36
b. Preference Axioms ............................................................................ 36
c. Utility Function ................................................................................. 37
4.2 Setting-Up a Consumer Demand Function ........................................................ 37
4.3 Properties of Marshallian Demand Function .................................................... 38
4.3.1 Adding Up .............................................................................................. 38
4.3.2 Homogeneity .......................................................................................... 39
4.3.3 Negativity ............................................................................................... 39
4.3.4 Symmetry ............................................................................................... 40
4.4 Cost Minimization and Hicksian Demand Function .......................................... 42
4.5 Special Utility Structures ................................................................................... 43
4.5.1 Separability ............................................................................................ 44
4.5.2 Aggregation Over Consumers ................................................................. 44
4.6 Empirical Demand Systems ............................................................................... 45
4.6.1 Linear Expenditure System (LES) ......................................................... 46
4.6.2 Indirect Addilog Model (IAD) ............................................................... 48
4.6.3 Rotterdam Model ................................................................................... 49
4.6.4 Transcendental Logarithmic Demand System ....................................... 52
4.6.5 Almost Ideal Demand System (AIDS) ................................................... 53
4.7 Welfare Analysis ................................................................................................ 55
4.7.1 Welfare Measures ................................................................................... 56
 4.7.2 Marschallian Consumer Surplus ............................................................ 57
4.7.3 Compensating and Equivalent Variations .............................................. 58
 Compensating Variation .................................................................... 58
 Equivalent Variation .......................................................................... 58

CHAPTER V. METHODS AND STATISTICAL PROCEDURES ............................. 60

5.1 Data Handling .................................................................................................... 60
5.1.1 Commodities Grouping .......................................................................... 60
5.1.2 Price of Commodity ............................................................................... 62
5.1.3 Price of Grouped Commodity ................................................................ 64
5.2 The Problem of Missing Observation ................................................................ 65
5.2.1 The Problem of Zero Price ..................................................................... 65
5.2.2 The Zero Expenditure Problem .............................................................. 66
5.3 Classification of Households in Income Groups ................................................ 67
5.4 Modelling Demand System ................................................................................ 68
5.4.1 Specification of System Model ............................................................... 68
5.4.2 Incorporating Income Group into the Linear AIDS ............................... 70
5.5 Estimation Procedures ........................................................................................ 73
5.6 Computation of Elasticities of Demand ............................................................ 74
5.7 Test of Restrictions ............................................................................................ 76
5.8 Welfare Analysis ................................................................................................ 78
 Compensating Variation ............................................................................. 78
 Compensating Variation Measurement of Proposed Price Change ............ 78

CHAPTER VI. ESTIMATION RESULTS AND DISCUSSION ................................ 81

6.1 The Budget Share ............................................................................................... 81
6.2 The Model‘s Performance .................................................................................. 83
6.3 The Results of the Restrictions Test .................................................................. 93
6.4 Price Elasticities ................................................................................................. 95
6.4.1 Own Price Elasticities ............................................................................. 96
6.4.2 Cross Price Elasticities ........................................................................... 98
 6.5 Expenditure Elasticities ...................................................................................... 107
 Variation Over Time ................................................................................... 109
 Variation Across Income Groups ................................................................ 109
6.6 Family Size Elasticity ........................................................................................ 112

CHAPTER 7. POLICY EXERCISE ............................................................................. 115

7.1 Background ........................................................................................................ 115
7.2 The Scenarios ..................................................................................................... 116
7.3 Welfare Measurement Revisited ........................................................................ 118
7.4 The Welfare Effects of the Scenarios ................................................................ 120
 Concluding Words ....................................................................................... 122
 Notice .......................................................................................................... 122

CHAPTER 8. CONCLUSIONS .................................................................................... 123

REFERENCES .............................................................................................................. 125

LIST OF TABLES Page

Table 2.1 GOI Foreign and Domestic Debt, 1995-2000 (USD Billions) ...................... 15
Table 2.2 Indonesia: Net Capital Inflows (USD Billions) ............................................. 15
Table 2.3 Selected Macro Indicators of Indonesian Economy ....................................... 17
Table 2.4 Percentage of Average Monthly per Capita Expenditure
on Food and Non-food, Indonesia .................................................................. 21
Table 2.5 Monthly Average Budget Share of Food by Commodity Groups
in Indonesia, 1990 – 1999 .............................................................................. 22
Table 2.6 Monthly Average Share of Food Expenditure
by Commodity Groups in Rural and Urban Indonesia 1990 – 1996 .............. 23
Table 2.7 Average Daily per Capita Consumption of Calorie and Protein
by Commodity Group, 1990, 1993, 1996....................................................... 24
Table 2.8 Price Increases of the Nine Essential Commodities July 1997-
April 1998 ...................................................................................................... 25
Table 3.1 Numbers of Households Co-operating in SUSENAS
in East Java, Indonesia from 1990 – 1996 ..................................................... 30
Table 5.1 Variable Description and A-priory-Hypotheses ............................................. 72
Table 6.1 Shares of Food Groups (mean value) to Weekly Total Food Expenditure,
East Java, Indonesia ....................................................................................... 82
Table 6.2 Parameter Estimates for the LA/AIDS Model Based on the 1990 -
SUSENAS Micro Data: Urban East Java, Indonesia ....................................... 84
Table 6.3 Parameter Estimates for the LA/AIDS Model Based on the 1990 -
SUSENAS Micro Data: Rural East Java, Indonesia ..................................... . 85
Table 6.4 Parameter Estimates for the LA/AIDS Model Based on the SUSENAS
Micro Data: 1993, Urban East Java, Indonesia ................................................ 86
Table 6.5 Parameter Estimates for the LA/AIDS Model Based on the 1993 -
SUSENAS Micro Data: Rural East Java, Indonesia ........................................ 87
Table 6.6 Parameter Estimates for the LA/AIDS Model Based on the 1996 -
SUSENAS Micro Data: Urban East Java, Indonesia ....................................... 88
Table 6.7 Parameter Estimates for the LA/AIDS Model Based on the 1996 -
SUSENAS Micro Data: Rural East Java, Indonesia ....................................... 89
Table 6.8 Parameter Estimates for the LA/AIDS Model Based on the 1999 -
SUSENAS Micro Data: Urban East Java, Indonesia ...................................... 90
Table 6.9 Parameter Estimates for the LA/AIDS Model Based on the 1999 -
SUSENAS Micro Data: Rural East Java, Indonesia ....................................... 91
Table 6.10 The Results Summary: Number of statistically significant Estimates
Summarized from the estimation results with LA/AIDS Model:
SUSENAS Micro Data East Java, Indonesia ................................................. 92
Table 6.11 Test Results .................................................................................................... 93
Table 6.12 Critical Values of 2 ...................................................................................... 94
Table 6.13 Ordinary Own Price Elasticities Based on the SUSENAS Micro Data
1990, 1993, 1996 and 1999: Urban East Java, Indonesia .............................. 95
Table 6.14 Ordinary Own Price Elasticities Based on the SUSENAS Micro Data
1990, 1993, 1996 and 1999: Rural East Java, Indonesia ............................... 96
Table 6.15 Compensated Own Price Elasticities Based on SUSENAS Micro Data
1990, 1993, 1996 and 1999: Urban East Java, Indonesia .............................. 97
Table 6.16 Compensated Own Price Elasticities Based on the SUSENAS
Micro Data 1990, 1993, 1996 and 1999: Rural East Java, Indonesia ............ 98
Table 6.17 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1990 SUSENAS Micro Data: Urban East Java,
Indonesia ....................................................................................................... 99
Table 6.18 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1990 SUSENAS Micro Data: Rural East Java,
Indonesia ........................................................................................................ 99
Table 6.19 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1993 SUSENAS Micro Data: Urban East Java,
Indonesia ...................................................................................................... 100
Table 6.20 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1993 SUSENAS - Micro Data: Rural -East Java,
Indonesia ....................................................................................................... 100
Table 6.21 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model for
Food Based on the 1996 - SUSENAS Micro Data:
Urban East Java, Indonesia ............................................................................ 101
Table 6.22 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1996 SUSENAS Micro Data: Rural East Java,
Indonesia ....................................................................................................... 101
Table 6.23 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1999 SUSENAS Micro Data: Urban East Java,
Indonesia ....................................................................................................... 102
Table 6.24 Ordinary Cross Price Elasticities Derived from the LA/AIDS Model for
Food Based on the 1999 SUSENAS Micro Data: Rural East Java,
Indonesia ....................................................................................................... 102
Table 6.25 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1990 - SUSENAS Micro Data: Urban East Java,
Indonesia ........................................................................................................ 103
Table 6.26 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1990 - SUSENAS Micro Data: Rural East Java,
Indonesia ........................................................................................................ 103
Table 6.27 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1993 SUSENAS Micro Data: Urban East Java,
Indonesia ....................................................................................................... 104
Table 6.28 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1993 - SUSENAS Micro Data: Rural- East Java,
Indonesia ....................................................................................................... 104
Table 6.29 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1996 - SUSENAS Micro Data: Urban East Java,
Indonesia ........................................................................................................ 105
Table 6.30 Compensated Cross Price Elasticities Derived from the LA/AIDS
Model for Food Based on the 1996 - SUSENAS Micro Data:
Rural East Java, Indonesia .......................................................................... . 105
Table 6.31 Compensated Cross Price Elasticities Derived from the LA/AIDS
Model for Food Based on the 1999 - SUSENAS Micro Data: Urban
East Java, Indonesia ................................................................................. .. 106
Table 6.32 Compensated Cross Price Elasticities Derived from the LA/AIDS Model
for Food Based on the 1999 - SUSENAS Micro Data: Rural East Java,
Indonesia ....................................................................................................... 106
Table 6.33 Expenditure Elasticities of Food Demand Across areas and
Income Groups Based on the SUSENAS Data:
1990, 1993, 1996 and 1999 East Java, Indonesia .......................................... 108
Table 6.34 Demand Elasticities of Food Items on Household Size Estimated
from the LA/AIDS Model Based on the 1990 –SUSENAS Data,
Urban East Java, Indonesia ............................................................................ 113
Table 7.1 The Weekly Average of the Household‘s Budget Share on
Food Groups across Income Groups and Areas: The Observed Data
1999 - East Java, Indonesia .............................................................................. 118
 CHAPTER I. INTRODUCTION

1.1 Structural Adjustment and Demand Analysis

Structural adjustment is a term used to signify an economic policy that requires a structural
change in the economy of any country. The policy has been introduced by International
Monetary Fund (IMF) and World Bank (WB) and usually applied as a measure aimed to
gain recovery. Although the application might be different from country to country the
policy is typically signified by the following principles: export-led growth; privatization
and liberalization; and the efficiency of the free market.

Since the mid 1980s, Indonesia has conducted a structural economic adjustment programs
in different intensity. In the pursuit of economic stabilization and promoting non-oil
domestic sectors competitiveness, the adjustment has been undertaken to fight the debt
crisis, balance of payment-deficits, and the growing fiscal burden. The measures taken were
typical for structural adjustment programs: currency devaluation, trade deregulation, budget
cuts, and reduced subsidies. However, due to domestic political reasons, the programs have
not been completely implemented as well as expected, despite external pressures.

In July 1997, the financial crisis shocked the world. While Asia in general the most critical
region to experience the crisis, Indonesia had the worst case among countries in this region.
These indicators reflected the situation at the end of 1997: (i) a currency depreciation of 80
per cent, (ii) inflation rate of 50 per cent, (iii) a sharp increase of unemployment rate, (iv)
loosening stock exchange value, (v) an increasing capital outflow, and (vi) economic
contraction of around 15 per cent/per year compared to previous year1.

This crisis has brought about dramatic changes, including the availability and accessibility
of foods. Food security of the country has been seriously affected through job losses, the
consequent decline in household incomes and access to food, and the rising price level of
consumer goods which led to a sudden drop of the purchasing power of the people.

1
Presentation by Dr. Syahril Sabirin, the Governor of Bank Indonesia at Pbancque de France,
(Paris Club) March, 1999.

1
To respond the crisis, Indonesia invited IMF and World Bank for an economic recovering
program. As a result, the government of Indonesia (GOI) faces a daunting list of obligation
and challenges. Indonesia has committed itself on what is termed letter of intent (LOI), and
this LOI called for a strictly monitored structural adjustment. Unlike the previous
adjustment programs, the post-crisis structural adjustment program has been accompanied
by an intensive pressure from international lending agencies, mainly the IMF and the World
Bank demanding for liberalization of the domestic market and trade deregulation. The
critical impact of these changes could be foreseen. Before the program, the agriculture and
food policy of the country has been characterized by heavy input subsidies and low
subsidized consumer prices for staple food. The policy used to be applied to protect low
income consumers (World Bank, 1986, 1999) who in most cases be also producers at the
same time (Timmer, Falcon, and Pearson, 1983).

It is believed that a substantial change in consumption pattern took place, but exactly how
that change took place is still unclear. Indonesia‘s policy makers are therefore being
challenged to design policies that are budgetary admissible, but enable the poor to be
rescued from hunger and under-nourishment.
There is another important consequence resulting from the current Indonesia‘s crisis.
Indonesia is now undergoing a profound political dynamic. Among other things, there is a
stipulation of new laws enforcing an adoption of decentralization concepts on previously
extreme centralistic administration system of the Indonesian government. Documented
under State Law number 22 1999, this law stipulates that decentralization of power and
responsibilities from the central authority to the local district authorities covers all aspects
of government administrative sectors except for security and defense, foreign affairs,
monetary and fiscal policy, justice and religious affairs. It is fairly justified to assume, that
local specific characters of each region or local (district) authority will be more determining
in shaping local economic policy. For the anticipation, policy makers have an urgent need
on accessing local specific information, for the sharpening of regional specific economic
policy making. This requires a deliberate and local specific study.
Since food budget is still dominating to the total household expenditure, information on
food consumption pattern is of significance. This study was conducted as an attempt to

2
provide information needed in designing new food policy with possible instruments, to
cope with new situation, with specification of East Java Province.
If the policy makers under allowing condition intervene to help those who are most
severely impacted, the policy makers have to identify those who have been most harmed
and the magnitude of the harm. In order to have a good scheme that brings forth the
conduct of a wise fiscal policy especially during the period when the economy is being
restructured, governments, including the government of Indonesia, need to know the
approximate magnitude of the elasticities of some important goods. For that, one needs a
knowledge that can be derived from a study of demand system. This study serves the
information on how households respond to the environments they face. The important
determinants are relative price and level of income. Policy makers on agricultural and Food
sector have a considerable interest on the matrix of elasticities derived from food demand
analysis. These are considered to be the decisive information for food supply planning and
correspondingly, food production related issues. It may also assist in structuring and
development of agricultural sectors policies as well. Knowledge on food consumption
pattern which is normally a by result in demand analysis, may be viewed as an indicator of
welfare barometers, and of course to design pricing policies of some strategic food
commodities in country of crisis. This study has been done on the setting of this
problematic economy.

1.2 Need for a New Study on Household Reaction

Unfortunately, the information on food consumption pattern in Indonesia is still rare and
deficient. Until the 1990s, demand studies available for Indonesia, although extensive, do
not include much information on cross price effects (Teklu and Johnson, 1987. They used
single model, were therefore difficult to accommodate restricting assumptions in the
preferential structure, and were inconsistent with demand theory, except with strict
assumptions. The functional forms applied were less restrictive and theoretically consistent
flexible (Teklu and Johnson, 1986).
To our knowledge, the study on demand pattern in East Java, especially that of after crisis
period, has not been made. This study was motivated to fill this failing information. It was

3
undertaken to attain a reasonable support and theoretical plausibility for policy analysis.
For that reason, a demand system approach was employed as a basis of the analysis.
The decisive impacts of economic changes and government policies (and programs) on
consumption are determined by the responses of households. It is the household which
functions as an intermediary between policies/programs and their impacts on individuals.
The predicted reaction of households to any intervention should be a crucial factor in
assessing the merits of various policy alternatives. On that ground, we used a micro data at
household level in this study. With this idea in mind, we hope that this study may enrich
our understanding of household demand behavior on food.
As noted above, though food demand studies in Indonesia are not new, there is an obvious
lack of knowledge about interrelationships among commodities and food consumption
behavior across regions. This knowledge on other hand 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, the present study will be
focused specifically on food consumption behavior of urban and rural consumers in East
Java, where a mixed of majority Javanese and a mix of other minorities reside.
The comprehensibility and importance for policy-makers of the elasticity concept have
been well elaborated by for instance in Timmer et al (1986), Deaton (1989).

1.3 Cross-sectional Demand Analysis

There are many methods available for estimating the impact of large price change on
consumer‘s behavior. One of them is the applied empirical method using an econometric
method. As a micro econometric research, this work deals with empirical analysis of
households, especially their behavior of household in allocating their disposable income on
food items. The study emphasizes the use of empirical applications of microeconomics,
with implications for efficiency and welfare analysis.
To capture the change, the estimation would have been done on the time series data of
thousands of households collected from the periods before and after the crisis. This would
be the ideal case, because such condition would allow us to examine the behavior change of
the consumer when facing the large price change. Another aspect that belongs to the ideal

4
condition is also the availability of price data of food items composing the household‘s
shopping basket. Such ideal conditions, unfortunately, have been not the case, in Indonesia.
The possible available data is an aggregate data set coming from some resources.
According to Leontief` (1993)2 however, it is deficient if economic analysis be done on the
basis of such models. Aggregative time series are mostly interdependent between
successive observations. Furthermore, in the macro-analytical approach, complicated
systems are usually formulated in terms of a small number of aggregative variables. Still,
there are real possibilities that the analytical results fail to conceive rapid structural change
due to aggregative measurement and the necessary attendant lengthening of the time series
data used. The average consumption level of any country, for example hides a considerable
variation among families within that country: as an expression of, among others; inequality
in income. More importantly, in Indonesia there is no long enough reliable record of time-
series data to allow the estimation of the price elasticities of certain goods to be executed.
The likely consequence of this could be a poor empirical performance of such models. To
overcome the drawback of modeling a large and complex aggregate economic system, one
may use disaggregated models based on individual household data, revealing real food
items consumed. The present study employed a household consumption/expenditure cross-
section data set to meet this proposition. It is a methodological-and empirical exercise on
the economics of household demand for a number of foods. The data used were micro-data
of household consumption/expenditure of East Java province, from 1990, 1993, 1996 and
1999. Its novelty, for current Indonesian context at least, lies primarily in the use of large,
detailed set of carefully compiled micro data from household consumption surveys. In
response to the economic crisis, thus, these four survey rounds provide us with data body
covering pre-and post crisis years. This may allow us to capture the changing pattern of
household economic behaviors, especially their consumption behavior.

1.4 Food Demand Study and Policy Analysis

Basic changes in regulation of market can provide an opportunity for low-income
consumers to improve their diets and real incomes; the information on variations in

2
Leontief suggests, that there is a need for the methodological reorientation of economic analysis. (See American Journal
of Agricultural Economics.75th Anniversary Issue. 75:2-5)

5
consumer preferences among income groups can be used to improve the cost-effectiveness
of food aid programs, through use of "self-targeted" commodities. Knowledge of
consumer‘s response to changes in economic parameters is of great relevance in policy
making. Therefore a well-known application of household demand study on consumption
commodities has been in the area of policy analysis. For policy makers information on
consumer demand behavior on certain commodities is important to the whole steps of
policy making process: to design, to implement or to evaluate a certain policy. Particularly
for food policy making, the policy makers 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. The demand study will provide them with knowledge of
consumption behavior of individuals or groups of individuals in their reaction on changing
prices or other economic parameters affecting their consumption behaviors.
This study is also guided by such pattern of reasoning. For us, this food consumption
analysis serves two specific functions. First, the analysis provides us with consumption
parameters to understand adjustments in the changing macro food economy. Precisely, we
might, empirically legitimated, conclude from the study, what happens to budget share of
rice in any household when prices or incomes fall. Second, this analysis may help us to
hypothesize at least, the likely nutritional impact of changes in the economic circumstances
of the poor: What happens to the consumption bundle of the poor when their incomes
change and prices fluctuate for the commodities they consume?
One of important province in Indonesia is East Java. A physical as well as statistical
observation indicates that this province is a typical region with different spatial
characteristics due to different endowment factors, pace of growth on development and at
certain level, cultural backgrounds that might differentiate their purchasing behaviors.
Because of this, using cross section data of several survey rounds might hopefully portray
such phenomenon.

6
1.5 The Need of Evaluation of Alternative Price Policies on
Different Income Groups
3
The economic development of Indonesia during the 80‘s was distinctive significantly from
that of previous episodes in the sense that in this period growth was brought about by what
might be called efficiency-led or supply side paradigm. This growth was in part due to a
deregulation of markets or otherwise market liberalization programs starting in 1983. An
important factor to explain this development is that the deregulation program shifted the
economic activity to non-subsidized sectors which further created its own momentum for
the economy to grow: (1) the increasing productivity raised income and domestic demand,
(2) expanding financial market mobilized savings and funded economic activities (3)
expanding economy increased the confidence of foreign investors and buyers to have an
idea that doing business in and with Indonesia is profitable.
The structural adjustment in Indonesia has been taken in response to internal and external
imbalances. The structural adjustment „package― which was adopted and implemented
covered four broad categories of measures relating to (Thoreback, 1992): exchange rate
management, fiscal policy, monetary and financial policies, and trade and regulatory
reforms. The trade and market liberating are also a manifestation of the country‘s position
as a ratifying nation of Uruguay Round as well as AFTA (ASEAN Free Trade Area). The
result of this adjustment is an economy, that more market-oriented is.
Market-oriented adjustments affect consumer‘s wellbeing differently. The welfare effects
of the adjustment vary significantly for different income groups, since the behavioral
parameters with respect to consumption are different across socio-economic classes. The
expected consequences would be severe for the poor and be moderate or otherwise
completely insignificant to the rich. Considering food security interest of the country‘s
poor, the presence of another government policies is required. The policy should be in a
position that can relieve the pressure on the government budget without risking the welfare
of the poor. At this stage, the need for analyzing welfare effect on each actual income group
of the country is essential.

3
Parker, S. Survey of Recent Development in Indonesia. Bulletin of Indonesian Economic Studies, Vol. 27 No.1, April
1991

7
When using conventional welfare analysis of price policy changes, it is important to
consider all consumers as a group4. That approach as a matter of fact, provides only a very
general measure of the change in welfare because we cannot further infer the effects on
specific groups of consumers. Since we are concerned with the effects of these adjustments
on the well-being of specific target groups, it becomes neither effective nor useful when the
focus is on all consumers as a group. The results could be misleading and erroneous.
Hence, there is a need to make use of the specified demand equations possibly not only to
measure accurate welfare effects caused by a given price policy on different income groups
of consumers but also to manage in creating the possibility of designing compensation
schemes for the poor.
Urgent agenda for the up-coming years includes the completion of reform and
democratization within government institutions, the resolution of current and potential,
fiscal and political decentralization, the establishment of civilian control over the military,
reform of the justice sector, including bringing Soeharto-era criminals to justice,
eradicating corruption, maintaining and advancing economic policy deregulation and
improving the investment climate.

1.6 The Scope

No one can deny that Indonesia revealed a very impressive achievement with respect of
food issues. For about 30 years, the main food staple in Indonesia has been treated as a
public good in the sense that price formation was being done politically instead of by
market mechanism. But in the last fifteen years pressures toward a relaxing politically
based pricing has been so increasing that the central government of Republic of Indonesia
is no longer in position to maintain their intervention in subsidizing foods.
Central budget pressure is up due to increasing deficit financed by foreign debt. And
foreign financial institutions are increasingly pushing the GOI to relax the pricing regime in
order for it to become more market oriented.
Indonesia committed itself in GATT and other international trade agreement. The spirit of
these agreements is to have an open market. Consumerism, a group movement representing

4
This is conducted using the concept of consumer surplus rendering an exact measure of consumer welfare only in
restrictive cases.

8
consumer‘s interest is now in the advent of getting political power, partly due to the
democratization of the politics in the country.

1.7 Objectives of the Study

Given the above description, the main objectives of this study is seek information
containing consumption preference of East Java Households based on household‘s budget
survey from 1990 to 1999. This is achieved through an estimation of demand parameters
using system approach of demand analysis across different income groups. Having such
demand parameters further objectives are
1. To analyze the effects of expenditure and price changes on demand of eleven food
groups for different income groups in the province of East Java, Indonesia;
2. To evaluate the specific welfare effects of selected price policies for different income
groups
3. To analyze a welfare change and at specific target group using welfare analysis
consistent with this study.

1.8 Organization of the Study

This study is organized in eight chapters. Chapter 1 describes the problem setting and the
significance and contribution of this study for the solution of the problem. This chapter is
closed with a statement of objective and the organization of this study. Chapter 2 presents
the general economic setting of Indonesia and a brief touch of East Java in particular, as an
environment in which the data for this study has been drawn. Chapter 3 describes the data
used in this study, including potential problems when using the data for study of this kind.
Theoretical framework of the study is then presented in chapter four. The chapter begins
with the underlying neo-classical consumer theory to adaptation needed to make the theory
applicable in an empirical works. Discussion in this chapter also addresses the development
of methods used to make this underlying theory applicable in empirical work, to the model
choice and adaptation to be used in this empirical study. The last sections of this chapter
present the working model to be estimated in this study. The methodological part of this
study describes all attempts in the forms of statistical manipulation and techniques that
have been applied in this study to bring the theoretical foundations be operational in

9
empirical work, or to make raw data we have meets plausibility needed for the estimation
process. This part is introduced in chapter five. Chapter six displays the results of
estimation and corresponding interpretation and discussions. Chapter seven demonstrates
the use of empirical results in policy assessment followed by the corresponding analysis.
By using the concept of compensation variation, it is demonstrated that this study provide a
merit and advantage in policy analysis. In this chapter we used the results for the purpose of
welfare analysis. The last chapter concludes the results and makes some suggestions based
upon them.

10
 CHAPTER II. REVIEW OF THE ECONOMIC SETTING

This chapter introduces in brief the economic development experienced by Indonesia as an

independent state. Of the interests are the development strategy applied, the macro
economic development and the role and performance of agriculture and food sectors Vis a
Vis the rest of the economy. The aim is to give an idea of the economic setting in which the
analysis of demand for food items is carried out. The information is required to provide a
logical relevance for the analysis in chapters that follow. Even though it is brief, the chapter
attempts not to forgo the clarity.

2.1 General Setting

The Republic of Indonesia is situated in Southeast Asia region. On August 17th 1945, it
released itself from the Dutch colony by proclaiming independence as Republic of
Indonesia. As an archipelagic country, Indonesia is a cluster, made up of estimated 17 000
islands. It consists of five main islands (Sumatra, Java, Kalimantan, Sulawesi, and Irian
Jaya), two major archipelagos (Nusa Tenggara and Maluku Islands), sixty smaller
archipelagos, and a myriad of small islands that spread around the Equator‘s line over about
5000 kilometers long. By the year 2002, Indonesia is a home of estimated 228 million
people. With this number, Indonesia is thus the world‘s fourth most populated country5.
From the location point of view, Indonesia has a strategic position in the region and may
play itself as an anchor country in Southeast Asia in politics as well as in economy. As the
fourth most populated country, Indonesia has a large number of labor forces with relatively
low wages.
Naturally, Indonesia is a rich country with various natural resources, such as oil, minerals,
rain forest timbers and biological diversity. In terms of size, its marine territory is bigger
than its terrain. This abundance of natural resources is thus a huge potential for agriculture,
deep rain forestry, fishery and mining. Even though it is rich in natural resources, Indonesia
belongs to the poor group of countries in the world. In its early age, Indonesia faced food
scarcity. And this phantom of hunger still haunts the country until nowadays.

5
Sources: Indonesian Central Bureau of Statistics (CBS), CIA’s fact book.

11
The economic performance of this country has been intensively determined by its political
setting. Because of that, this study presents a result of survey on its economy. The
information on the development of its social, economic, and political aspects will help the
readers understanding on the issues being addressed in the next sessions.

2.2 The Economic System

According to the constitution, Indonesia is a socialistic state who gives a mandate to its
government to undertake economic activities for the prosperity of the nation. The
constitution states that "branches of production which affect the life of most people shall be
controlled by the state", so, the government of Indonesia has an important role in the
economy of the nation. It has not only the authority to regulate, but also the legitimate
capacity to be a player in the economy. The government of Indonesia (GOI) is therefore a
regulating agent, development agent, and economic agent at the same time. As an economic
agent, the GOI may be an industrialist, who runs banking, utilities, industries, trading firms,
domestic transportation etc. through its large number of state owned companies. By
government controlling private sector in this sort, the economy was slowly turning from
colonial capitalism to state capitalism. To be regulator, the government sets up and imposes
measures on economic life. Given the very dominant role of GOI, there is a lack of clear
guidance in implementing these principles in real economy. As a result, the role is
interpreted differently by the ruling regimes of Indonesia‘s government. This principle
together with double roles status of the president, as the head of state and the head of
government, are believed to become source of economic mismanagement prevailing until
this moment.
So in the time of the first President Soekarno (1945 – 1966), the principle of state
dominance was interpreted as Berdikari (of self-reliance6) and later as principle of Guided
Economy. Soekarno interpreted the economic mandate he had by undertaking ambitious
building projects, nationalizing foreign enterprises, and refusing to undertake austerity
measures recommended by foreign donors, because such measures would have weakened
his support among the masses.

6
in reality it was an expression of distance from the west while inclining to east block.

12
In the time of Soeharto,  the second president  the economy was also used for political
ends. But Soeharto has run a generally orderly process of development, supported by large
inflows of foreign aid and investment. In a break from the socialistic Soekarno‘s Guided
Economy, Suharto‘s New Order regime welcomes the seemingly private market
development. Closeness to the west, politically as well as economically, was a paramount
distinction of Soeharto era. So, Soeharto interpretation on the role of government in
economic live was probably the single greatest discontinuity from Soekarno.

2.3 The First Two Decades

Following economic pattern of the colonial era, Indonesia was in principle an agriculture
exporting country in the 1950s. For the complete dismantling of colonial economy and
under the spirit of state capitalism, the state owned companies took over all plantations and
colonial or private enterprises in 1957. There were about 300 Dutch plantations and 300
firms in various areas such as mining, trade, finance, and utilities, which were finally under
the control of the Indonesian government. Inexperienced military officers and unskilled
civil servants replaced the management in these companies.
Under the Guided Economy, Soekarno seeks to industrialize the country through the path of
"socialism a la Indonesia". In this regime, development planning and control is centralized.
The nation's first five-year development plan (1956-60) proposed government investment
in public infrastructure, but offered little regulation or overall guidance to the private
sector. This plan was shadowed by dramatic developments in the political and economic
aspects.
By the mid-1960s, half of credit endorsement of the central bank was for the government
expenditures. This deficit spending led in turn to mounting inflation, which peaked at 1,500
percent between June 1965 and June 1966. At the same time, foreign debt increased, both
from the West and increasingly from the Soviet Union. In spite of a highly visible public
building campaign, the economy stagnated and by 1966 per capita production was below
the 1958 level.
To sum up, Indonesia‘s economy in this period suffered from neglect and economic
mismanagement. As a result, the inflation was rampant, the export revenues decreased
dramatically, government expenditure was high leading to budget and current account

13
deficits. Scarcity in basic necessities forced people to depend only on ration. The economic
bankruptcy brought Indonesia into a period of severe political turmoil in the 1960s. In
1967, this political turbulence brought the old government under president Soekarno to an
end; it gave an opportunity to a new government with new economic regime to come to
power. Under the new regime, thanks to its economic reorientation, economy has
transformed from virtually stagnant entity into pre- industrializing economy.

2.4 1966 – 1980: Period of Recovery

In 1966, following the downfall of Soekarno, a new regime came into power under
President Soeharto. The New Order regime pursued, with financial assistance from the
International Monetary Fund, a variety of emergency stabilization measures to recover the
economy. The development plan was made on basis of five-years planning cycles or
REPELITA. The first REPELITA started from 1969. The REPELITA is a medium run
development guidance which is introduced centrally and is applied nation-wide. There is a
national planning Agency (BAPPENAS) headed by a minister-equivalent officer. The task
is to design a development plan that rules nationally. Indonesia under Soeharto has
completed sixth REPELITA until 1998.
The main feature of economic development strategy under Soeharto is the principle of
balanced budget. Under this regime, the current account of the country is kept balanced.
The development pursuit was a pragmatic growth, with fiscal and monetary conservatism
as a hallmark of the economic regime. Growth in the money supply was restricted to
contain inflation. Similarly important is the role of government as an industrialist by state
direct investment, increasing regulations and offering special protection for favored
industries.
This period was marked by an increasing oil price and thus booming oil revenue for
Indonesia. At the same time, Indonesia experienced a massive capital inflow foreign aid. As
a consequence, the Indonesian economy has experienced a major expansion in the first
three REPELITAs, marked by GDP growth at 7.2 per cent yearly and rice self-sufficiency
by mid-1980s (Thorbecke, 1992).

14
2.5 1980 – 1996: Stabilization and Growth
Oil crisis has forced Indonesia made two important changes: stabilization programs and
promotion of non-oil industries as an alternative fuel of growth. The stabilization program
was aimed at solving the balance of payment deficit and the growing fiscal burden. Other
government response took the forms of devaluation of rupiah (Indonesian currency),
deregulating measures to promote non-oil exports, budget cuts and reduced subsidies.
Several capital and import intensive projects were postponed and subsidies on fuel,
agriculture and states enterprises were reduced (Nasution, 1991).
Under this program non-oil resources have been worked out, work force has been trained in
basic skills, and the strategic geographical location has been promoted seriously. As a
result, Indonesia maintained most of the advantages that fuelled rapid economic growth
during the 1980s and early 1990s and a large and expanding internal market of
approximately 210 million people have been developed, until the crisis broke out in 1997.
Despite of the crisis at the end of the 90s, these factors will remain attractive for other
countries, especially if the government of Indonesia makes significant advancement on
their policy challenges.

2.6 The Debt

As the role of debt in shaping the policy choice of Indonesia‘s government is so obvious, it
is important to review the debt of Indonesia. As reviewed previously, Indonesia adopted
what is called a floating balance of account: The deficit is funded by foreign debt from
many overseas lending agencies. In practice, the government of Indonesia borrowed abroad
each year, primarily from the World Bank, Asian Development Bank, and a group of
bilateral donors grouped in the Consultative Group on Indonesia (CGI). The proceeds were
used to fund the development budget. By long-established convention, the GOI of the new
order regime – contrast to the predecessor-- avoided domestic borrowing, and Indonesia's
debt - GDP was sustainable. Indonesian debt management policies were an important part
of what was widely viewed as a prudent macroeconomic management strategy.
Prior to the 1997-98 financial crises, Indonesia‘s debt was considered not too critical. This
situation changed in 1998-99, when Indonesia for the first time developed a large domestic
debt stemming from the costs of the country's banking sector bailout. At that time,

15
Indonesia's official debt burden increased from 27 percent of GDP prior to the financial
crisis to approximately 100 percent of GDP at the end of 2000. Although Indonesia has
shouldered high debt - GDP ratios in the past (most recently in the late 1980s), the costs of
servicing the country's official debt placed a heavy burden on the budget. In 2001, interest
payments on Indonesia's domestic and foreign debt were forecast to reach almost 35
percent of central government expenditures. By way of comparison, development spending
accounted for only 17.5 percent of domestic government expenditures, and more than half
of this sum stemmed from donor-financed development projects.
Table 2.1 GOI Foreign and Domestic Debt, 1995-2000 (USD Billions)
DEBT/GDP
YEAR TOTAL RATIO
Foreign Domestic
1995 63.5 0.0 63.5 31%
1996 56.3 0.0 56.3 25%
1997 57.9 0.0 57.9 27%
1998 67.3 0.0 71.5 72%
19991) 75.8 68.7 144.5 102%
20002) 74.8 78.0 152.8 100%
Source: Bank of Indonesia.
1)
Domestic debt figure based on Rp. 312 trillion in bank recapitalization bonds issued, plus Rp. 228 trillion in
bonds issued to repay Bank Indonesia for liquidity credits, converted at the 1999 average exchange rate of Rp.
7855/USD.
2
) Domestic debt figure based on Rp. 430 trillion in bank recapitalization bonds plus RP 228 trillion in bonds
issued to BI, converted at the 2000 average exchange rate of RP 8430/USD. Foreign debt figure is presented
through October 2000.

In order to reduce the short-term burden to the budget, Indonesia has concluded debt-
rescheduling agreements with the Paris Club group of official bilateral creditors on two
occasions. In September 1998, the GOI and Paris Club agreed to reschedule USD 4.6
billion in principal payments falling due from August 1998 to March 2000. In April, they
concluded a similar agreement rescheduling USD 5.8 billion in principal payments falling
due from April 2000 to March 2002. However, Indonesia cannot take advantage of the
latter reduction until it reaches agreement with the IMF on a new letter of intent.

16
 Table 2.2 Indonesia: Net Capital Inflows (USD Billions)

1993 1994 1995 1996 1997 1998 1999 20002)

Private 5.2 3.7 10.3 11.5 -0.4 -13.8 -9.9 -8.5
FDI (1) 2.0 2.1 4.3 6.2 4.7 -0.4 -2.7 -4.1
Other 3.2 1.6 5.9 5.3 -5.0 -13.5 -7.2 -4.4
Official 12.8 0.3 0.3 -0.5 2.9 10.0 5.4 3.8
Total 18.0 4.0 10.6 11.0 2.5 -3.9 -4.5 -4.7
Source: Bank of Indonesia; (1) Foreign direct investment; (2) Preliminary data.

2.7 Economic Structure

Thanks to richness in natural resources, Indonesian economy has been for almost three
decades extractive in manner, in the sense that resource extracting sectors like mining,
forest, cash crops agriculture dominated the contribution in the country‗s general domestic
product (GDP).
Before the mid-1970s exports consisted mainly of a small number of primary commodities,
including natural rubber, coconut oil and copra, tin, and crude oil. By the end of the 1970‘s
oil sector has been the main foreign currency earner. Because of deregulation, the country
has become less dependent upon exports of oil and gas since 1980, and efforts to increase
other exports have been encouraged.
It is only from the mid 1980s on, that Indonesia has a reasonably well-balanced economy in
which all major sectors, including manufacturing industry and services play an important
role. Agriculture (including animal husbandry, fishing and forestry) has historically been
the dominant sector, in terms of both employment and output. There is a vast range of
mineral resources, the extraction and exploitation of which have proceeded rapidly in the
past three decades, enabling the mining sector to make an important contribution to the
balance of payments. The manufacturing sector also expanded dramatically during the New
Order period, especially since the mid-1980s. The decline in petroleum prices after 1983
resulted in a concerted push towards industrialization, as a result of which semi-processed
and manufactured products increasingly came to dominate exports. A determined effort to
promote tourism since the mid-1980s has also had a big impact on invisible export earnings
during the past decade. In 1991, the share of manufacturing in GDP exceeded that of
agriculture for the first time. More recently, the services sectors have expanded rapidly, and

17
in 2000 jointly accounted for approximately 40% of GDP and employed about one-third of
the working population. Exports provide the main impetus for growth. Low levels of
domestic disposable income mean that exports have been the primary engine of growth.
Because of enduring degradation of natural resources, it is expected that the contribution of
primary products becomes less and less. Forests, for example, are declining by as much as 1
million hectares per annum, and Indonesia is expected to become an oil importer early in
the next century. As for primary commodities, its relative share in total GDP was 60
percent in 1970. It was 39 percent at time of growth, and became only 8.6 per cent in 19987.
The sharp contrast is performed by valued added of this sector, meaning that manufacturing
of agricultural produces experienced an increasing tendency.
The natural resource base of the country is increasingly degraded, leaving less for the
regime to exploit, and less for the growing rural population to seek its livelihood from.
Indonesia economic development performance in the 1990s, which is the period of the
survey for this study, is briefly displayed in the following table.
Table 2.3 Selected Macro Indicators of Indonesian Economy
1990 1993 1996 1998 1999
GDP ($ Billions)* 114.4 158 227 96.8 141,3
Real GDP Growth (%) 7.2 6.5 7.82 -13.2 0.3
Per capita GDP (US $) 623 630 1146 1070 448
GDP by Sectors
Agriculture 20.6 18.79 16.3 18.4 27.8
Manufacturing 37 39.42 41.6 23.4 36.2
Services 29 41.79 80.5 35.7 56.9
Government n.a** nab 10.5 4.1 7.2
Labor Force (Millions) 75.9 n.a 94 92.6 94.8
In Agriculture 55.9 n.a n.a n.a 43.2
Currency Rate (Rp/1 US$) 1842.8 2087.1 2342.3 14.850 7855.200
Sources: Central Bureau of Statistics, CIA Fact book, * CIDES, Bank Indonesia, Ministry of
Finance, U.S. Commerce Dept. +June 1998  the IPCC Data Distribution Centre.
n.a = not available

Government development emphasis in early 1990s was typically characterized as less

interference in private business and greater support on technology inputs. The agriculture
sector predominated and benefited from the infusion of modern technology by the
government. Indonesia was at that time a major recipient of development aid from

7
Source: Bank of Indonesia, Source: U.S. Department of Commerce, Bureau of Economic Analysis.

18
international donors. Major trade partners were Japan and United States, and the trade with
ASEAN fellow members was increasing.

2.8 Food and Agricultural Policy

Compared to other sectors in the economy, the contribution of Indonesian agriculture sector
to the gross domestic product has been steadily declining. However, its role remains
important to the economic development of the country.
Agriculture sector includes forestry and fisheries. It is the most important sector of the
economy in terms of employment. As shown by the 1990 census, its share in total
employment was 55 percent. That percentage practically remained unchanged from the
previous census in 1980. The current indicator shows that the sector provided
approximately 41 percent of employment for the country‘s labor forces.
In term of output, the share of this sector to the GDP declines steadily. While the
contribution to the GDP in the 1970s assumed approximately some 33 percent, its share by
the early 1980s was around 23 percent, and further declined to 16.3 percent in 1996.
However, its important role in the provision of food and employment for the population of
the country is still evident.
Indonesia has been internationally recognized in its achievement in agricultural
development due to increasing rice production which has led to rice self-sufficiency for the
country in 1984 (Dillon, 1992). The achievement of agriculture sector in term of domestic
production and the improvement of food security was still remarkable until 1996.
Governmental programs, introducing modern agricultural techniques, infrastructure in the
rural areas, heavy subsidies for both input and output of food producing sector, have been
identified as factors contributing to that achievement.
Self-sufficiency on food or rice particularly, has been a main concern of food policy in
Indonesia, likely for social-political reasons. Because rice occupies about two thirds of the
total harvested area under food crops, its performance dominates overall agricultural
growth. The interest on self-sufficiency on food for a country like Indonesia is clear, when
one takes into account of the potential risks that might come out of its dependence on world
market supply. The risks could be in the form of e.g. trade embargo, or food supply
insufficiency in the international market, say because of intended or unintended stopping in

19
some major producing countries. However, the extreme dominance of rice in both
production and consumption is such that the achievement of self-sufficiency until this
moment poses important policy issues which have ramification for the entire Indonesian
economy.
Therefore, following the achievement of rice self-sufficiency, agricultural policy makers
turned their attention from a rice-based to a multi crop-based food policy. The reason for
that is quite obvious. As described before, due to declining oil revenues and oil price
prospects, changes in agricultural production technology, current level of government debt,
environmental concern, Indonesian is considering reducing its subsidy on agricultural
chemicals, pricing water to recover more of its cost and buffering the domestic rice price
less relative to world market (Teklu and Johnson, 1988). Furthermore, the changing pattern
of rural and urban consumption associated with the development of the country has
imposed a new demand on a policy change.
The adherence into open market mechanism is now becoming imperative as Indonesia, due
to the dramatic crisis, is now committed to implement letter of Intent with the IMF, World
Bank and other international lending agencies.
The changes in consumption pattern associated with these policies must be correctly
anticipated if the agricultural sector is to be properly positioned in the upcoming
development plan. Food policy that is based on accurate information on food consumption
pattern may contribute a just solution in production, distribution and price of agricultural
products. These changes are therefore of interest to agriculture policy makers.
As far as agriculture policy concerned, there have been four applied measures dealing with
crops sector in Indonesia. The measures covered the following areas:
I. Product and area targets extending to the first crop and in irrigated areas also to second
crops. This policy was conducted simultaneously under the title of BIMAS or mass
guidance program, which led Indonesia into the achievement of self-sufficiency in rice
production in 1984. This policy was relaxed in the early 1990s.
II. Price policy consisting of floor price for selected crops, i.e. rice, maize, and soybean, at
appropriate levels through.
III. Agriculture subsidy covering both the agriculture producer as well as consumer
subsidies. The producer subsidy program was called Ponca Usaha, literally means five

20
 items subsidy package; it consists of seed, fertilizers, chemical pesticides, credit, and
consultation service. In addition to that, irrigation tax was not charged to the
agriculture producers. The consumer subsidy was executed through a purchasing
scheme and special market operation for selected target groups.
IV. Price support for agriculture producers. This policy was imposed by using of several
instruments, such as high level price for some domestic produces, import control
mechanism, and domestic marketing boards for selected crops. The marketing board
was in the reality a quasi-state monopoly, in which producer participation to the
marketing scheme was obligatory. Examples for the schemes were marketing boards
for clove and citrus, where the growers of these two commodities were forced to
participate in the schemes. In addition to these schemes, there was also an obligatory
registered channeling of fresh produce through central urban markets, and gradual
relaxation of this rule in the late 1980s;
The above listing policies and recent changes indicates the very broad spectrum of
interventions by the state in the production, collection and marketing of annual food crops
and industrial crops.
All these policy events indicates a strong move towards centralization in the 1970s and the
early 1980s, and after the reform of the banking system in 1986, an increasing emphasis on
local initiatives, and since the early 1990s a reliance on entrepreneurships and abandonment
of central guidance in food crop agriculture. A market-led diversification became a leading
direction in agricultural sector (Affif, 1992) and Timmer (1989:7). A liberalization of the
agricultural and food sectors is now becoming an obligation for Indonesia, as this country
has no longer other choice to take.
To sum up, as a typical phenomenon of developing country, Indonesia started for their
agriculture development from food price policy environment that used food imports and
budget based subsidies for across-the-board consumer protection, while a host of
production-oriented government projects attempts to increase food output.
To the current regime such price policy orientation is deemed backwards. A contending
mainstream of food policy maintains that the government of Indonesia can more effectively
meet the full range of food policy objectives by using price policy, not to keep food prices
low for consumers, but as part of the incentive package that induces greater food

21
production from millions of small farmers. Programs and projects can provide targeted food
subsidies to protect the very poor until they find jobs and higher incomes that result from
the new policy environment. Reversing the prevailing policy toward dealing with hunger
does not mean a new emphasis on production while food consumption problems are
ignored, because such a strategy would fail on both political and humanitarian grounds. The
reversal of policy and project roles does mean dealing with both production and
consumption issues in a manner that creates fewer-not more-problems of poverty and
hunger for the future.
This argument, while gaining full supports from ranges of influencing groups (including of
course the IMF and World Bank, - two determining giants in current Indonesian economic
policy setting), is however politically not yet amenable.

2.9 Consumption Pattern

A brief look on the consumption situation in Indonesia over the survey periods (1990 –
1999) may be followed in Table 2.4.

Table 2.4 Percentage of Average Monthly per Capita Expenditure

on Food and Non-food, Indonesia
NO 1990 1993 1996 1999
1 Food 60.36 56.86 55.34 62.94
2 Non Food 39.64 43.14 44.66 37.06
Total 100.00 100.00 100.00 100.00
Source: Central Bureau of Statistics, Indonesia, 1990 -1999

In the 1990s, the economic development of the country has brought about some changes: an
increasing disposable income, changing price ratios and level, altering population structure,
changing tastes and habits, and the incoming of new products by multinational food
companies. This situation allows a new food variety to be offered to the Indonesian
households. Food categories itself have changed, among other thing, due to increasing food
manufacturing and retailing industries. These factors may affect the consumption habit and
consequently consumption pattern of the people. Some indicators confirmed these changes:
1. An obvious change of per capita income: In the mid-sixties, the Indonesian per capita
income was $ US 70. In 1996, before the crises broke out, the per capita income was $
US 1100 (Bank of Indonesia, 1996).

22
2. Changing rate of population growth: The growth rate was 2.0 % per annum in 1990.
After a slight decrease in 1996, it raised to 2.3 annum again in 1999. However, this
growth has been accompanied by better other demographic indicators, such as the
increase of life expectancy and the decrease of infant mortality, etc.8
3. The changing of retail structure entailing urbanization; the change has induced the
retail price structure as well. This has caused changing relative prices, i.e. money price
of any good after being deflated by consumer price index.
4. On the other hand, the intensive effort of the government of Indonesia to achieve rice
self-sufficiency seemed to bring a crucial impact on consumption pattern of Indonesian
food consumer.

These factors are believed to be the main determinants in shaping of consumption pattern in
Indonesia. Along with Table 2.4, Table 2.5 to 2.7 displays some other indicators for
consumption pattern in Indonesia.
Tables 2.4 reveals that, in the period of economic growth (1990–1996)9 the food
consumption in Indonesia followed the pattern of Engel‘s Law: the percentage of food
budget decreases as the purchasing power of consumer increases. The figure of the year
1999 indicates that the economic crisis in Indonesia has set the level of people‘s wellbeing
back into the period before the crisis, roughly formulated. The loss of purchasing power has
forced people to meet their basic need first.
Table 2.5 Monthly Average Budget Share of Food by Commodity Groups
in Indonesia, 1990 – 1999
Food Groups 1990 1993 1996
1 Cereals 29.89 24.30 23.12
Tubers 1.66 1.49 1.22
Fish, Meat, Eggs and Milk 18.78 19.47 19.84
Vegetables, Legumes, Fruits 18.42 17.47 17.69
Other Items 14.88 14.87 14.61
Prepared Foods& Drinks 8.40 13.51 15.35
Alcoholic beverages 0.20 0.19 0.14
Tobacco & betel 7.77 8.70 8.03
Total Food Share 100.00 100.00 100.00
Source: SUSENAS –1990 – 1996

8
Source: Statistics Division, United Nations Economic and Social Commission for Asia and the Pacific (ESCAP)
9
In this period, average economic growth was 6.5 percent.

23
The consumption structure of average Household in Indonesia is presented in Table 2.5,
while Table 2.6, decomposes the consumption pattern across the areas: rural and urban.
Table 2.6 Monthly Average Share of Food Expenditure
by Commodity Groups in Rural and Urban Indonesia 1990 – 1996

COMMODITY 1990 1993 1996

GROUP Rural Urban Rural Urban Rural Urban
1. Cereals 22.87 11.9 18.41 9 17.4 8.47
2. Tubers 1.38 0.52 1.2 0.47 0.95 0.42
3. Fish 6.39 4.73 5.96 4.41 5.58 4.03
4. Meat 2.54 3.4 2.43 3.54 2.54 3.7
5. Eggs and Milk 2.25 3.38 2.41 3.42 2.69 3.2
6. Vegetables 6.16 4.32 5.84 4.05 5.89 4.08
7. Legumes 2.61 2.25 2.42 2.04 2.22 1.67
8. Fruits 3.4 3.22 2.71 2.76 2.78 2.98
9. Oil and Fat 3.26 2.08 3.1 1.95 2.99 1.85
10.Beverages Stuff 4.1 2.76 3.82 2.52 3.71 2.31
11. Spices 2.38 1.61 2.25 1.47 1.68 1.15
12. Alcoholic beverages 0.14 0.09 0.13 0.08 0.09 0.07
13. Tobacco and betel 5.33 3.88 5.78 4.08 3.88 3.43
14.Miscellaneous 0.78 0.66 0.96 0.75 1.42 1.14
Food Budget Share 63.59 44.8 57.42 40.54 53.82 38.5
Source: CBS, Indonesia, 1990 - 1996

Table 2.7 shows another indicator of consumption situation in the study area East Java. The
table presents an average structure of consumption in term of calorie and protein intake.
Food groups presented in the table are constructed from around 250 food items listed in the
survey of SUSENAS10. The prepared food group covers all food items that are produced
mostly by food manufacturing industry, not prepared by the households themselves, in the
form of either purchased or given as a gift, or both, to the households.
Though the four tables express consumption pattern differently, there is a generality that
can be caught up from them, especially with regard to the consumption in the study areas.
The consumption pattern follows more or less the same tendency.

10
See chapter 3 for further information about SUSENAS

24
 Table 2.7 Average Daily per Capita Consumption of Calorie and Protein
by Commodity Group, 1990, 1993, 1996
CALORY (Calorie) PROTEIN (gram)
NR. COMMODITY GROUP
1990 1993 1996 1990 1993 1996
1 Cereals 1247.20 1 210.42 1 152.86 24.08 23.26 17.03
2 Tubers 106.57 93.70 58.12 0.88 0.81 0.44
3 Fish 38.33 40.14 42.62 7.01 7.26 7.16
4 Meat 20.02 20.91 38.74 1.31 1.40 2.52
5 Eggs and Milk 21.53 27.79 34.82 1.33 1.67 2.07
6 Vegetables 40.33 37.75 36.25 2.85 2.63 2.45
7 Legumes 49.17 51.07 60.48 4.65 4.97 5.08
8 Fruits 42.88 37.83 40.43 0.51 0.43 0.41
9 Oil and fat 201.33 212.49 221.53 0.75 0.71 0.55
10 Beverage Stuff 90.19 94.17 113.64 0.72 0.79 0.95
11 Spices 26.41 27.54 15.55 0.81 0.84 0.67
12 Miscellaneous 12.09 15.71 34.17 0.43 0.53 0.62
Total (without no.13 &14) 1 896.05 1869.52 1849.21 45.33 45.30 49.93
13 Prepared Food 87.03 149.31 170.46 2.06 3.59 4.56
14 Alcoholic beverages 0.15 0.14 0.12 - - -
Source : Central bureau of Statistic, Indonesia, 1996

Following patterns of consumption are learned from the previous tables:

1. Over time of observation, food still embodies a main component of monthly
expenditure of average Indonesian. This holds for Indonesia nationwide.
2. Staple foods (cereals and tubers) account for majority of the expenditure. Among the
staples, rice is the prime item. Rice is the main staple crop in all parts of the country.
This has been the case, partly because of government food policy that put rice as a
single leading indicator to define self-sufficiency for the whole country. The
government defines self-sufficiency as the condition in which national production of
rice has reached a pre determined quantity. The quantity is determined simply by
multiplicationon per capita basic need of rice with the number of Indonesian
population, including those who are traditionally use other stuff as a staple. With this
approach applied in the copuntry rice becomes more and more available all over the

25
 country, and people who are previously not rice eater learned to eat rice, simply
because rice is more accessible for them. Rice is therefore consumed by more than 95
% of the population; although in some parts of Indonesia people have previously their
own local staple food. As secondary staple foods follow maize, cassava, sweet
potatoes, peanuts and soybeans.
3. The prepared foods in aggregate gain more and more preference from the Indonesian
households. Its pattern follows an upward trend. Urbanization is deemed as the factor
responsible for this tendency. The tendency may represent the consumption of people
in some central-industry areas, which has been reducing their consumption on rice and
substituting it with wheat based processed food. However this phenomena is still of
minor significance, compared to the whole country tendency.
4. Share of (supposedly to be) luxurious food groups (meat, fish) to the total food budget
of the east Java households are still low.
5. The food share to the total expenditure is bigger in a rural area than in an urban one.
This holds for almost food groups. This mught happen, because the absolute value of
the expenditure in monetary term is low. Stated in monetary term, the average per
capita monthly expenditure for food in 1990 was Rp. 22 633, - (Rupiah11) for urban
area, and 16 379 Rupiah for rural area. In 1996, the value was Rp. 48 278, - in the
urban, and was Rp. 33 345, - in the rural.

2.10 Food Security and Distribution

Food security has been an ocupying issue in a the Indonesian history, and the economic
crisis in 1997 has made the issue become more critical in the country. Not only the
declining of purchasing power and induced price rise (Table 2.8), that threats directly the
food accessibility, but also the climatic condition, like the phenomena of El Nino bring a
growing concern on food security.

11
Indonesian currency

26
 Table 2.8 Price Increases of the Nine Essential
Commodities July 1997-April 1998
Food goods Java Off Java
Rice 50% 37%
Salted Fish 56% 42%
Palm Oil 134% 80%
Granulated Sugar 36% 31%
Salt 66% 32%
Kerosene 8% 6%
Washing Soap 77% 72%
Textiles 38% 39%
Batiks 25% 30%
General 51% 39%
Source: Central Bureau of Statistics. 1999

With the approval of the IMF, the government began subsidizing imports of rice and other
essential commodities in early 1998. Bulog12, whose role was supposed to be cut sharply as
part of the liberalization of the real economy, instead expanded its role as food wholesaler
for rice, soybeans, wheat, sugar and other commodities. The government provided
exchange rate and consumer subsidies for basic commodities, incurring significant costs as
the exchange rate continued to deteriorate. Apart from their burden on the budget, subsidies
introduced other complications.

Food security
Indonesia is presently facing serious problems with regard to food security: The droughts
that came in the time of crisis have worsened the situation. El-Nino phenomenon as well as
forest fires have affected the food production. To fill the gap between the low domestic
production and the consumer demand, the government had to increase food imports. In the
mean time, Indonesia was receiving food aid from Japan, Australia and the UN due to the
crisis. Indonesia is therefore classified as a "Low-Income Food-Deficit Country" (LIFDC)
by the FAO. The government has made efforts to target the subsidies at lower income
groups rather than across-the-board.

12
a statal agency established in time of Soeharto; It engaged in food procurement with monopoly power. It losed
monopoly power after economic liberalisation.

27
 CHAPTER III. THE DATA SET: SUSENAS DATA

This study employed the Household Consumption/Expenditure data: a data set provided by
a so called National Socio Economic Survey (widely known as SUSENAS) in Indonesia.
The set thus consists of grouped, namely a household, rather than individual micro level
data.
This chapter provides explanation on the characteristics of the survey and the data it
produced, so that the feasibility the strengths and weaknesses of it in relation to its
applicability for empirical study of demand can be discerned. A deliberate discussion will
address the nature of data, the related concepts and definitions used in the SUSENAS. In
addition, some aspects or problems encountered when using SUSENAS data for analyzing
consumer behavior is also addressed.

3.1 SUSENAS Data

The National Social Economic Survey (SUSENAS) is a household survey carried out
annually by the Indonesia Central Bureau of Statistics (CBS). The aim is to collect socio –
economic data of households. The following main variables are collected in the survey: the
characteristics of the people‘s education, health/nutrition, housing/environment,
criminality, socio-cultural activity, and domestic travel, community opinion about welfare,
and consumption and income. These variables are classified into two categories: core and
module. The core variables appear on a yearly basis, while variables in the module are
collected and appear only once in every three years. The data sets used in this study belong
to consumption/expenditure module.
Actually, the survey has been started since 1963 and the micro-data produced has also been
there ever since. So the CBS has among their data holdings the micro-data set from 1963 to
1999. However, due to heterogeneity in sample units, sample size, region coverage, and
information contents in some points of survey -years, not all of data set can be merged and
an analysis over time using all rounds available is not viable.
The data sets used in this study are the micro-data of household consumptionn/expenditure
of East Java province, from the SUSENAS rounds of 1990, 1993, 996 and 1999. With

28
respect to economic crisis, thus these four survey rounds provide us with a data body
covering the pre-and post-crisis situations. Each round of survey is broken into rural and
urban households and the estimation of demand parameters was done accordingly. The
selection of four survey coverage is based on the ground that there is an important change
happened between these time span. The change has been so profound, that especially policy
makers need to know, how that economic event impacted the consumption pattern of the
consumers.
A major advantage of the study is that the study included also the most recent data available
in Indonesia, as long as the SUSENAS Household expenditure concerned. So it should
reflect the behavioral response of the consumer of the economy under crisis, in addition to
data for previous periods. Thus the data used in this study covered also adjustment made by
households under investigation as a reaction of economic crisis. Thus both economic
environments, -the pre and the post crisis, are accordingly captured in the survey.
As mentioned previously, household consumption/expenditure data set is one of three
modules contained in the SUSENAS. The original purpose of the Central Bureau of
Statistics (CBS) of Indonesia to run the survey is to collect socio-economic information for
establishing an aggregate data. Three modules mentioned previously are (1) consumption
and income module, (2) health, education, and housing environment module, and (3) socio-
culture, criminality, and domestic travel module. Each module is collected from a sample of
households every year but at different points of time. To protect confidentiality of the
respondents, the household identity is kept anonymous. However, for research purpose of
this study, the data set containing information on individual households was accessible after
anonymization.
The Consumption and income module of the SUSENAS survey covers all household
expenditure during a week of enumeration with full specification of commodities. Listed in
the questionnaires are 231 consumption items, for which data on quantities and values were
gathered. For non-food consumption only value data were generally asked except a few for
which quantity questions were included. The questionnaires include also a section on
income, even though income is generally regarded as delicate subject and omitted in many
budget studies. Questions on household income were formulated so that all income sources

29
were covered. Information on salaries and wages was requested from employee respondents
while profits were reported by entrepreneurs.
Table 3.1 Numbers of Households Co-operating in
SUSENAS in East Java, Indonesia from 1990 – 1996
YEAR NUMBER OF HOUSEHOLDS
Urban 856
1990 Rural 4836
Urban 2762
1993
Rural 4876
Urban 2832
1996
Rural 5183
Urban 3250
1999
Rural 5302
Sources: SUSENAS, 1990 – 1999
Transactions relating interest, rent, gifts, grants, money pooling, loans and commercial
papers were all specified.
The SUSENAS sample was selected as to represent 27 provinces of13 Indonesia - all parts
of the country. With 65 000 households in the sample, representing both rural and urban
areas, nation-wide, the survey was capable of obtaining both national and regional level
estimates. Exactly the same forms of questionnaires, field instructions and coding notes are
used for the data collection in all provinces.

3.2. Data Collection Methods

Data collection was undertaken through direct interview to the selected households. For
information related to household characteristics such as information on housing,
consumption and expenditure, the respondent is the head (the wife or husband) of
household. For information related to individual characteristics such as information about
health, education, activities, and demographic characteristics, the questions should be asked
to each household member, except the information of children where the respondent must
be their parents.
For enumerators, i.e., the people who did the interviews, BPS usually used its permanent
staffs, called mantri statistik (Mantis). There is one Mantis for each sub district (called
Kecamatan). From the 1993 Susenas, where the sample size increased significantly, the

13
The study employed data of one province, East Java representing data of 8 832 urban and rural Households.

30
BPS office branches in all provinces and districtt (regencies) had to hire additional non-
permanent workers (called Mitra Statistik) as interviewers.
Before taking to the field, the enumerators were rigorously trained in a nationally organized
training on the implementation of the survey. The training topics were technical aspects of
the survey such as the definitions of variables used, the procedure of filling out the
questionnaires, interview technique, etc.

3.3 Sampling Methods

The CBS applied the three-stage stratified sampling for the SUSENAS with the following
procedures: At the first stage, the CBS chooses survey enumeration areas systematically.
This is done by use of a list of enumeration areas for the Population Census as the sampling
frame. At the second stage, the CBS selects one segment randomly from the enumeration
area. In the last stage, the CBS selects 15 households from the segment systematically, to
represent each selected segment. Every household selected in the survey was visited by an
enumerator who was given the responsibility of interviewing directly the respondents to
excerpt the required data items. For information related to household characteristics such
as information on housing, consumption and expenditure, the respondent is the head or the
spouse of the head of household. For information related to individual characteristics such
as information about health, education, activities, and demographic characteristics, the
question should be asked to each household member, except the information of children
where the respondent must be the parents.
For food consumption the survey reference period was one week prior to the enumeration
of data. For non-food consumption the reference period was either one month or one year.
In this case, the first concern of the survey design is that each individual household answer
shall accurately report the actual expenditures of the household concerned.
This survey calls for highly detailed information on food expenditure which can only be
obtained reasonably accurately by having the participants keep diary or record of all
purchased. However, the interview method used in the SUSENAS does not rely on
household‘s diary or record keeping. In the case of infrequent purchases, the individual is
frequently incapable of recalling the exact expenditure, and his recollections may be
seriously biased. The interview method, therefore, may have reporting errors. This

31
particularly true if the question of expenditure on food items is answered by the husband
(man) and not by the wife (woman) of the household.

3.4 The Concepts and Definitions

The concepts and definition adopted in the National socio-economic Survey (SUSENAS)
with regard to household, household members, consumption expenditure are given below.

3.4 1 Household
Households were classified as consisting of two types, i.e., ordinary households and special
households. The latter type was excluded by the survey. An ordinary household was
defined as a person or a group of person living in a (physical/census) building or a part
thereof and usually sharing the same pot‘ Sharing the same pot means that everyday needs
of the group were managed together as one unit. Other than the commonly found, i.e., a
household consisting of a man, his wife, and his children, there are other types of ordinary
household such as:
 A person who rented a room or part of census building and managed his/her own
meals;
 A family who lived in two separate buildings, but ‗share the same pot‘, where both
buildings were in the same segment
 A boarding house with not more than ten boarders;
 The household of the managers of boarding institutions when it was separated from
the institution, they managed.

3.4.2 Special Household

Special household includes:
 People living in an orphanage, residence hall, dormitory, hostel, prison, military
barracks, and the like. However, military personnel who live with his or her family
in barracks but managed his or her own meals separate from that of the barracks
was not included;
 People living in boarding houses with not less than ten persons.

32
3.4.3 Household Member
Household members included each of those persons who formed a household regardless of
whether he or she was present or temporarily absent at the date of enumeration. However,
a household member who was on a journey for six months or longer, or less than six
months but intended to move away for more than six months or longer, was not regarded as
a household member. On the other hand, a person who has stayed for six months or longer,
or had stayed for less than six months but intended to stay for more than six months, was
regarded as a member of the household.

3.4.4 Consumption Expenditure

Household consumption was distinguishable between food and non-food and was limited to
goods and services used for household purposes, omitting those items purchased for
business or to be given away, regardless of their source of origin. The reference period for
food consumption was one week. Average per capita expenditure of a household was
obtained by dividing the number of household members into total consumption expenses of
the household. Consumption expenditure includes items purchased as well as those
produced on own account and used in final consumption.

3.4.5 Household income

Income is a flow concept and must be measured in amount per unit of time. The money
income of a household encompasses rupiah (Indonesian currency) amounts received by
household members from all sources, including salaries, wages, profits, return on saving,
gifts, and inheritance etc. Many households receive income from several sources. The
diversity and irregularity of income sources add to the difficulty of obtaining accurate
reports of income. Income is probably under-reported by many respondents in SUSENAS.

33
 CHAPTER IV. THEORETICAL FRAMEWORK

The main interest of this study is parameter estimates of demand function reflecting budget-
allocation behavior of households in East Java, Indonesia. From these estimates, we can
derive more indicative and predictive information in the form of food demand elasticity.
This information will allow us to conduct a welfare analysis in order to measure the effect
of any change in demand exogenous variable(s), on the household‘s consumption. Hence,
the task of this study is to find a consumer expenditure model that delivers such
information with most high capacity in describing and predicting the consumption behavior
of the households in East Java, Indonesia, the study area. Our methodological choice is the
econometric approach.
This study is therefore an empirical work in the form of econometric estimation of demand
functions using a system approach. A system approach of demand analysis is characterized
as follows:
 it is based on a certain underlying economic theory;
 it involves some assumptions to confine the scope of analysis;
 it uses an advancement of econometric techniques of estimation and inference;
 It follows any econometric algorithm or calculation methods to come to any
expected quantitative results.
This chapter is presented to deal with the above issues. It addresses (i) the underlying
economic theory; (ii) the assumptions made;.(iii) the principle(s) involved and methods
used to transform the underlying theory to econometrically estimable model (iv)
econometrical algorithm and techniques applied in this study.
The aim is not to expose in detail the treatment imposed on the theory itself. It is rather, to
demonstrate the relevance between the model applied in this study with the theoretical
plausibility and its fitness to the available data.
While the data issues have been addressed in previous chapter, we introduce the theoretical
components in this chapter. The statistical procedure and its related issues will be presented
in the next chapter.

34
The first part of this chapter will review briefly the consumer demand theory. The emphasis
is in introducing (i) what the economic theory tells us about the behavior of consumer, and
(ii) what treatment is needed to make that theory useful for our research question.

4.1 Neo-classical Theory of Consumer Behavior

Neo-classical consumer economics is concerned with the question of choice that consumers
make during their budget allocation activities. It gives answer to the question of ―which
consumption basket the consumers take at time of shopping and why?” The consumer may
be an individual or ―a representation of individuals‖ acting as a decision unit of its member.
This study regards the second type of consumer, i.e. the household instead of individual person,
as a representing the decision unit of family members. The consumption basket contains a mix
of all goods14 the consumers purchase at given commodity prices and consumer‘s purchasing
power.
The theory of consumer‘s demand behavior is founded on a very established building
Blocks (Raunikar and Huang, 1987) constituting concepts of utility function, commodity
set, and the axioms governing the ordering of consumer‘s preference. Expositions to
describe how this theory works have been provided, among others, by Deaton &
Muellbauer (1980b); Phlips (1983); Theil (1975), (1976) and Varian (1984). Delineation of
the principle ideas is presented as follows.
Methodological questions of the work addresses the following theme:
(i) A Specification of a utility or Expenditure Functions.
(ii) A Specification of demand functions.
(iii) The Formula for theoretical restrictions.
(iv) The Formula for calculating elasticities: price (own price, cross price; ordinary,
compensated), expenditure, specifics factors.

4.1. 1 Basic Concepts

The standard approach to demand analysis involves an attempt to estimate the following
general demand equation (Houthakker and Taylor, 1992):
qit  f i ( yt , Pit , z1t , z 2t ,...., z nt , uit ) (4.1)

where
14
The goods we refer to, are any goods or services which generates satisfaction to the consumer

35
 q it = per capita consumption,
yt = per capita income,
Pnt = Price of commodities,
znt = Household characteristics, and
uit = error term.

a. Commodity Set
Commodities in the neo-classical framework are assumed to have following properties:
 non-negative
 Divisible
 Unbounded.

b. Preference Axioms
To assure that consumer preference may be represented by utility function with nice
properties, consumer‘s preference is assumed to have to following properties:
 Completeness
 Asymmetry
 Transitivity
 Continuity
 Monotonicity
 Convexity
 Differentiability.

The first three properties ensure that consumer‘s preference is rational. Completness
assumption implies that the consumer is able to rank the bundles and choose between them.
This axiom is also called a comparability axiom (Phlips, 1983). The reflexivity assumption
states that each bundle is as good as itself. The transitivity axiom assures that the consumer
preference is consistent.
These three axioms are sufficient to guarantee a set of commodity bundles to which a
consumer would be indifferent in preferring one over the other (Phlips, 1983). For the
existence of a utility function which transforms the commodity bundles into utility, the
continuity assumption is needed. The last four properties ensure that a continuous utility
function that depicts a satisfaction level acquired from the consumption of the commodity
bundle is ―well behaving‖. Well behaving utility function enables someone to get a

36
quantitative information to be used to explain, to describe, or to forecast the consumer‘s
behavior. Once the existence of utility function is postulated, two additional axioms are
used to guarantee the best choice that maximizes consumer‘s utility. They are (a) non-
satiation axiom, and convexity axiom. Non-satiation implies that consumer prefers more
goods than less; and the convexity tells us that the average is preferable than the extremes.
This axiom assures the existence of strict quasi-concave utility function. The assumption of
quasi concavity of the utility function along with twice differentiability is used
conventionally in consumer demand theory (Deaton, 1986).

c. Utility Function
Utility is the economic term for the satisfaction that consumer obtains from consuming
good(s) and service(s). A rational consumer chooses the consumption basket, which
generate the highest level of utility. On the ground of this postulate, one can derive a set of
demand equations, the parameters of which an empirical researcher seeks to estimate. Once
these parameters consistently are estimated, one can describe and predict the demand
behavior of the consumer. Provided that the preference axioms are met and the commodity
set performs properties as described above, the utility function will also in possession of
nice properties: i.e. order preserving, monotonic, quasi concave, real valued and continuous
(Debreu, 1959).

4.2 Setting-Up a Consumer Demand Function

The system of consumer demand functions is derived from a constrained optimization
process. By budget-constrained maximization of utility function one obtains the
Marshallian demand function. It begins with the supposition that consumer maximize their
utility (u) as a function of what they consume (x i) subject to a budget constraint. Noted
mathematically, the objective of consumer is
(4.2)

Using optimization techniques like the Lagrangian, one comes to the Marshallian
demand function:
x = x (y, p). (4. 3)

It is a relationship between the quantity of goods purchased, the prices and income.

37
4.3 Properties of Marshallian Demand Function
By manipulating the first order conditions presented earlier in (4.2) and (4.3) one can
derives important properties pertaining the parameters of demand functions. The
knowledge on this properties helps researcher in resolving the problem of estimation.
The results of changes in prices of commodities and income level of the consumer are
described by the partial derivatives of first order conditions. There are in general, four basic
properties of demand functions: namely, adding up, homogeneity, negativity, and symmetry
that are important in providing a testable hypothesis to test the rationality of consumer
behavior. The properties of demand functions guide the empirical analysis in testing
consumer behavior from real world data. The properties are always effective irrespective of
the form of utility function and take the form of mathematical restrictions on the derivatives
of the demand functions.

4.3.1 Adding Up
Adding up condition comes from the budget restriction and the monotonic property of the
preference. According to this property, because the representative consumer is assumed to spend
exhaustively all of their income, the income or total expenditure should be the addition of the values
of the Marshallian demand function. Formally, it is expressed as
P1q1  P2 q2 ... Pn qn  y
which is the linear budget constraint given in (4.1). Substituting (P, y) for q i we get

 p i q i ( P, y )  y (4.4)
i
By writing in the elasticity notation, one obtains the following equation:
w1e1 y  wneny  1 ; (4.5)

where w1 is the budget share of good 1, and e1x is the income elasticity. This condition, which says
the sum of weighted share of income elasticities, is equal to one; the weights being the budget
shares of the commodities. This condition is known as an Engel aggregation.

38
By writing in the elasticity form, the following expression is obtained

w1e1 j  w 2 e 2 j  . . . w j e jj  w j  . . . w n e nj  0 (4.6)

w e
i
i ij  w j

Where w i is the budget share of good i, and e ij is the cross price elasticity‘s of i th and j the

commodity and e jj is the own price elasticity. This condition, which says that the sum of

the own price and cross price elasticities weighted by their budget shares due to change in
th th
price of j commodity is equal to the negative of the change in price of j commodity, is

known as Cournout aggregation.

4.3.2 Homogeneity
The Marshallian demand functions are homogeneous of degree zero in price and income,
meaning that if we multiply all the prices and income by a constant k, the optimal quality
demanded of commodities is unchanged.
According to Euler‘s theorem, if a function f(y) is homogeneous of degree  , then
derivatives of this function satisfy the following properties:
e i1  e i 2 ... ein  e ix  0 i (4.7)

where eij is the elasticity of demand i with respect to a change in price of good j; eii is
referred to as own piece elasticity and eij is referred to as cross price elasticity; e iy is the

elasticity of demand with respect of a change in income known as income elasticity.

Rewriting (4.7) as

ej
ij  eiy ,

one says that the sum of all own and cross price elasticities is equal to the negative of the
income elasticity. This condition given by the homogeneity property of demand function is
also referred to as the row constraint. If there are n demand equations then there will be n
restriction of the utility maximization problem with a budget constraint.
The homogeneity restriction is not particularly useful for single commodities studies since
such studies seldom, if ever, include all prices and income (Currie, 1972).

39
4.3.3 Negativity
 qi
This property states that the (n x n) matrix formed by the is negative semi-definite.
 Pj

The negativity property places the following inequality restriction on s ij ; the diagonal

∂qi
elements must be non-positive for all i, sii ( )  0 . This also follows from the
∂Pi
u
assumption of quasi concavity of the utility function by which the second derivative with
respect to any price is negative.

4.3.4 Symmetry
The Symmetry property of demand function follows from the Slutsky equation; for any
price j
wj
eij  e ji  w j (e jy  eiy )
wi

Symmetry condition implies that if budget shares and one set of cross prices elasticities are
known along with income and own price elasticities, another set of cross price elasticities
could be calculated.
In applied demand analysis, the properties of demand functions discussed above have
important implications in terms of testing the hypothesis of consumer theory, in imposing
certain restrictions on the parameters of estimations and the expected signs of elasticities.
By Engel aggregation in adding up property,

w ei
i iy  1,

So only n-1 of the income elasticities are independent.

By homogeneity property,

e
j
ij  eiy  0

for each demand function there is one redundant elasticity and, therefore, n redundant
elasticities for n equations. By the symmetry property it applies, that knowing the budget
shares and one of the off-diagonal elements, the other set of off-diagonal could be
calculated, which reduces the number of independent elasticities by 1 2 ( n 2  n ) . In practice

40
to derive all price and income elasticities we need to estimates n 2  n parameters ( n 2 price
and n income elasticities). Using the properties of demand functions, namely homogeneity,
Engel aggregation and symmetry, the number of independent parameters to be estimated
can be reduced to
 2  n 2  n  n 2  n
(n  n)  n  1    
 1 .
  2  2

This is very useful when the applied researcher is faced with the problem of a small number
of goods to be analyzed is 10, and then the total number of elasticities to be estimated is
110. However, using the above restrictions, only 54 parameters need to be estimated. Also,
expected signs of elasticities can be derived from these restrictions. For example, using
homogeneity, if all cross elasticities and the income elasticity for a good are positive then
own price elasticity should be negative.
An alternative to derived demand functions is made available by duality principle. This is
achieved by introducing the indirect utility function. This is done by inserting demand
function q = q (y, p) into a utility function U = U (q) to give maximum attainable utility as
a function of y and p, noted as
U* = U q* (y, p)  =  (y, p).
The function  (y, p) is the indirect utility function. It indicates the monetary value of
maximum utility as a function of price and income. The indirect utility function has the
properties of (i) continuity; (ii) monotonicity (non decreasing in Y; (iii) quasi-convexity;
(iv) zero homogeneity in P and Y; and (v) using derivative property, the Marshallian
demand function could be retrieved from indirect utility functions by applying Roy‘s
Identity (Roy, 1942) according to which
 V ( P, Y ) V ( P, Y ) 
qi ( P, Y )    / 
 Pi Y 

That is duality approach. As demonstrated above, duality theorem suggests that by making
YU(), the value of the income constraint in the maximization problem defined by eq. (2),
then the optimized value of utility in that problem will be equal to Uo, parallel to utility
maximization subject to an expenditure constraint, one can derive demand functions by
minimization of expenditures subject to a utility constraint.

41
 n
min xi  Y   Pi xi so that U 0  U ( x1 ,, xn ) (4.8)
i 1

The duality theorem implies that the solution to the maximization problem is identical to the
solution of its minimization dual when the constraint to the maximization problem is
appropriately defined. By applying this principle, the behavior functions for the xi's are
solved simultaneously for the primal and the dual. Simply put, this means that the levels of
the solution values are the same. That is, xiM = xiU as can be seen in the graph in the two
goods case. But it also means that equations (3) and (4) are equal at that point.
Marshallian demand function is observable but not predictive. The Hicksian is predictive but
not observable. Combining both is the advantage of using the duality principle.
By using of duality principle, the demand functions may be established from derivation of
cost function, the minimum cost of obtaining a fixed level of U at given price. Deriving the
cost function with respect of price (Shepard‘s Lemma) leads to Hicksian or compensated
demand functions. Hicksian demand function is the relationship between the goods
purchased, prices and utility.
This result has important implications in applied demand analysis. If a functional form is
assumed for V (P, Y) then the estimable form of Marsallian demand equations could be
derived using Roy‘s identity and will have the same structure as the ones derived from
direct utility function (Barten and Bohm, 1982). The approach to derive demand functions
using indirect utility function is also amenable for applications in welfare economics and
index number analysis since it represents the allocations to achieve the maximum utility
levels under different prices and income (Jorgenson et. al., 1982).

4.4 Cost Minimization and Hicksian Demand Function

In the utility maximization approach of deriving demand functions for commodities, the
consumer‘s problem was to maximize utility for a given level income. The optimizing
solution of this problem was used to attain some utility level of U. If reformulated as one of
choosing the commodities to minimize the total expenditure to reach the same level of U,
then this problem is described as a dual problem of the former approach. The expenditure
minimization problem is given by
Min P1q subject to U (q)  U . (4.9)
q

42
The solution of this constrained optimization problem is a set of quantity demand functions
which are functions of P and U,
qi*  h i ( P,U )

The demand function is called Hicksian or compensated demand functions. The minimized
expenditure or cost to achieve a certain level of utility U given the price vector P is
P l q *  Ph( P,U )  C * ( P,U ) .

This is known as the expenditure function or cost function. The properties of C* (P, U) are
useful in understanding the restrictions on the demand functions. They are useful in
understanding the restrictions on the demand functions. They are summarized as (i) C (P,U)
is continuous in P and U; (ii) it is non-decreasing in P and U (monotonically); (iii)
homogenous of degree one in price; (iv) concave in prices; and (v) by the derivative
property the Hicksian demand functions could be retrieved from the cost functions using
Shepard‘s lemma (Shepard, 1953),
C
hi ( P, U )  .
 Pi

The indirect utility function could be derived by inverting the cost function and vice versa
using Shepard-Uzawa duality theorem (McFadden, 1978; Diewert, 1974, 1980). The
Marshallian demand functions could be derived by substituting the inverse of expenditure
function into the Hicksian demand function (Deaton and Muellbauer, 1980b). Using
appropriate functional forms in the cost function the demand function could be derived for
applied empirical work (Deaton and Muellbauer, 1980a).

4.5 Special Utility Structures

In addition to the information provided by the theoretical properties of demand functions,
the assumption of specific form of utility function imposes yet further conditions. The
choice of a specific form for the utility function depends on prior knowledge of the
consumer‘s preference structure. Very common behavior assumptions embedded in
specifications of utility functions are separability and additively. The restrictions that these
concepts imply for the demand parameters along with their behavioral consequences are
summarized below.

43
4.5.1 Separability
The separability idea postulates that commodities which interact closely in yielding utility
can be grouped together while which interact only in a general way through the budget
constraint are kept in separate groups (Sadoulet and Janvry, 1995). In most of the empirical
analysis of the demand system, the use of aggregate data for quantities such as food,
clothing and housing is common along with their price indices, rather than individual
quantities of elementary commodities qi and their prices Pi as described in the theory. The
use of such aggregate data requires the assumptions that the utility function is separable in
these aggregate, a condition under which decisions involving these aggregates give a utility
level that is equivalent to the one that would be articulated in terms of individual
commodities.
Separability in general implies that the marginal rates of substitution between pairs of good
in separated goods are independent of the level of commodities consumed outside that
group (Phlips, 1983: The concept of separability allows the use of aggregate data and is
consistent with optimization by stages.

4.5.2 Aggregation over Consumers

The market demand functions are defined as the horizontal summation of the individual
demand functions of all the consumers under competitive market conditions. If the

individual household‘s demand function for the ith good is given by

 
qih  qih P, y h ; h = 1, . . . H (4.10)

Aggregation over consumers deals with the transition from the individual household
behavior to the analysis of the market demand. Two issues are important in the analysis of
aggregation: (1) under what demand function does

( P, y1 , y 2 ,..., y h  qi P, y h 
and (2) does qi P, y  have the same properties as


qih P, y h .
Intuitively, if in aggregate demand function q i P, y  , price and average income do not

change then q i remains the same; Hhowever, if the marginal propensity to consume (MPC)

44
changes due to income redistribution, then the above aggregation may not hold. Let us
have the following linear function qih  aih ( P)  bi ( P) y h in which bi (P) , the marginal
propensity to consume is the same for all consumer in the market, then
f i ( P, y1...y H )  qi ( P, y ) is satisfied.

4.6 Empirical Demand Systems

In general, there are three broad approaches to specify applied demand systems. The first
approach is to derive a system of demand equations from utility maximization problem
assuming specific form of utility functions. The linear expenditure system and indirect
addilog model are example of this approach. The second approach is deriving demand
system based on an approximation on an arbitrarily specified functional form, as in the
Rotterdam Model, transcendental logarithmic system, and almost ideal demand system
(AIDS) (Deaton, 1986, Pollak and Wales, 1992). The third methods is to construct model
with ad hoc (naive model, see Deaton 1986) specifications directly imposing theoretical
restrictions as in generalized addilog model and Theil‘s multinomial extension of the linear
logit model. An understanding of the use of these demand systems for different purposes
and situations and their limitations is helpful in selecting appropriate models to work with
and assess the validity of the empirical results from implementing them. Detailed reviews
of these demand systems are given by Barten (1977), Johnson et al. (1984), Deaton (1986)
and Bewley (1986). Also a number of comparative studies of demand systems have been
carried out to evaluate the appropriateness of this model. (Park, 1969, Yoshihara, 1969,
Goldberger and Gamaletsos, 1970, 1973; Deaton, 1974; Theil, 1975, 1976; Lybeck, 1976;
Pollak and Wales, 1978). Examples of these approaches to empirical demand systems are
reviewed here to form a basis for further review of applied food demand studies of
developing countries in general and the demand studies from Indonesia in particular. Some
of these extensions of these models with recent advances are also discussed.

4.6.1 Linear Expenditure System (LES)

The LES was introduced by Klein and Rubin (1948 - 1948) in an attempt to construct a true
cost-of living index. This system is the first empirical estimation of a system of demand
equations satisfying all general restrictions. Stone (1954) showed that the only linear form

45
of demand systems that satisfy the theoretical restrictions of adding up, homogeneity, and
symmetry is the LES.
The direct utility functions for the LES is of the Stone-Geary form

U (q)  (qi  Ci )bi b = 1

i
i (4.11)

Where C‘s are subsistent requirement of q i and b i is marginal budget share.

Maximizing (4.12) subject to budget constraints results in Marshallian demand equations
given by
bi
qi  Ci  (Y   Pj C j ) (4.12)
Pi j

Where (Y   Pj C j ) is the available income to allocate among goods in a fixed proportion

b i , and is termed as a supernumerary expenditure. Substituting it in U (q), the indirect

utility function is given by

(Y   Pj C j )
V ; b0  bibi (4.13)
b0P bj
j

The cost function derived from the inversion of the above indirect utility function can be
written as (Deaton and Muellbauer, 1980b)
n
C (U , P)   Pk Ck  U  Pkbk (4.14)
k 1

Where  Pk Ck is the fixed cost on subsistence requirement with no substitution, and Pkbk is
the term that allows for the utility to be attained at a constant price per unit.
The linear expenditure system can be derived by differentiating (4.14) with respect to prices
using Shepard‘s lemma and substituting in the indirect utility function;
Pi qi  Ci Pi  bi(Y   Pj C j ) = 1, n goods (4.15)

The number of parameters to be estimated is 2n (n of C‘s and n of b's). Given constraint

imposed by  bi  1 , the LES needs only (2n-1) parameters to be chosen independently.

The adding up and symmetry conditions are satisfied by imposing the following restrictions
on the parameters of the utility function (12),  bi  1 and 0 < bi < 1; q i  Ci respectively.

The income elasticity for commodity is given by

46
 bi
eiy  , (4.16)
wi

bi
where wi  is the average budget share. The marginal budget shares for LES are then
y

given by bi  eiy wi .

The own price elasticity for commodity i is given by

Ci
eii  1  (1  bi)( ) (4.17)
qi
and the cross-price elasticity between good i and price of good j is given by
Pj C j
eij  bi ( ) for all i = j (4.18)
Pi qi
It means that for LES, all goods are gross complements (Jonhson et al. 1984).
The LES incorporates the restriction implied by an additive utility structure, thus
maintaining strong separability assumptions with price and expenditure elasticities
proportional (Deaton, 1974). This condition is appropriate only for broad groups of
commodities, which restrict the use of LES in disaggregated commodity analysis. The
restriction 0 < bi < 1 implies that all income elasticities are positive, thus not allowing for
negative income elasticities such in inferior goods. Because of view parameters, LES is
applied in cases where data are scarce and less parsimonious model cannot be used (Lluc,
Powell, and Williams, 1977). Despite these limitations, however, experience with LES
shows that it is a reasonable model, if the good are broadly grouped and price variations
within these groups are restricted (Phlips, 1983).
A generalized version of the linear Expenditure System (GLES) was developed by Wales
(1971) incorporating the concept of elasticity of substitution between the uncommitted
expenditures; in this version Budget share is given by
n
Pi (Y   PiCi )
k 1

  Y

Pi  qi pi 
wi  Ci   n  (4.19)
Y  
 
k 1
q k Pk1 

47
Where  the elasticity of substitution and LES is is a special form of GLES with   1 .
Another extension of the linear expenditure system has been to incorporate inter- temporal
effects (Lluch, 1973) and could be written as

Pit qit  Pit Cit  bit (w   Ptk*Ctk )  Vit (4.20)

t k

where C it and bit are parameters specific to periods t which vary over the life cycle, W is

the current discounted value of present and future income and current financial assets, Ptk*
current discounted price of good k in the future period t and Vit is the error term. Blundell
and Ray (1982) and Ray (1985) have proposed and estimated non-additive versions of LES.
Green et al. (1980) and Blanciforti et al. (1986) formulated dynamic version of the LES in
which habit formation has been incorporated.

4.6.2 Indirect Addilog Model (IAD)

The indirect addilog demand system or simply the addilog model is derived from additive
indirect utility function (Houtakker, 1960).
n
y
V ( P, y)   ai ( )b (4.21)
i 1 Pi

Where a and b are parameters with  ai  1; qibi  0 and 0  bi  1 . Using Roy‘s

identity to derive addilog demand functions,

y
ai bi ( )1bi
Pi
qi ( P, y )  n (4.22)
y
k 1
ak bk ( )bk
Pk
In the log form the addilog model is written as (Someryer, 1974; Theil, 1975)
y y
ln qi  ln ai bi  (1  bi ) ln( )  ln  aibi ( )bi (4.23)
Pi Pi
Which satisfies Engel aggregation and Cournot aggregation; and the substitution matrix is
negative semi definite given by b i  1 . However, only quasi-concavity of the utility
function is required for negative definiteness, which implies that at most one b can be equal
to minus one (Murty, 1982. The income and price elasticities of addilog demand systems
are given by Johson et al., 1984.

48
 The income elasticity is

eiy  (1  bi )  b j w j , for all i (4.24)

where eiy  b j w j

the own-price elasticity is

eij  (1  bi )  bi wi , for all i (4.25)

where 1  eii  0 with  1  bi  0 and wi  0 ;

and the cross price elasticity is

eij  b j w j for all i  j (4.26)

that depends only on the commodity, whose price is changing, and not on the good whose
quantity is responding. The complete set of demand parameter in addilog demand system
can be estimated with 2n -1 independent coefficient (n for bi‘s and n-1 for ai‘s.). A review
of demand systems that are approximations of true unknown demand structures such as the
Rotterdam Model, the transcendental logarithmic demand systems and the almost ideal
demand system is presented below.

4.6.3 Rotterdam Model

The demand system is started with specific algebraic demand system and then the general
restrictions are imposed to make it consistent with the theory of consumer demand (Theil,
1965, Barten, 1964). The relative price version of this system begins with Stone‘s (1954)
logarithmic demand function:

ln qi  ai  eiy ln y eij ln Pj . (4.27)

Writing the above equation (3.48) in differential form yield

d ln qi  eiy d ln Y   eij d ln Pj (4.28)

Multiplying by the budget share (2.50) can be expressed as

wi d (ln qi )( bi d ln Y   Cij lnPj , (4.29)

49
where
d ln Y  d ln Y   wk d ln Pk

 qi
bi  wi eiy  Pi ,
Y
and
Pi PjS ij
C ij  wi eij  .
Y
S ij is the (i, j) term in the Slutsky substitution matrix. The total differential of the

budget constraint is
n n

 Pi dqi  qi dpi  dY
i 1 i 1
(4.30)

In logarithmic terms,
n n

 wi d ln qi    wi d ln Pi   d ln Y
i 1 i 1
(4.31)

The basic idea underlying the Rotterdam model is to view the demand theory as a budget
sharing process for the consumer. Accordingly, budget shares and the changes in them are
of interest rather than the actual quantities consumed. Changes in value shares consist of
three components: changes in income, prices, and quantity consumed. Since changes in
income and prices are assumed to be given in demand theory, the only component
behaviorally determined is the changes in quantity consumed. Following Barten (1969), a
typical equation of the absolute price version of Rotterdam model can be written from
(4.31).

wit  ln qit  a i  b i  wkt  ln qkt   Cik ln Pkt (4.32)

k k

Where  stands for the first difference operator over time, q it and Pt are respectively the
quantity consumed of, and the price paid for the ith commodity in period t and t-1. a i , b i , Cik
(i, k = 1, 2, n) are the parameters interpreted as the intercepts, the income and price
coefficients, respectively.
The adding-up restriction in Rotterdam model implied
n n

 b i  1;
i 1
C
k 1
ik  0; k = 1, 2 . . . n.

50
The homogeneity can be enforced by imposing the restriction
n

C
K 1
ik 0

and the Slutsky symmetry restriction is given by

S ik  S ki .
The income elasticity is given by
b
e  i.
iy w
i

The income elasticities are positive if bi‘s are positive. Also ir could be noted that .
 
e  1 when b i  1 .
iy
 

Thus, if b i  wi , the commodity is a luxury item. The own price and cross price
elasticities are given by

(C ii  C ii b i  b i w i )
eii  (4.33)
wi

and

(C ijb j  b i w i )
eij  (4.34)
wi
The parameters of the demand system can be significantly reduced if additivity restriction
is further imposed (Johnson et al., 1984). Then, the required number of parameters is only
(n + 1) to form a complete set of demand elasticities.

4.6.4 Transcendental Logarithmic Demand System

Instead of starting with a specific indirect utility function, Christensen, Jorgensen and Lau,
(1975) approximate the true indirect utility function with a second order Taylor series
expansion. The indirect utility function of the trans log model is given by,

n Pj 1 Pj P
ln V  α0   a ji ln ( )  b ji ln ( ) ln ( i )
i 1 y 2 y y (4.35)

51
Using Roy‘s identity the trans log demand system can be written as
n
Pi
a m   bij ln (
)
y
wj  i 1
(4.36)
P
a m   b mi ln ( i )
j y
where
n n
a m   a j ; b m   b ji .
j 1 j

Thus, the demand system uses normalized prices with respect to income. Normalization
am  1 is imposed to identify the parameters of the consumer demand or expenditure
share equation in (4.35).
The income elasticity for the indirect translog demand system is given by

  b ij / w j   b ij
e jy  1 
i j i
. (4.37)
 1   b ji ln ( Pi / Y )
j i

The own price elasticities is given by

b jj / w j   b ji
e jj  1 
j
; (4.38)
 1   b ji ln( Pi / Y )
j i

And the cross-price elasticity is given by

b ji /w j   b ji
e ji 
j
. (4.39)
 1   b ji ln ( P / Y )
j i

The translog demand system has been widely in applied demand analysis (Christensen et
al., 1975; Jorgensen and Lau, 1975).

4.6.5 Almost Ideal Demand System (AIDS)

The almost ideal demand system (AIDS) that is introduced by Deaton and Muellbauer
(1980a) starts with a class of preference called the price-independent generalized

52
logarithmic (PIGLOG). The aim of using this class of preference is to ensure that the
necessary and sufficient conditions for consistent aggregation across consumers are
satisfied. The log of the cost or expenditure function is represented by the following
equiation:

log C( ; u) = (1- u) log a(p) + log b(p).

This cost function gives an arbitrary first-order approximation to any demand system. It
satisfy the axioms of choice, aggregates perfectly over consumers, has a functional form
which is consistent with household budget data, and simple to estimate and test the true
restrictions of demand theory. It also combines the best of theoretical features of both
Rotterdam and trans log models. The formulation of AIDS uses the duality theory and
expenditure function instead of utility or indirect utility function. By taking a specific
functional form for log a(p) and log b(p) as, then the AIDS cost (expenditure) is function
in natural logarithmic form is specified as

1
ln C (U,P)  a0   ai ln Pi    γ*ij ln Pj Pi  U b0  Pbi
i (4.40)
i 2 i j i
Where ai , b i and  *ij are parameters. U is the utility level and Pj are prices. The expenditure

function is linearly homogenous in P, provided

* *
 ai  1;   ij    ij   b j  0 .
j i

It is also consistent with aggregation over consumers. Differentiating the expenditure

function using Shepard Lemma yields


w j  a j   ij ln Pi  U b0 b j exp b jy ln( Pij )  (4.41)

Substituting for U , which is the indirect utility function derived the expenditure function
Y
w j  a j   aij ln Pi  b j ln ( ). (4.42)
P*

where,
1
ln P*  a0   ai ln Pi    ij* ln Pi ln Pj (4.43)
i 2 i j

53
is an overall price index, which could be replaced by Stone‘s (1954) index in empirical
applications since (2.41) is highly non-linear. The Stone‘s index is given by

ln P*   wi ln Pi (4.44)
i

When Stone‘s index is used in (4.41) the model is termed as linear approximation of almost
ideal demand system (LA/AIDS).

There are three sets of restrictions on the AIDS model given by

n

a
i 1
i  1,


i 1
*
ij  0,

b
I 1
i  0;


j
ij  0,

and
γij  γ ji

which should hold for the AIDS model to represent a system of demand equations (which
add up to total expenditure (  wi  1 ), are homogenous of degree zero in prices and total

expenditure, and satisfy Slutsky symmetry. The Slutsky coefficients are given by

S ij  γij  wi w j  wi δij (4.45)

where ij is the Kronecker delta (i.e. δ ij  1 when i = j and equals 0 when otherwise). The
Marshallian and Hicksian measures of elasticities can be computed from estimated
parameters of the LA/AIDS model as follows:

54
  ii
eii  1   bi (4.46)
wi

 ij w
j
eij  bi ( ) (4.47)
wi w
i
 ii
d ii  1  wi ; (4.48)
wi

and
 ij
d ij   ej (4.49)
wi  wi
where e denotes Marshallian elasticities and d denotes the income compensated or Hicksian
measure. Expenditure elasticities can be obtained using
bi
eiy  1  (4.50)
wi
The restrictions of demand theory can be imposed during estimation and tested easily with
AIDS. There are several applied studies using AIDS including Deaton and Mullbauer
(1980a) for Great Britain; Sergenson and Mount (1985), Hein and Pompelli (1985),
Blanciforti and Green (1983), and Hayes et al. (1988) for the United States; Mergos and
Donatos (1988) for Greece; Fulponi (1989) for France; and Ray (1980, 1982) for India.

4.7 Welfare Analysis

One advantage that dual approach offers in exploring the properties of optimal decision
rules is that it provides a way to investigate benefit cost analysis of any proposed policy
change. As expressed before, the expenditure function measures the household willingness
to pay to reach a certain utility level at given good‘s price level. So it suggests measuring
welfare effects of any economic change, for example price change.
So do the demand estimates from the AIDS model. As the model is derived from cost
function, it is useful not only because they can be used to construct price and income
elasticities but also useful to characterize the structure of the underlying cost function. By
using the estimated elasticities, we can infer in what direction and how much demand are
going to change. By using the information on consumer‘s preferences contained in the cost

55
function, we can evaluate the welfare effects of different price policies on households of
different income groups. Therefore, in this study attempts was made by static simulation,
to measure welfare losses for each income group under different pricing or subsidy
scenario. This includes changes in prices of commodity groups for which the government
of Indonesia intervenes directly or indirectly in fixing the consumer prices. Also changes
in prices of commodities mainly consumed by low income household. This may be single
and or multiple changes for example in price of rice, tobacco etc.

4.7.1 Welfare Measures

In analyzing demand function, cost functions and consumer demand functions themselves
provide the basis for welfare analysis of price changes. There are at least four alternative
welfare measures derived from theory of demand. These are:

 Index numbers (e.g. true index of cost of living)

 Marshallian consumer surplus
 Compensation variation
 Equivalent variation
Index numbers are commonly used for two reasons:
 as relative welfare measures (e.g., cost of living index, standard of living index);
and
 as means of generating price and quantity index for a commodity group (e.g., food,
capital, etc.).
A true cost-of-living index gives us an idea on how the minimum cost to consumers of
achieving a particular reference level of economic welfare changes as prices change. The
key assumption underlying the idea is that consumers behave as if they are minimizing the
cost of achieving any given level of economic welfare. N principle the objective of using
the index numbers is to present a summary of economic information about a group of
commodities.
For relative welfare measures, the group of commodities includes all commodities relevant
to an economic agent. For price and quantity index, the group of commodities involves
only a subset of all the commodities relevant to a decision-maker (e.g., food, capital, etc.).

56
Consumer surplus (CS), compensating variation (CV), and equivalent variation (EV) rely
on the demand functions. Consumer surplus uses Marshallian demand function, whereas
CV and EV use Hicksian demand function as a basis of analysis. In the absence of income
effect of price change, both Marshallian and Hicksian demand functions are identical. So
the measure derived through CS, CV and EV are the same? Conversely, if the income
effect of price change is present, this money metric resulted from the three is not equal.

4.7.2 Marshallian Consumer Surplus

Marshallian consumer surplus measures the change in welfare resulting from a price change
in monetary terms. It expresses the benefit that a household obtains from buying good.
This welfare measure can be described by way, in which demand of any household on a
certain good is expressed as a demand schedule.
Consumer‘s surplus equals the total that any household would pay minus the amount that it
actually does pay for the quantity bought.
It is expressed mathematically as follows:

CS    xi  p, m dpi  m
1
(4.51)
i 0

Where 0 represents an initial situation and the 1 is the final situation.

It should be noted that consumer‘s surplus as a measure of welfare consumer‘s welfare is
only valid consistently when the marginal utility of income is constant. This condition is
only fulfilled when the preferences is homothetic; and secondly if there is no income effect
when the price of a commodity changes.
As a welfare measure, the surplus derived from the Marshallian consumer‘s surplus are not
unique, i.e., path dependent (see Silberberg, 1978).

4.7.3 Compensating and Equivalent Variations

Introduced by J. R. Hicks (1942), compensating variation and equivalent variation are
money metric (monetary measures) of the gain or loss in consumer‘s welfare following an
economic change. The economic change may be introduced trough price change by
government or another shocks taking place in economic environment.

57
 Compensating Variation
Compensating variation is the compensating payment (amount of money) that leaves the
consumer as well of as before the economic change. It may positive or negative. It is
positive, if the economic change makes consumer worse off, and negative, if the economic
change brings betterment to the consumer.
Welfare analysis utilizing compensating variation approach uses past information to
estimates amount money needed to compensate Household to keep them stay at the level of
wellbeing before the (price) change. Technically it means a nominal money value required
to keep the consumers at utility level they enjoyed before he change. When a consumer‘s
situation is changed from situation 1 to situation 2, the compensation variation is defined as
(Just et al. 1982:85) „ the amount of income which must be taken an away from a consumer
(possibly negative) after a price and/or income change to restore the consumer‘s original
welfare level―. It is formalized as

CV  C u0 , p  u0 , p
1
  0
 (4.52)

ce the Hicksian demand functions are the derivatives of the cost function, integration
also gives the differences in costs of reaching the same level of well-being two
different price situations. And so
p0
CV     xi  p,u0  dpi  m (4.53)
i
p1

 Equivalent Variation

Equivalent Variation is the compensating payment that in the absence of the economic
change moves the consumer to the welfare level associated with the change. So EV is a
maximum amount the consumer would be willing to pay to avoid the change.
. Formally it is stated as

  
EV  c u 1 ,p 1  c u 1 ,p 0  (4.54)

In term of compensated demand function it is expressed as

p1

EV     xi p, u 1 dpi   m )
i
 (4.55)
p0

Hicksian demand functions can be derived from cost function using Sheppard‘s lemma:

58
The compensating Variation is especially important for policy analysis because it gives the
actual amount of money required to leave the consumer at least as well off as before the
change in the pricing policy.

59
 CHAPTER V
METHODS AND STATISTICAL PROCEDURES

This chapter presents the procedures applied in the estimations and related works. It
includes the discussions on data handling, classification of expenditure classes for rural and
urban households, the procedures of estimation and the statistical tests used. In addition,
this chapter presents also a discussion of econometric issues on the data used. Specific
section of this chapter addresses the estimation procedures for the AIDS Model.

5.1 Data Handling

Because the SUSENAS data was compiled not directly for the purposes of this study, we
need first to transform the data to meet the requirement for this study. This includes issues
on the grouping methods for the commodities, the issue on price of individual commodity
versus price of commodity-group, and the issue on the zero expenditure phenomena.

5.1.1 Commodities Grouping

For conciseness, and moreover for estimation reasons, we need in the empirical work a
small number of commodities to reduce the variables to be analyzed. Or, we need to
summarize the information through a grouping of the goods, when they display a similar
role in determining consumer‘s behavior. In addition, the price of close substitutes such as
meat, eggs, and fish, are very likely to move together, and hence grouping them into one
commodity would bring no serious problem. We need to group the goods, because there is
a believe saying, that nutritional superiority of any food or group of commodities may lead
consumers to make a priority of spending. Moreover, it is justified to assume, that cross
price effects among highly aggregated good is vanish; so that, the grouping of commodities
is justified (Theil, 1975),
Since economic theory does not provide any easy guidance on the number of composition
of food groups in an empirical work, we decided to group the commodity items on an ad-
hoch basis. However, the spirit of plausibility is highly respected. Accordingly, in this
study we grouped the food items based on the following considerations:

60
1. Nutritional content and sources. Based on this principle, food items with similar
nutritional constituents or sources (e.g. carbohydrate source or cereal, animal products,
etc.) were brought together into one commodity group;
2. The food price policy perspective: Food items being subject of food policy measure
were considered to be one group. Special for Indonesian case, the policy makers might
be interested to know the relationship between rice as a group to other food groups,
especially a group of foods assumed to be its substitute, like sweet potatoes, cassava,
wheat, sago and other starch containing food stuff. Because of that, these food goods
are then grouped to be a single group of food. Recently, there is also an interest to
know, if there is a potential for process foods (manufactured foods) to be the substitute
for rice. It might be the case, that through processing, food groups previously
considered to be inferior by Indonesian households have become upgraded culturally.
So that, it might become a substitute for instance, for rice. If this is the case, then food
diversification strategy may be achieved by manufacturing domestically endowed food
stuffs, like the ones mentioned above. To capture such information, one needs to have
a clear cut guideline in distinguishing the rice to manufactured food. This reasoning is
adapted into this study as strategy to compose the food group.
3. Consumption or expenditure pattern on food commodities, i.e. the substitution or
complementarity of food items.
4. The form of aggregation in which the data is available.
5. Consideration of a parsimonisity: This principle seeks to include a minimum number of
commodity groups with a powerful explanatory character. On this basis, thus, all non-
food expenditures for example, were aggregated into a single group.
6. The past studies of the Indonesian food sector.
7. Pattern of diet of the households, the behavior of which is under investigation.
8. The need to have relatively small group of food items.
In this study, non-food goods have been excluded from the demand systems by assuming
separability of the utility function. This exclusion should not be so harmful in the context
of a developing country like Indonesia, where a great portion of the budget goes to food
expenditure as shown by the following table. The exclusion of the durable goods group is
also based on the fact, that this study used a static model. To capture preference on durable

61
goods, one needs to cope with time dimension. This however, cannot be explained by a
static demand system since time dimension is very crucial in the decision to spend on a
durable good.
Based on these arguments it was decided to estimate a demand system for eleven
commodity groups. Food is composed of eleven (11) commodities groups: consisting of
rice (denoted as WR), non-rice staples (WNR), Fish (WFS), meat (WM), eggs and Milks
(WE), legumes (WL), fruits (WFR), oils and fats (WOL), tobacco and betel (TBCW),
prepared or manufactured food (WOPF) and spices and the miscellaneous (WSP). This
method of grouping is not based on knowledge about elasticities among them as suggested
by Hicks (1981), but rather based on our a priori knowledge about food needs and food
habits on the areas of studies, and the reasons mentioned above.
1. The food groups covered in the study are assumed to represent total food consumption
of the household. This may only be realistic assumption and therefore justified when
they contributed to a major expenditure of respondent being studied.

5.1.2 Price of Commodity

Conventional practice of cross sectional demand analysis focuses its attention on behavioral
change of consumers due to changing income level, household‘s demographical
characteristics, and space-related demand determinants, like e.g. rural vs. urban. However,
some studies have indicated, that also in cross section based analysis, estimating price
elasticities is possible.
The major problem, when possibility of estimating elasticities from cross section data is
proposed, concerns the degree of variation in price observed in this type of data set, and the
reason why that variation exists. The question weather or if sufficient price variation exists
to enable robust estimates of price elasticities to be made is actually empirical. So its
justification is based on the actual conditions of the population under investigation.
In the literatures, there are arguments maintaining the existence of price variation at any
point of time (cross sectional based variation): ―That there is considerable spatial variation
in prices in most developing countries should not be doubted‖ (Deaton, 1987).
1. Transport difficulties make it hard to bring the price in uniformity.

62
2. Price variation is there due to the fact that, as indicated by casual inspection, the price
of a commodity depends on where it is purchased. Some observation revealed that the
same good has different prices at different outlet in the market (Pratt et al, 1979);
3. Price variations reflect perceived or actual differences in quality, service agreements,
location, or information imperfections;
4. Furthermore, price variation on commodities are caused by (i) the nature of firm‘s cost
of production and weather they differ, (ii) the search strategy employed by consumers
and weather search costs differ across consumers (iii) the nature of the demand for
products;
Following assumptions meets the situation in East Java:
2 Price variation exists due to quality mix from one outlet to another at time of
purchasing. This is still the case in East Java, both in urban and especially in rural
areas: one warung15 - a most generally found outlet in East Java - may serve the buyer
differently. This difference creates a buying preference among potential buyers,
therefore one buyer may be loyal to one outlet, whereas the other buyers be loyal to the
other outlet;
3 Price variation is a reflection of quality effects, region, price discrimination, service
purchased with the commodity, seasonal effects, quality differences;
4 Price variation reflects opportunity cost of time and marginal cost/benefit of
information search;
5 Price variation may still exist as a reflection of cost of information, brand loyalty, brand
loyalties through distribution network.
The inclusion of price in the demand function estimation with a cross-sectional survey data
of household dated back on the works of Deaton (1978, 1988) and Cox and Wohlgenant
(1986). Deaton maintained that household surveys contain information on the spatial
distribution of prices, while Cox and Wohlgenant hold that knowledge of all factors
affecting price differences and price variation induced by region and season is desirable
from the standpoint of estimating commodity curves.
In this study, we assume that structure of demand is relatively constant, and consequently
price variation can attributed to changes in supply condition. It is to say that a range of

15
a traditional village-level outlet for foods and various consumer‘s goods.

63
prices for similar commodities can be generated, allowing estimation of cross -sectional
demand functions. Corresponding works with this assumption are those of Timmer and
Alderman (1979); Chernichovsky and Meesok (1982); Teklu and Johnson; Blundell et.al
1993).
The implied price of each good is calculated by dividing the total expenditure on each
Ei
commodity by the total quantity of commodity ( Pi  ). This is the definition of an
qi
implicit price. It does not necessarily reflect the marginal price that consumer face, but it is
the only information available from the observation indicating the price.

5.1.3 Price of Grouped Commodity

The data we have are on value and quantity of consumed food items such as quantity and
value of rice of type 1, quantity and value of mutton, number and value banana etc. The
households noted both these quantities and expenditures value during the survey.
Therefore, in the data we found for example, that a certain household for a certain period of
time spent 20 000 Rupiah (Indonesian currency) for buying 10 kilograms of rice. Dividing
the former by the latter which would be the unit value of rice could be used as an indicator
that the price of rice is Rp. 2000, - per kilo. It is then straightforward to derive the own- and
cross-price elasticities by running a regression of the quantity purchased on the unit value,
total expenditure of food, and several other characteristics.
In this study, some of food items for reasons described in the previous section are grouped
into any category. This handling creates the need of weighing price for each individual
items being grouped. In this study, this is done by weighing each of them according to their
share of consumption in their category. Likewise, the weighted price for each category is
the sum of weighted prices of each item in that category. Hence for any particular group (k)
consisting of n items, the price (Pk) is defined as
n wi
Pk   Pi
i 1 n
 w1
i 1

where wI is the share in the category or group being made.

64
5.2 The Problem of Missing Observation
Missing data of item‘s observation unit in each of household sample is a matter of fact in
this study. Although this makes the data set incomplete, they can still be used in the
analysis after some adjustment. There is a range of reference on this issue: Little and Rubin
(1987) provides researcher with techniques to cope with data missing phenomena.
Multivariate statistics Text of Tabacknick and Fidell (1996) address also problem of
missing data and its associated solution. The same theme may also be found in Cohen and
Cohen (1983). Various methods are there, to substitute missing data (e.g., by mean
substitution, various types of interpolations and extrapolations). Also, parries deletion of
missing data can be used.

5.2.1 The Problem of Zero Price

The problem of zero prices arises, when the information on unit value (price) is not
available for all items and households. In any survey, the case of uninformed unit value
(prices) occurs in two appearances: first when expenditure on a given food items are zero,
second, in the case of so called „other categories of food― in which various food items with
different measures are assorted as one category, that there is no unit value representing
price of that category. At the other side, in order to estimates a complete demand system,
the unit value (prices) must be available for all items, and for all households, regardless of
whether or not a particular household consume that good. To take care of this problem,
there is strategy proposed by Heien and Wessells (1988) and Heien and Pompelli (1988,
1989). This procedure was based on the estimation of the missing prices. The estimation of
missing prices was done by performing a regression of observed prices on regional
dummies and household total expenditures. The estimated prices replace the missing prices
in the estimation of the demand system.
This study applied a strategy of mean substitution, in which the missing price data is
substituted by an average value at kabupaten level. In this case, kabupaten represent a
cluster of households with the same supply condition.

65
5.2.2 The Zero Expenditure Problem
The problem of zero expenditure rises when some households do not consume a certain
food items or group of items being analyzed. An example mostly found in developing
countries is low income group of households do not consume meat and milk, that their
participation rates for meat and milks zero. Other typical example of zero expenditure
phenomena is found in an economy where certain value in the community restrains the
community to consume any food items. Typical for Indonesia for example is that Moslem
families abstain from eating pig meats (De Vega and Fisher, 1983). In that phenomena we
find that in the sample some households do not consume a certain food items or group of
items being analyzed. It means that certain proportion of households in this income group
have zero expenditures on these commodities.
Another possibility for the zero expenditure to exist in the collected data is due faulty
records. There are two explanations for that. Firstly, zero purchases can result from false
reporting by either respondent or enumerator. Secondly, the additional zeroes may arise
because purchases are not made frequently. During a one week survey period many
households record zero expenditure on many food items. An offsetting influence is
occasional relatively large purchases, many of these presumably to be stored, for later
consumption. Consumers who, prior to the survey, have made a recent purchase of
infrequently purchased item, and concerned that the expenditure will escape enumeration,
falsely record the purchase as having taken place during the survey period. However, in
common the cause of the zero expenditure is not known. Also from the SUSENAS data, it
was impossible to determine whether the household did not consume the particular products
at all or simply did not consume during the one week period.
In general, the phenomena of zero expenditure recalls a specific method in the estimation
process, because expenditure share of the commodity group to the total expenditures is a
dependent variable in the estimation of demand system. Solutions for this phenomena are
proposed by some authors. Cox and Wohlgenant (1986) applied a method to overcome
missing prices, by discarding first all incomplete observations, and estimate population
parameters using the remaining observations; and secondly, by using a zero order method
which substitutes an appropriate sample mean for missing values.

66
Other more complex solution includes first order methods, where missing data is estimated
in a more complex way and the missing value is viewed as unknown parameters The
parameters are estimated by least squares on the completed sample, but the method by
which the missing data are estimated has many variation (Amemiya (1984).
This study chooses to retain sample observations with zero expenditure or consumption
levels being replaced by sample mean value at Kabupaten level. This is done, in order to
portray adequately the full range of observed behavior. Because we maintain that price
differences reflect mainly commodity supply conditions, then average prices methods are
appropriate zero-order solutions for missing prices associated with these zero expenditures.
Therefore, we use this procedure to determine missing prices. That way, we assume that the
non-consuming households faced an average commodity price as maintained by Cox and
Wohlgenant (1986).

5.3. Classification of Households in Income Groups

After the data is segregated into two area-based groups, the rural and the urban data, each
of group is then further classified into three different income groups, by applying following
procedures.
1. We arrange the data to get the range.
2. We determine class interval
3. We determine to have three income classes
4. Based on above requirements we split the data into three income classes: the lower
income class, the middle, and the higher income class.
The following is an example from the rural area data set 1996.
 From the first step we got, that the range was 119 530
 For three class we got 119530 /3 = 39 000 (to be rounded). Based on this number, we
used a width of approximately Rp. 40 000, - for each income class.
 Next step is then to place the household sample into one of these three income classes.
 The constructed expenditure groups are treated as a household income class, which is,
assumed in this study to accommodate the household characteristics.
Applying this method we have three income groups with the following classification:
Income Group I < Rp. 45 000 per week

67
Income group II: Rp. 90 000 to < Rp. 130 000, - per week
Income Group III > Rp. 130 000 per week.

5.4. Modeling Demand System

5.4.1 Specification of System Model

One crucial step in empirical estimation process is to choose the working model. And for
that, there are many criteria for selection of demand model to follow. But the important
ones are: theoretical consistency, relative explanatory power of the model, simplicity and
ease of estimation (Wang, Halbrend and Johnson, 1996). An examination of the literature
reveals, that the class of complete demand systems, which both satisfy the theoretical
constraints and admit non-linear Engel curves is very limited. Three of them are the
Indirect Translog System of Christensen, Jorgenson and Lau (1975), generalized Linear
Expenditure System of Carlevaro, and Quadratic Expenditure System of Pollak and Wales
(1978 and 1980). The fourth type is the Almost Ideal Demand System (AIDS), a class of
demand systems proposed by Deaton and Muellbauer (1980).
The following assumptions are made in specifying system models:
1. The functional form is identical for all commodities in the system.
2. The households operate in competitive food markets;
3. Food items constitute a weakly separable branch of the household‘s utility function
which identical for all households. This is made to permits us to treat food
consumption as if the household pre-allocates a particular budget to food before
entering the markets and maximize its utility subject to this budget allocation. In that
case, we apply the two-stage budgeting assumption (Thomas, 1987).
According to Deaton and Mullbauer (1980) the AIDS model has advantages in view of
following characteristics.
1. It satisfies adding-up, homogeneity in price and income, and Slutsky symmetry;
2. Although highly non-linear in its parameters, it can be approximated into linear in
parameters;
3. The model gives arbitrary first-order approximation to any demand system;
4. AIDS model satisfies the axioms of choice exactly;
5. It aggregates perfectly over consumers;

68
6. It has functional form consistent with household budget data;
7. It is simple to estimate (in its approximation linear form);
It may be used to test for homogeneity and symmetry constraints;
The choice of demand system in this study is based on (1) the agreement to theoretical
constraints, (2) flexibility of functional form which is necessary for confronting them with
the micro data used in this study.
The explanatory power of the AIDS models has been tested in both developed and
developing countries context, that some author having research experiences of food demand
studies in developing countries recommended the use of the AIDS model in studying
demand in developing countries. Based on these advantages, this study applied the AIDS
model to the existing data. Deaton and Muellbauer (1980) approximated the cost function
of the price-independent generalized logarithmic (PIGLOG) class of preference, with the
following cost function defined as a flexible functional form:

Log C (u, p) = (1 - u) log {a (p)} + u log {b (p)} (5.1)

where u is the utility level lying between zero and one, p is the price vector, and a(p) and
b(p) are the cost of the subsistence and the bliss, respectively.
1
log a ( p)   o   i log pi    ij* log pi log p j (5. 2)
i 2 i j
Log b (p) = log a ( p)   Pi βk , (5. 3)
k
where i and j indicate the ith and jth commodity (i and j = 1, 2, . . . n). Thus the AIDS cost
function can be written as:
1
Log c (u, p) = α0  i αi log pi  i  j γ* log p log p j  U β0 πi piβi (5.4)
2 ij i

where  0 ,  i ,  ij* are parameters, U is utility level, and p i are the prices of the ith and the jth

commodities.
The AIDS model in budget share is obtained by firstly differentiating the cost function with
respect to log price followed by substitution of U by using the cost function. The model
specified bellow corresponds broadly to that of Blanciforti and Green (1987)16

16
detailed derivations of the model are available in Deaton and Muellbauer (1980)

69
 Y
wi  αi   γ log Pj  βi log
ij P (5.5)
th
where w is the budget share, Pj is price for j good and P is an overall price index defined
in terms of individual prices as follows:
1
log P  α0  α j log Pi    γij log Pi log Pj . (5.6)
2 i

In this function the adding-up restriction implies:

 i  1,
i

i
i  0, 
i
ij  0. (5.7)

The equation system (5.6) with the adding up restriction (5.7) constitute the unrestricted
system. For it to be consistent with utility theory, the following additional restrictions must
hold:
1. homogeneity :


j
ij  0 , and

2. symmetry :
 ij   ji
.
where  i , βi , γij (i. j = 1,2, ..., n) are parameters.

When we use the Stone‘s (geometric) price index as P, then we can avoid the non-linear
estimation (Deaton and Muellbauer (1980), and we get so called the linear Approximate
AIDS (LA/AIDS) (see Blanciforti and Green, 1983).
So instead of using P we used P*, defined as:

LogP  log P*   0   wk log Pk (5. 8)

5.4.2 Incorporating Income Group into the Linear AIDS

To consider the effect of household size in the system, the size variable is introduced in the
model:
Y
w  α   γ log P  β log ( )  θ log S (5. 9)
i i ij j i i
P*
or
wi  αi  βi ( logY  log P* )  j γij log p j  θ i log S (5. 10)

70
The reason for allowing household size into the model is because our data set covers
households with widely different demographic characteristics. In this study however, only
household size represents household characteristics. It is also worth noting, that concerning
the household characteristics, effort has been made to find a scale by which different family
members be assigned different scale according to age. This is what one call equivalent
scale methods (see Pollak and Wales, 1981, 1992). Another notice that should be taken
into account is the effects of individual factors (tastes, habits, expectations, experiences,
and other unobservable variables). These may affect the consumption composition of the
household. Technically speaking, if these variables are omitted from the model, their
effects should be embedded into the disturbances assuming that their effects sum to zero.
This is what we followed in this study. So, we let the household characteristics be
indicated by one qualitative variable only that is the income group. Assuming this to be
true, we let D to represent income group in the model, and therefore is embedded as
dummy variable.
Y
wi  (αi  δiei )  ( β  δiei ) log ( )   j γij logp j  θi log S (5.11)
i *
P
The mathematical formulation of the linearized approximated AIDS equation is presented
as follows:
WR = α WR + D1WR + D2WR + D3WR + 1lPr1WR + 2 lPr2WR + 3lPr3WR + 4lPr4WR + 5 lPr5WR +6 lPr6WR
+ 7 lPr7WR + 8 lPr8WR + 9 lPr9WR + 10 lPr10WR + 11 lPr11WR +  LYPWR + 1 D1LYPWR +
2 D2LYPWR + 3 D3LYPWR +  lJARTWR + eWR.

WNR = α WNR + D1WNR + D2WNR + D3WNR + 1lPr1WNR + 2 lPr2WNR + 3lPr3WNR +4lPr4WNR +5 lPr5WNR+
6 lPr6WNR + 7 lPr7WNR + 8 lPr8WNR + 9 lPr9WNR + 10 lPr10WNR + 11 lPr11WNR +  LYPWNR +
1 D1LYPWNR + 2 D2LYPWNR + 3 D3LYPWNR +  lJARTWNR + eWNR.
WFS = α WFS + D1WFS + D2WFS + D3WFS + 1lPr1WFS + 2 lPr2WFS + 3lPr3WFS + 4lPr4WM +5 lPr5WFS +
6 lPr6WFS + 7 lPr7WFS + 8 lPr8WFS + 9 lPr9WFS + 10 lPr10WFS + 11 lPr11WFS +  LYPWFS +
1 D1LYPWFS + 2 D2LYPWFS + 3 D3LYPWFS +  lJARTWFS + eWFS.

WM = α WM + D1WM + D2WM + D3WM + 1lPr1WM + 2 lPr2WNR + 3lPr3WNR + 4lPr4WNR +5 lPr5WNR +

6 lPr6WNR+ 7 lPr7WNR+8 lPr8WNR +9 lPr9WNR + 10 lPr10WNR + 11 lPr11WNR + LYPWNR +
1 D1LYPWNR + 2 D2 LYPWNR+3 D3 LYPWNR + lJARTWNR + eWM.

WE = α WE+D1WE+D2WE+D3WE + 1lPr1WE+2 lPr2WE+3lPr3WE+ 4lPr4WE+ 5 lPr5WE + 6 lPr6WE +

7 lPr7WE + 8 lPr8WE + 9 lPr9WE + 10 lPr10WE + 11 lPr11WE + LYPWE +  1 D1LYPWE + 2 D2
LYPWE + 3 D3 LYPWE +  lJARTWE + eWE.

71
WL = α WS + D1WL+ D2WL+ D3WL + 1lPr1WL+2 lPr2WL+3lPr3WL+ 4lPr4WL+ 5 lPr5WL + 6 lPr6WL +
7 lPr7WFS + 8 lPr8WFS + 9 lPr9WFS + 10 lPr10WFS + 11 lPr11WFS + LYPWFS + 1 D1LYPWFS + 2
D2 LYPWFS + 3 D3 LYPWFS +  lJARTWFS + eWL.
WFR = α WFR + D1WFR + D2WFR + D3WFR + 1lPr1WFR + 2 lPr2WFR + 3lPr3WFR + 4lPr4WFR + 5 lPr5WFS +
6 lPr6WFS + 7 lPr7WFS + 8 lPr8WFS + 9 lPr9WFS + 10 lPr10WFS + 11 lPr11WFS +  LYPWFS +
1 D1LYPWFS + 2 D2LYPWFS + 3 D3LYPWFS +  lJARTWFS + eWFS.

WOL = α WOL + D1WOL + D2WOL + D3WOL + 1lPr1WOL + 2 lPr2WOL + 3lPr3WOL + 4lPr4WOL + 5 lPr5WOL +
6 lPr6WOL + 7 lPr7WOL + 8 lPr8WOL + 9 lPr9WOL + 10 lPr10WOL + 11 lPr11WOL +  LYPWOL +
1 D1LYPWOL + 2 D2LYPWOL + 3 D3LYPWOL +  lJARTWOL + eWOL

WTB = α WTB + D1WTB + D2WTB + D3WTB + 1lPr1WTB + 2 lPr2WTB + 3lPr3WTB + 4lPr4WTB + 5 lPr5WTB +
6 lPr6WTB + 7 lPr7WTB + 8 lPr8WTB + 9 lPr9WTB + 10 lPr10WTB + 11 lPr11WTB +  LYPWTB +
1 D1LYPWTB + 2 D2LYPWTB + 3 D3LYPWTB +  lJARTWTB + eWTB.

WOPF = α WOPF + D1WOPF + D2WOPF + D3WOPF + 1lPr1WOPF + 2 lPr2WOPF + 3lPr3WOPF + 4lPr4WOPF +

5 lPr5WOPF + 6 lPr6WOpf + 7 lPr7WOPF + 8 lPr8WOPF + 9 lPr9WOPF + 10 lPr10WOPF + 11lPr11WOPF +
 LYPWOPF + 1 D1LYPWOPF + 2 D2LYPWOPF + 3 D3LYPWOPF +  lJARTWOPF + eWOPF.

WSP = α WSP + D1WSP + D2WSP + D3WSP + 1lPr1WSP + 2 lPr2WSP + 3lPr3WSP + 4lPr4WSP + 5 lPr5WSP +
6 lPr6WSP + 7 lPr7WSP + 8 lPr8WSP + 9 lPr9WSP + 10 lPr10WSP + 11 lPr11WSP +  LYPWSP +
1 D1LYPWSP + 2 D2LYPWSP + 3 D3LYPWSP +  lJARTWSP + eWSP.

Table 5.1: Variable Description and A-priory Hypotheses

VARIABLE DESCRIPTION/
CODE VARIABLE NAME
COMPOSING FOOD ITEMS
WR rice groups Domestic rice, imported rice, sticky rice.
WNR Non-rice staple Corn, cassava, sweet potatoes, talas, Sago
WFS Fish group Sea and fresh water fishes, shrimps, squids; crabs:
fresh, preserved, canned.
WM Meat group Beef, buffalo Beef, Mutton, Pork, Chicken
WE Eggs and Milk Eggs, Milk and milk products
WL Legumes/Nuts Peanut, Soya Beans, Mung Bean Cashew Nut, Soya
cakes (Tempe), Tofu
WFR Fruits and Vegetables vegetables, and fruits: fresh and preserved, canned
WOL Edible Oil and Fat Cooking oils, margarine, coconuts
WTB Tobacco and Betel Clove filtered cigarettes, unfiltered cigarettes,
cigarettes, tobaccos, betel, cigars
WOPF Prepared Foods Bottled water ( carbonated, non-carbonated),
alcoholic beverages, energy enriched beverages,
packed cakes & foods, syrups, breads, instant
noodles, packed served foods (fried chicken, fried
rice, sate salads), snacks, ice cream.
WSP Spices, Miscellaneous Salts, candle nut, coriander, pepper, nutmeg, cloves,
fish paste, Soya sauce, tomato sauce, packed-and
mixed spices, crispy, Getup chips, macaroni. Etc.

72
Dj Income Group Dummy, 1 : lower group; 2: middle group; 3: higher group
J=1,2,3
LPrij Log of Price of Food Prices (unit values) of the estimated food groups
Group
I, j = 1- 11;
LYPi Log of Total Food
Expenditure
LJART Log of Household Size
1 Expenditure Parameter Total expenditure
α Parameter Intercept coefficient
 Price Parameters Price coefficients
i Di Income group Parameters Coefficients of the lower, middle, and higher income
groups
 Parameter Coefficient of household size
e Disturbance parameter

5.5 Estimation Procedures

The LA/AIDS incorporating household size of equation (5.13) will be the empirical version
to be estimated. For empirical implementation, any demand system model must be
embedded in a stochastic framework. A disturbance term for each equation in the system is
required since some factors not implicitly introduced into the model may influence
household consumption behavior. The stochastic assumption are that E ( e i ) = 0 and E

( e i e i1 ) =  ii1 . Where e i is an n x 1 vector and  ii1 the Kronecker product. It means
that error term is assumed to have expectation zero, to be uncorrected across commodities
and have a contemporaneous variance-covariance matrix  . Due to the adding up
conditions (the sum of budget shares equals one) the variance-covariance matrix of the
disturbance term is singular. Each disturbance term can be written as a linear combination
of the remaining disturbance terms. The singularity of variance-covariance matrix 
prohibits the estimation of the demand function by system approaches. To overcome this
singularity, it is necessary to delete arbitrarily one commodity from the full set.
To estimate the LA/AIDS while imposing both homogeneity and symmetry, Zellner‘s
Iterative Seemingly Unrelated Regression Estimation (ITSURE) will be used. The
seemingly unrelated regressions methods may improve the efficiency of parameters
estimates when there is contemporaneous correlation of errors across equations (Zellner,

73
1962). This permits cross-equation restrictions to be imposed and with the iterative
solutions estimates are independent of the deleted equation (Barten, 1969).
In this study, the equation for the last commodity (other) will be dropped in order to form
the joint density function. The computer program available for estimating of equations in
SAS (Statistical Analytical System) program is called ITSUR (iterative seemingly unrelated
regression or iterative joint-generalized least square). This program obtained the
contemporaneous correlation matrix by using OLS residual and the final parameter
estimates take this information into account.

5.6. Computation of Elasticities of Demand

The elasticities of particular interest are own and cross price, expenditure, and household
size elasticities of expenditure. The main advantage of the use of elasticities is that these are
independent of the units measurements, thus results are comparable even if derived for
countries with different currencies or where the commodities are measured in different
physical units. The formulae and procedure used for calculation of elasticities and related
statistics of the AIDS model in this study followed Teklu and Jonhson (1988), and Green
and Alston (1990).

Expenditure elasticities

βi  βi'
ε iy  1 (5.12)
wi
Own price elasticities
γii  ( βi  βi' ) wi
εii  1 (5.13)
wi
Cross price elasticities
γij ( βi  βi' ) w j
ε  (5.14)
ij wi

74
 θi  ( βi  βi' )
Household Size: ε is  (5.15)
wi
The compensated or hicksian price elasticities are derived by transforming the ordinary or
marshallian price elasticities through the Slutsky equation. Thus, the compensated own
price elasticities becomes
εiiH  εii  wi . εiy (5.16)

and the compensated cross price elasticities becomes

εijH  εij  w j . εiy , (5.17)

where εii and εij are the compensated own price and cross price elasticities respectively,
H H

the rests are defined as previously.

The price and household size elasticities for the standard model, i.e. equation (5.11) without
income group variables, can be obtained by imposing β i1 = 0 in equation (12) to (17)
respectively. The elasticities will be calculated at the mean sample, assuming the mean
budget share fixed. The standard errors of the elasticities will be calculated using the usual
formulas for the distribution of linear transformation of a normally distributed random
vector. Equations (5.13) to (5.17) can be rewritten in matrix form as:
ε  Ab (5.18)

where ε is the vector of estimated elasticities ( ε iy ,ε ii ,ε ,ε is ), b is the vector of estimated

ij
AIDS model parameters (  ' s,  ' s, and  ) and A is a matrix. The variance covariance of ε ,
Var( ε ) is
Var ε   AVar bc A T (5.19)

where Var (b) is the variance-covariance matrix of b.

The values of all elasticities used in this study therefore are not independent of the
distribution of the budget share. Unlike the LES model, the AIDS model allows the
expenditure elasticity to decrease with respect to a decrease in the budget shares for the
necessities (  i  0). Expressed mathematically this is to say
 εiy β
  i2  0 for  i  0 (5.20)
 wi wi

75
Thus, in this situation, the AIDS model possesses a more desirable property than the LES.
Concerning the properties of the own-price elasticities in the AIDS, the sign of  εii  wi

depends on the relative magnitudes of  ii and  i w i . A priori, it is difficult to assign a

positive or negative value to change in  ii with respect to a change in the budget share, w i
(Blanciforti and Green, 1983).
The method selected for calculation of elasticities for the AIDS model is important if
reliable results are to be expected. In the test of alternative formulae for the calculation of
the elasticities of demand in the AIDS models used by analysts, Green and Alston (1990)
concluded that not all were reliable and correct. According to Green and Alston, the
elasticity estimates for any commodity are similar across the AIDS model and the linear
approximation of the AIDS model using formulae (5.13) to (5.17). Green and Alston
recommended these estimators as the ones, which provide similar elasticities to the AIDS
model.

5.7 Tests of Restrictions

Test of restriction embodies our study and concerns to answer the questions of the
reliability model assumptions. One advantage of using AIDS model is that it gives the
researcher possibility to impose and test the validity of underlying consumer theory in
economics. The main interest in the inference with regard to the estimation of demand
system is the question if the underlying theory is supported by the existing data. As
explained previously, symmetry, adding-up and homogeneity are restrictions that can be
tested and imposed in the AIDS model. For that purpose test is conducted using the
likelihood ratio test. The test statistic Likelihood Ratio (denoted by) is the ratio of the
maximum value of two likelihood ratios under the more restrictive hypothesis (L) to that
of a less restrictive hypothesis (L). Symbolically, if (L) and (L) are the maximum
values of the likelihood function with and without a set of specified restrictions
respectively, the likelihood ratio is defined as:
Lω
λ .
LΩ
For the Null Hypothesis of the form
H0: R = r,

76
as in the case of this study, where R is a (J x K) matrix that selects the appropriate elements
from  so as to specify the linear combinations of  that are of interest; r is a (J x 1) vector
of value to which we hypothesize that the linear combinations are equal, with normally
distributed error; It can be shown that the  may be simplified into: LR = T (in SSER -
lnSSEU)
In other words, the likelihood ratio test statistic can be written in terms of the restricted and
unrestricted sums of square errors and have a 2 distribution with J (number of restriction)
degree of freedom. (see for instance Griffiths et al, 1993). For normally distributed
disturbance,

 T/2
 ˆ
 
 ω 
λ  ,
 ˆ 

 Ω 
or

ˆ
– 2 Log  = T ( Log  ˆ ).
 Log 
ω Ω

Under the null Hypothesis that the restriction valid, the less restricted L is distributed,
asymptotically as a chi square distribution with degree of freedom equal to the number of
restrictions to be tested.
Decision Procedure
To test the restrictions the following procedure is applied:
1. We calculate the determinant of variance covariance matrix of the residual of the model
with and without restrictions. When using ITSUR of SAS program, this estimate, terms
as S matrix, provided as part of estimation result given out by the program.
2. We form the ratio the lambda, which is () = L0/L1. This ratio is always between 0
and 1 and the less likely the assumption is, the smaller  will be. This can be quantified
at a given confidence level as follows:
3. We calculate the Chi-square, which 2 = -2 ln. The smaller is, the larger 2 will be.
4. We can tell when 2 is significantly large by comparing it to the upper 100 × (1- )
percentile point of a Chi Square distribution with k degrees of freedom. 2 has an
approximate Chi-Square distribution with k degrees of freedom as defined previously

77
5. The likelihood ratio tests computes 2; and rejects the assumption, if 2 is larger than a
chi square with k degrees of freedom percentile, where the percentile corresponds to the
confidence level chosen by the analyst.

5.8 Welfare Analysis

As explained previously, three methods are available for the measurement of welfare
change. They are the consumer surplus (CS) concept, the compensating variation (CV)
comcept, and the equivalent variation (EV) concept. Technically, consumer surplus is
relevant to our analysis; if we were secured that income effect of price change is zero.
Since we know, that this is not the case (compensated and ordinary price elasticities are not
the same), we use rather welfare measure based on Hicksian demand function. Between CV
and EV, CV is preferable, because it allows us to make an ex – ante analysis of welfare
change. To measure changing household‘s welfare from a changing policy, welfare must be
measurable. However, utility which indicates welfare is not directly measurable. So an
alternative measure must be chosen. The CV measure is based on new prices, and the EV
measure is based on initial prices. Information on the distribution of welfare gains and
losses among household groups should be useful to policymakers in making judgments on
whether this policy result is inferior or superior to an alternative policy result. Since we
refer to the new price in our analysis, Compensating Variation was used in this study.

 Compensating Variation
To find money metric expression for this measure from the observable data, one who
follows the path of analysis should be going from the demand function back to the
underlying cost function.
Since the Hicksian demand functions are the derivatives of the cost function, integration
also gives the differences in costs of reaching the same level of well-being from two
different price situations. And so
p0
CV =   xi ( p,u 0 ) d pi x  m (5.21)
i
p1

78
 Compensating Variation Measurement of Proposed Price Change
The general expression of (5.14) is

CVi  C (ui0 ,pi1  ui0 , pi0 ) (5.22)

where: C Vi = compensating variation of a price change for the i th income group

th
u i0 = original utility level for the i income group

pi0 = original mean price vector for the i th income group

p1i = new mean price vector for the i th income group.

Both vector of prices are observable (the original vector of prices is the observed data, and
the new vector of prices is set exogenously) but utility levels are not.
As can be seen, the expression for CV contains the utility level u which not observable. In
order to estimate the CV‘s by income groups, we need to transform the utility function into
money metric indirect utility function m (P, V), namely, the income (expenditure) needed
to attain utility level v at the vector of prices P. This expresses consumer‘s willingness to
pay to attain the said utility level.

The calculation of an observable utility level may be done by employing the indirect utility
function. This is conducted by transforming the expenditure or underlying cost function for
the AIDS:
ln m  a  p 
ln c  p, u   a  p   u  b  p   u  p, Y   . (5.23)
b p 
Expressed in detail, it becomes
 11
1 11 11 
lnm  (α0   α j ln Pj    γ jk lnPj lnPk )
Uo V (p, m) =  
j  1 2 j 1 k 1


β0  (pk ) βk
 (5.24)

where m 0i  mean of original income for the i th income group and

11 11 11
ln m = α 0   ln Pj  12   γ jk ln Pj ln Pk  U  Pjβi (5.25)
i 1 j 1 k 1

79
In the optimum condition, it holds that c = m
To do so,

mi0  c(u i0 , pi0 ) (5.26).

Equation (5.24) can be used to estimate the money metric value of u at starting and end
points of any economic change.
For the CV‘s calculation, we can use the estimated results of equation (5.11), namely
α i ,γ ij , and β i .

Finally, the CV‘s for each income group are found by subtracting the value of the original
cost functions ( m 0i ' s ) from the value of the new cost functions (found by replacing the new

vector of prices and the original utility levels U i0 in equation (5.24).

80
 CHAPTER 6
ESTIMATION RESULTS AND DISCUSSION

This chapter presents the results of estimation we made on the linearized AIDS Model
applied on data bodies we have. The model describes the consumption behavior of
Households in rural and urban - East Java recorded in four rounds of the SUSENAS-
survey. The estimation is conducted using the SAS program version 6.12. The estimation
applied the iterative seemingly unrelated regression (ITSUR) procedure. This procedure
allows the estimation of eventually contemporaneous correlation in error terms across
equations, which then to be used to derive more efficient estimates.
A descriptive statistics, in terms of mean of budget share of each food groups to the total
food expenditure was derived as a part of the estimation‘s results. They are presented in the
first section, as a prelude for the assessments to follow. Its importance lies in giving an idea
on the dominance of each group relative to the other. The mean values presented in Table
6.1 were used to estimate the point elasticities.

6.1 The Budget Share

Table 1 presents the average budget share of each food groups across survey rounds and
areas. This information is important to give an idea on the composition and relative
significance of each food groups to total household‘s expenditure on food.
Information on food share is an important element in assessment of the results for policy
designing.
A quick glance at the share numbers given in Table 6.1 shows that rice assumed the most
important role followed by prepared foods and tobacco and betel, while the rest of food
groups shares a fairly balanced portion to the total household‘s food expenditure. With
respect of area, rice took a bigger portion of household‘s food expenditure in rural than in
urban areas. This share declined only slightly in the course of surveys. It means that
irrespective of time and space, rice constitutes a main menu for households in East Java,
Indonesia.

81
 Table 6.1 Shares of Food Groups (mean value) to Weekly Total Food
Expenditure, East Java, Indonesia

1990 1993 1996 1999

FOOD GROUPS
Urban Rural Urban Rural Urban Rural Urban Rural
Rice 0.28 0.29 0.21 0.26 0.22 0.27 0.21 0.25
Non Rice Staple 0.03 0.09 0.02 0.06 0.02 0.06 0.02 0.07
Fish 0.05 0.06 0.06 0.06 0.06 0.06 0.06 0.05
Meat 0.06 0.09 0.09 0.08 0.08 0.08 0.07 0.08
Eggs and Milk 0.08 0.04 0.07 0.05 0.06 0.04 0.05 0.04
Legume 0.06 0.05 0.06 0.06 0.06 0.05 0.06 0.05
Fruit & vegetable 0.04 0.04 0.04 0.03 0.04 0.03 0.04 0.04
Edible Oil 0.05 0.05 0.05 0.06 0.05 0.06 0.05 0.06
Tobacco & betel 0.10 0.08 0.11 0.10 0.12 0.10 0.11 0.09
Prepared Foods 0.21 0.17 0.26 0.20 0.27 0.22 0.31 0.24
Spices 0.04 0.04 0.03 0.04 0.02 0.03 0.02 0.03
Sources: own calculation, from the SUSENAS, 1990, 1993, 1996, 1999

With respect to the share of prepared food, its relative importance may be attributed to its
many items that composed it. Prepared food is a mix of all food goods that households did
not prepared by themselves. It includes mainly bottled or packaged drinks and snacks in
various forms, mostly offered to households in line with increasing numbers of food
manufacturing industries. The inclusion of these items into one composite in this study is
made, to capture the consumption attitude of households towards manufactured foods. It
may be interpreted as their willingness to ―diversify‖ their daily menu into more various
brackets. Food diversification strategy gains an increasing attention from policy makers in
Indonesia, due to an increasing concern of rice scarcity in the country.
A close inspection into the share‘s number in the table shows us a consistently increasing
portion of prepared foods to the total household‘s budget for food in the course of the
surveys. It holds for both rural and urban areas. Too, the portion of household‘s budget
disposed to prepared foods is bigger in urban areas than in rural areas.
Tobacco and betel in this study represent convenience goods which take an important role
in household budget. Table 6.1 shows that Tobacco and betel ranked the third in term of

82
household‘s budget share. The fact that the budget share of tobacco and betel assumes
relatively high amount in both areas indicated the importance of local specific convenience
goods in the household‘s pending pattern.
A combined picture of a slightly declining rice share and increasing share of prepared foods
is consistent to the expectation. It is also desirable in perspective of food diversification is
perceived.
However, the consumption pattern of households in East Java still follows a typical pattern
of consumption in Indonesia in general, in which rice constitutes a single main food groups
to the total household food budget. It implies that food policy should be still focused on rice
as a main agenda. Rice policy thus, remains an important and crucial agenda of this sector,
for political and economic reasons.

6.2 The Model’s Performance

The demand system that we specified consists of budget share to the total food expenditure
of the following food groups: rice, non-rice staples, fish, meat, eggs and milk, legumes,
Fruits and Vegetables, Edible Oils, tobacco and betel, prepared food , and spices and
miscellaneous.
Demand estimates for spices and miscellaneous food group are not estimated directly,
because it was dropped from the system. Instead, they were estimated by using adding up
principle. These eleven food groups accommodated all items recorded in the questionnaire
of SUSENAS module for household expenditure survey. It is worth noting, that the group
of prepared foods into this group is motivated by the expectation that this group may
represent the consumer‘s preference on processed foods that possibly be substitutive for
rice. This type of information is needed to address issues of food diversity for rice. When
prepared food is substitute for rice, and has more than unity elasticity of income, then it
seems to be reasonable to hypothesize that the promotion of processing food industry might
be means to cut the dependence of people on rice. Tables 6.2 to 6.9 display the parameter
estimates resulted from the model estimation. The estimation was done with the imposition
of homogeneity and symmetry restrictions. The adding-up restriction is automatically
imposed.

83
 Table 6.2 Parameter Estimates for the LA/AIDS Model
Based on the 1990 - SUSENAS Micro Data: Urban East Java, Indonesia

Income Group Based- Dummy COMMODITY PRICES

MODEL Intercept
Lower Middle Higher Rice N.Rice.S Fish
Rice 0.43617*** 0.13790** -0.10532 -0.03258 0.09962*** -0.00793** -0.01435***
Non Rice Staple -0.03085 0.01261 -0.04879 0.03618 -0.00793** 0.00947*** -0.00192**
Fish 0.10379*** -0.03910 0.04389 -0.00479 -0.01435*** -0.00192** 0.01673***
Meat 0.17200*** -0.12176** 0.10025 0.02151 -0.00661 0.00078 0.00126
Eggs and Milk 0.11893*** -0.06385 0.20018 -0.13633*** 0.00369 0.00392*** 0.00027
Legume 0.11005*** -0.04797 0.18887 -0.14090*** -0.00575 0.00050 -0.00219
Fruit & Vegetable -0.04509 0.00903 -0.01951 0.01048 -0.01720*** -0.00281* 0.00357***
Edible Oil 0.09978*** -0.02950 0.07468 -0.04518** 0.01809*** 0.00071 -0.00128*
Tobacco 0.05767 -0.02161 -0.09180 0.11341* -0.03402*** -0.00238 -0.00151
Prepared Foods -0.00978 0.14303** -0.27308 0.13005** -0.03082*** -0.00062 -0.00294**
Spices c -0.01267 0.02122 -0.06937 0.04815 -0.00472 0.00028 0.00236

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice -0.00661 0.00369 -0.00575 -0.01720*** 0.01809*** -0.03402***
Non Rice Staple 0.00078 0.00392*** 0.00050 -0.00281** 0.00071** -0.00238
Fish 0.00126 0.00027 -0.00219 0.00357*** -0.00128* -0.00151
Meat 0.01404*** -0.00297* 0.00498 0.00400 -0.00498** 0.00552
Eggs and Milk -0.00297* -0.00499*** 0.00037 -0.00407*** 0.00089 0.00361*
Legume 0.00498 0.00037 0.01557*** 0.00116 0.00060 -0.01272***
Fruit & Vegetable 0.00400 -0.00407*** 0.00116 0.01735*** -0.00564^*** 0.00371
Edible Oil -0.00498** 0.00089 0.00060 -0.00564*** -0.00507*** 0.00026
Tobacco 0.00552 0.00361* -0.01272*** 0.00371 0.00026 0.03555***
Prepared Food -0.01391*** -0.00193 -0.00366** -0.00188 -0.00291*** -0.00058
Spices c -0.00210 0.00119 0.00114 0.00182 -0.00068 0.00255

COMMODITY PRICES Total Dummy Effect on Budget Share

House-hold
MODEL Budget Size
Prepared
Spices Shares Lower Middle Higher
Foods
Rice -0.03082*** -0.00472** -0.12972*** -0.03731* 0.03561 0.00169 0.17013***
Non Rice Staple -0.00062 0.00028 0.01845*** -0.00036 0.01386 -0.01350 0.00186
Fish -0.00294** 0.00235*** -0.01826** 0.01113 -0.01460 0.00347 0.00356
Meat -0.01391*** -0.00210* -0.03429*** 0.03257*** -0.03615 0.00358 -0.01343***
Eggs and Milk -0.00193 0.00119* -0.00986 0.01588 -0.05472 0.03884*** -0.01556***
Legumes -0.00366** 0.00114 -0.01593* 0.01038 -0.04685 0.03648 -0.00544
Fruit & Vegetable -0.00188 0.00182* 0.03572*** -0.00537 0.00694 -0.00157 -0.02746***
Edible Oil -0.00291*** -0.00067 -0.01338*** 0.00895 -0.02050 0.01154* -0.00021
Tobacco -0.00058 0.00255* 0.02912** 0.00399 0.02482 -0.02881 -0.03601***
Prepared Foods 0.06152*** -0.00226*** 0.11961*** -0.03520** 0.07227 -0.03707*** -0.06646***
Spices c -0.00226 0.00043 0.01854 -0.00467 0.01931 -0.01464 -0.01098
Note:*:p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

84
 Table 6.3 Parameter Estimates for the LA/AIDS Model
Based on the 1990 - SUSENAS Micro Data: Rural East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.82731*** -0.22646*** 0.54562 -0.31916*** 0.06853*** -0.01220*** -0.01132***
Non Rice Staple -0.35235*** 0.31185*** -0.57096 0.25911*** -0.01220*** 0.02004*** -0.00197
Fish -0.03556*** 0.04994*** -0.11649 0.06655*** -0.01132*** -0.00197*** 0.01742***
Meat 0.21759*** -0.12273*** 0.14292 -0.02019 0.00338 0.00006 -0.00442***
Eggs and Milk 0.07959*** -0.02084* 0.02881 -0.00797 -0.00904*** 0.00176*** -0.00036
Legume 0.09160*** -0.05213*** 0.07103 -0.01890 0.00027 -0.00173** -0.00145***
Fruit & Vegetable -0.00591 -0.05147*** 0.05091 0.00056 -0.00210 -0.00610*** 0.00299***
Edible Oil 0.06991*** 0.01874 -0.02778 0.00904 0.01701*** 0.00053 -0.00017
Tobacco 0.06124*** -0.04802** 0.00327 0.04475** -0.01610*** -0.00152** 0.00225***
Prepared Foods 0.06600*** 0.11356*** -0.05987 -0.05369** -0.03801*** 0.00114 -0.00546***
Spices c -0.01942 0.02756 -0.06746 0.03990 -0.00041 -0.00001 0.00248

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice 0.00338 -0.00904*** 0.00026*** -0.00210 0.01701*** -0.01610***
Non Rice Staple 0.00006*** 0.00176*** -0.00173** -0.00610*** 0.00053 -0.00152*
Fish -0.00442*** -0.00036 -0.00145*** 0.00299*** -0.00017 0.00225***
Meat 0.00460** 0.00143*** 0.00547*** 0.00024 -0.00526*** 0.00169
Eggs and Milk 0.00143*** 0.00249*** 0.00292*** -0.00043 0.00131*** 0.00038
Legume 0.00547*** 0.00292*** 0.00475*** -0.00147 -0.00138* -0.00481***
Fruit & Vegetable 0.0002 -0.00043 -0.00147 0.01073*** -0.00332*** 0.00125
Edible Oil -0.00526*** 0.00131*** -0.00138* -0.00332*** -0.00795*** 0.00206***
Tobacco 0.00169 0.00038 -0.00481*** 0.00125 0.00206*** 0.01893***
Prepared Food -0.00536*** -0.00056 -0.00404*** -0.00223*** -0.00261*** -0.00276***
Spices c -0.00183 0.00011 0.00146 0.00045 -0.00021 -0.00136

Commodity Prices Total Dummy Effect on Budget Share

House-hold
MODEL Prepared Budget
Spices Lower Middle Higher Size
Foods Shares
Rice -0.03801*** -0.00041 -0.22686*** 0.05454*** -0.15401 0.09948*** 0.10296***
Non Rice Staple 0.00114 -0.00001 0.12192* -0.07672*** 0.16518 -0.08846*** 0.03111***
Fish -0.00546*** 0.00248*** 0.03234*** -0.01303*** 0.03217 -0.01914*** -0.00949***
Meat -0.00536*** -0.00183*** -0.04674*** 0.03439*** -0.04449 0.01011** 0.00368***
Eggs and Milk -0.00056 0.00011 -0.00999*** 0.00529 -0.00923 0.00394 -0.00493***
Legumes -0.00404*** 0.00146*** -0.01373*** 0.01201** -0.01953 0.00752 -0.01003***
Fruit & Vegetable -0.00223*** 0.00045 0.01898*** 0.01532*** -0.01705 0.00173 -0.02035***
Edible Oil -0.00261*** -0.00021 -0.00133 -0.00535* 0.00840 -0.00305 -0.00503***
Tobacco -0.00276*** -0.00136*** 0.01722*** 0.01240** -0.00320 -0.00920 -0.03090***
Prepared Foods 0.06120*** -0.00133*** 0.08533*** -0.03201* 0.02185 0.01016* -0.04397***
Spices c -0.00132 0.00067 0.02286 -0.00684 0.01992 -0.01308 -0.01305
Note: *: p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

85
 Table 6.4 Parameter Estimates for the LA/AIDS Model
Based on the SUSENAS Micro Data: 1993, Urban East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.37317*** 0.08145*** -0.04648 -0.03496 0.10559*** -0.00395*** -0.01430***
Non Rice Staple 0.01471*** 0.00132 -0.01228 0.01096 -0.00395*** 0.00335*** -0.00039
Fish 0.07548*** -0.01378 -0.01647 0.03025** -0.01430*** -0.00039 0.01919***
Meat 0.16219*** -0.04560*** 0.04161 0.00399 -0.00423 0.00210*** 0.00250***
Eggs and Milk 0.05588*** -0.01792 0.05660 -0.03868* -0.00999*** -0.00010 0.00027
Legume 0.10210*** -0.01140 0.03598 -0.02458* 0.00079 0.00006 -0.00236***
Fruit & Vegetable 0.02516*** -0.03800*** 0.02264 0.01536 -0.01400*** -0.00016 0.00408***
Edible Oil 0.08358*** 0.01145 -0.01144 -0.00001 0.00297 0.00064 -0.00282***
Tobacco 0.04514*** -0.04929** -0.04225 0.09154*** -0.02425*** -0.00126 0.00041
Prepared Foods 0.02703*** 0.05927** 0.01051 -0.06978** -0.03772*** -0.00090** -0.00666***
Spices c 0.03556 0.02251 -0.03841 0.01590 -0.00091 0.00060 0.00008

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice -0.00423 -0.00999*** 0.00079 -0.01400*** 0.00297 -0.02425***
Non Rice Staple 0.00210*** -0.00010 0.00006 -0.00016 0.00064 -0.00126
Fish 0.00250*** 0.00027 -0.00236*** 0.00408*** -0.00282*** 0.00041
Meat 0.01848*** 0.00145 -0.00114 -0.00138 -0.00312*** -0.00677***
Eggs and Milk 0.00145 0.00455*** 0.00280*** 0.00026 0.00170*** 0.00251**
Legume -0.00114 0.00280*** 0.00752*** -0.00219* 0.00376*** -0.00741***
Fruit & Vegetable -0.00138 0.00026 -0.00219** 0.02003*** -0.00182** -0.00220
Edible Oil -0.00312*** 0.00170*** 0.00376*** -0.00182** 0.00323*** -0.00117
Tobacco -0.00677*** 0.00251** -0.00741*** -0.00220 -0.00117 0.03983***
Prepared Food -0.01020*** -0.00297*** -0.00413*** -0.00061 -0.00454*** 0.00056
Spices c 0.00232 -0.00048 0.00230 -0.00201 0.00118 -0.00026

Commodity Prices Total Dummy Effect on Budget Share

House-hold
MODEL Prepared Budget
Spices Lower Middleb Higher Size
Foods Shares
Rice -0.03772*** -0.00092 -0.10207*** -0.01599** -0.00030 0.01629 0.13599***
Non Rice Staple -0.00090** 0.00060* 0.00136 0.00102 -0.00353 0.00251* -0.00044
Fish -0.00666*** 0.00008 -0.01113*** 0.00311 -0.00715 0.00404* 0.00342
Meat -0.01020*** 0.00232*** -0.03627*** 0.00910* 0.00784 -0.01693* 0.00557**
Eggs and Milk -0.00297*** -0.00048 0.00287 0.00209 0.01430 -0.01639*** -0.01216***
Legumes -0.00413*** 0.00229*** -0.01464*** 0.00287 0.00476 -0.00762 0.00716***
Fruit & Vegetable -0.00061 -0.00201*** 0.01288*** 0.00686 -0.00094 -0.00592 -0.01912***
Edible Oil -0.00454*** 0.00118*** -0.01018*** -0.00303 0.00009 0.00294 -0.00014
Tobacco 0.00056 -0.00026 0.03902*** 0.01472** -0.02861 0.01389*** -0.03217***
Prepared Foods 0.07041*** -0.00323*** 0.11804*** -0.01407* 0.01850 -0.00443** -0.08230***
Spices c -0.00323 0.00043 0.00012 -0.00667 -0.00495 0.01162 -0.00581
Note: *: p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

86
 Table 6.5 Parameter Estimates for the LA/AIDS Model
Based on the 1993 - SUSENAS Micro Data: Rural East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.38858*** 0.09829*** -0.05124 -0.04705* 0.04422*** 0.00202 -0.01158***
Non Rice Staple -0.10448*** 0.06942*** 0.08470 -0.15412*** 0.00202 0.00193* -0.00172**
Fish 0.01383 0.00956 0.03856 -0.04812** -0.01158*** -0.00172*** 0.01707***
Meat 0.14945*** -0.05462*** -0.00133 0.05595 0.00707*** -0.00099* 0.00006
Eggs and Milk 0.07183*** -0.02164** -0.01563 0.03727 -0.00157 0.00068 0.00004
Legume 0.08848*** -0.01942* -0.00290 0.02232 -0.00294 0.00150** 0.00038
Fruit & Vegetable 0.02056 -0.02212* -0.00809 0.03021 -0.00487*** -0.00114* 0.00180***
Edible Oil 0.12488*** 0.00054 -0.00922 0.00868 0.02943*** 0.00085** -0.00194***
Tobacco 0.06284*** -0.05305*** 0.01943 0.03362 -0.02811*** -0.00283*** 0.00042
Prepared Foods 0.15469*** 0.00234 -0.05321 0.05087*** -0.02829*** -0.00160** -0.00409***
Spices c 0.02934 -0.00930 -0.00107 0.01037 -0.00538 0.00130 -0.00043
MODEL
COMMODITY PPRICES
Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice 0.00707*** -0.00157 -0.00294 -0.00487*** 0.02943*** -0.02811***
Non Rice Staple -0.00099* 0.00068 0.00150** -0.00114* 0.00085** -0.00283***
Fish 0.00006 0.00004 0.00038 0.00180*** -0.00194*** 0.00042
Meat 0.01557*** 0.00119** -0.00132 -0.00425*** -0.01136*** 0.00177*
Eggs and Milk 0.00120** 0.00152*** 0.00177*** 0.00072 0.00079** -0.00182**
Legume -0.00132 0.00177*** 0.00577*** -0.00197** 0.00131* -0.00402***
Fruit & Vegetable -0.00425*** 0.00072 -0.00197** 0.01158*** -0.00326*** -0.00051
Edible Oil -0.01136*** 0.00079** 0.00131* -0.00326*** -0.01226*** 0.00012
Tobacco 0.00177* -0.00182** -0.00402*** -0.00051 0.00012 0.03298***
Prepared Food -0.00947*** -0.00292*** -0.00333*** 0.00074 -0.00503*** 0.00361***
Spices c 0.00173 -0.00041 0.00284 0.00116 0.00135 -0.00160

Commodity Prices Dummy Effect on Budget Share

Total Budget House-hold
MODEL Prepared
Spices Shares Lower Middleb Higher Size
Foods
Rice -0.02829*** -0.00538*** -0.10811*** -0.02877*** 0.01252 0.01626 0.12759***
Non Rice Staple -0.00160** 0.00130*** 0.04420*** -0.01085** -0.03185 0.04271*** 0.01367***
Fish -0.00409*** -0.00043 0.01774*** -0.00181 -0.01120 0.01300** -0.01121***
Meat -0.00947*** 0.00173*** -0.03149*** 0.01257*** 0.00341 -0.01598 0.00611***
Eggs and Milk -0.00292*** -0.00041 -0.00839*** 0.00542* 0.00636 -0.01179** -0.00524***
Legumes -0.00333*** 0.00284*** -0.00496 0.00275 0.00220 -0.00495 -0.01147***
Fruit & Vegetable 0.00074 0.00116*** 0.01387*** 0.00430 0.00383 -0.00813 -0.01463***
Edible Oil -0.00503*** 0.00134*** -0.01197*** -0.00022 0.00277 -0.00255 -0.00347***
Tobacco 0.00361*** -0.00160*** 0.02766*** 0.01408*** -0.00293 -0.01115 -0.03905***
Prepared Foods 0.05183*** -0.00146*** 0.05407*** -0.00155 0.01482 -0.01327** -0.04797***
Spices c -0.00146 0.00091 0.00738 0.00408 0.00007 -0.00415 -0.01433
Note: *: p < .10; **: p < 0.05; ***:bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

87
 Table 6.6 Parameter Estimates for the LA/AIDS Model
Based on the 1996 - SUSENAS Micro Data: Urban East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.25999*** 0.14216*** -0.17063 0.02847 0.10770*** -0.00062 -0.01199***
Non Rice Staple 0.00871 0.00538 -0.00866 0.00328 -0.00062 0.00610*** -0.00152***
Fish 0.13138*** -0.04917*** 0.10218 -0.05301*** -0.01199*** -0.00152*** 0.01930***
Meat 0.19436*** -0.10959*** 0.10043 0.00916 -0.01586*** 0.00254*** 0.00040
Eggs and Milk 0.10731*** -0.03267 0.12443 -0.09176*** -0.00357 -0.00063 0.00179**
Legume 0.03742*** 0.02751* -0.03625 0.00874 -0.00164*** 0.00156* -0.00268***
Fruit & Vegetable 0.04594*** -0.08497*** 0.08216 0.00281 -0.01210*** -0.00039 0.00235***
Edible Oil 0.07617*** 0.02069** -0.02989 0.00920 0.01617*** -0.00053 -0.00270**
Tobacco 0.02618 -0.05489** 0.04031 0.01458 -0.03263*** -0.00288*** -0.00047
Prepared Foods 0.06642*** 0.13026*** -0.19534 0.06508*** -0.04090*** -0.00415*** -0.00396***
Spices c 0.04612 0.00529 -0.00875 0.00346 -0.00456 0.00054 -0.00051

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice -0.01586*** -0.00357*** -0.00164 -0.01210*** 0.01617*** -0.03263***
Non Rice Staple 0.00254*** -0.00063 0.00156** -0.00039 -0.00053 -0.00289***
Fish 0.00040 0.00178** -0.00268*** 0.00235*** -0.00270*** -0.00047
Meat 0.02352*** -0.00090 -0.00349** -0.00052 -0.00342*** 0.00243
Eggs and Milk -0.00090 -0.00259** 0.00094 -0.00006 0.00095** 0.00369***
Legume -0.00349** 0.00094 0.00915*** -0.00206** 0.00478*** 0.00434**
Fruit & Vegetable -0.00052 -0.00006 -0.00206** 0.02609*** -0.00252*** -0.00101
Edible Oil -0.00342*** 0.00095** 0.00478*** -0.00252*** -0.00746** 0.00043
Tobacco 0.00243 0.00369*** 0.00434 -0.00101 0.00043 0.04536***
Prepared Food -0.00529*** -0.00048 -0.01110*** -0.00930*** -0.00459*** -0.01797***
Spices c 0.00058 0.00088 0.00020 -0.00048 -0.00111 -0.00128

Commodity Prices Total Dummy Effect on Budget Share

House-hold
MODEL Prepared Budget
Spices Lower Middleb Higher Size
Foods Shares
Rice -0.04090*** -0.00456*** -0.07069*** -0.03064*** 0.04600 -0.01536** 0.11999***
Non Rice Staple -0.00415*** 0.00054 -0.00047 -0.00037 0.00168 -0.00131 0.00294***
Fish -0.00396*** -0.00051** -0.02381*** 0.01057** -0.02430 0.01373*** 0.00187
Meat -0.00529*** 0.00058 -0.04119*** 0.02553*** -0.03032 0.00480 0.00003
Eggs and Milk -0.00048 0.00088*** -0.00828* 0.00600 -0.03292 0.02691*** -0.01276***
Legumes -0.01109*** 0.00020 -0.00265 -0.00710* 0.01036 -0.00326 0.00439**
Fruit & Vegetable -0.00930*** -0.00048 0.00154 0.01889*** -0.02334 0.00444 -0.02293***
Edible Oil -0.00459*** -0.00111*** -0.01020*** -0.00498** 0.00728 -0.00230 0.00342***
Tobacco -0.01797*** -0.00128* 0.02258*** 0.01344** -0.00831 -0.00513 -0.02863***
Prepared Foods 0.10071*** -0.00297*** 0.13703*** -0.03014*** 0.05162 -0.02149*** -0.06753***
Spices c -0.00298*** 0.00872 -0.00386 -0.00120 0.00225 -0.00105 -0.00079
Note: *: p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

88
 Table 6.7 Parameter Estimates for the LA/AIDS Model
Based on the 1996 - SUSENAS Micro Data: Rural East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.30791*** 0.04109 0.04096 -0.08205** 0.11334*** -0.02431*** -0.01601***
Non Rice Staple 0.05591*** 0.02025 -0.01523 -0.00502 -0.02431*** 0.03077*** -0.00314***
Fish 0.05055*** -0.00721 -0.03924 0.04645*** -0.01601*** -0.00314*** 0.01732***
Meat 0.14639*** -0.07173*** 0.02880 0.04293*** -0.00614*** 0.00366*** -0.00191***
Eggs and Milk 0.08341*** -0.02115* 0.03310 -0.01195 -0.00296*** -0.00155** -0.00070*
Legume 0.04397*** 0.00147 -0.00622 0.00475 -0.00562*** 0.00335*** 0.00124*
Fruit & Vegetable 0.03687*** -0.04477*** 0.06456 -0.01979 -0.00460*** 0.00014 0.00311***
Edible Oil 0.14115*** -0.01719* 0.03168 -0.01449 0.03231*** 0.00034 0.00094***
Tobacco 0.01769 -0.03844** -0.01041 0.04885** -0.02968*** -0.00148 0.00244***
Prepared Foods 0.08151*** 0.12830*** -0.10962 -0.01868 -0.05162*** -0.00489*** -0.00408***
Spices c 0.03464 0.00938 -0.01839 0.00901 -0.00471 -0.00289 0.00080

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice -0.00614*** -0.00296*** -0.00562*** -0.00460*** 0.03231*** -0.02968***
Non Rice Staple 0.00366*** -0.00155** 0.00335*** 0.00013 0.00034 -0.00148
Fish -0.00191*** -0.00070* 0.00124** 0.00311*** 0.00094** 0.00244***
Meat 0.01903*** -0.00108*** -0.00005 -0.00233*** -0.00820*** -0.00015
Eggs and Milk -0.00108** 0.00062 0.00092** 0.00052 0.00192*** 0.00136**
Legume -0.00005 0.00092** 0.00521*** 0.00102 -0.00042 0.00022
Fruit & Vegetable -0.00233*** 0.00052 0.00102 0.00578*** -0.00364** 0.00363***
Edible Oil -0.00820*** 0.00192*** -0.00042 -0.00364*** -0.02326*** 0.00361***
Tobacco -0.00015 0.00136** 0.00022 0.00363*** 0.00361*** 0.03112***
Prepared Food -0.00231*** 0.00147*** -0.00716*** -0.00529*** -0.00329*** -0.01023***
Spices c -0.00051 -0.00053 0.00130 0.00167 -0.00032 -0.00084

COMMODITY PRICES Total Dummy Effect on Budget Share

House-hold
MODEL Prepared Budget
Spices Lower Middleb Higher Size
Foods Shares
Rice -0.05162*** -0.00472*** -0.09119*** -0.01376* -0.00440 0.01817 0.10056***
Non Rice Staple -0.00489*** -0.00288*** -0.01157*** 0.00222 0.00102 -0.00324 0.03235***
Fish -0.00408*** 0.00080*** -0.00093 0.00261 0.00836 -0.01096*** 0.00103
Meat -0.00231*** -0.00051 -0.02606*** 0.01805*** -0.01154 -0.00651* -0.00101
Eggs and Milk 0.00147*** -0.00053*** -0.00788*** 0.00480 -0.01021 0.00542* -0.00421***
Legumes -0.00716*** 0.00130*** 0.00108 -0.00117 0.00274 -0.00158 -0.00768***
Fruit & Vegetable -0.00529*** 0.00167*** 0.00199 0.00930** -0.01816 0.00887** -0.01296***
Edible Oil -0.00329*** -0.00032 -0.01893*** 0.00457* -0.00798 0.00341 -0.00081
Tobacco -0.01023*** -0.00084* 0.02819*** 0.00872* 0.00240 -0.01112** -0.03488***
Prepared Foods 0.09046 -0.00307*** 0.12417*** -0.03323*** 0.03282 0.00041 -0.06816***
Spices c -0.00307 0.00910 0.00113 -0.00210 0.00497 -0.00287 -0.00423
Note: *: p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

89
 Table 6.8 Parameter Estimates for the LA/AIDS Model
Based on the 1999 - SUSENAS Micro Data: Urban East Java, Indonesia

MODEL Intercept Income Group Based- Dummy Commodity Prices

Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.20433*** 0.04987** -0.03882 -0.01105 0.08999*** -0.00381*** -0.01542***
Non Rice Staple 0.01927** 0.02442** -0.02381 -0.00060 -0.00382*** 0.00722*** 0.00022
Fish 0.11637*** -0.01589 0.03230 -0.01641 -0.01542*** 0.00021 0.02296***
Meat 0.16798*** -0.06886*** 0.03986 0.02900** -0.00239 0.00436*** 0.00016
Eggs and Milk 0.13368*** -0.03789*** -0.00176 0.03965** -0.00763*** -0.00257*** 0.00034
Legume 0.02743** 0.01071 -0.00185 -0.00886 0.00092 -0.00129* -0.00184***
Fruit & Vegetable 0.05634*** -0.00903 -0.00163 0.01066 -0.01184*** -0.00136** -0.00035
Edible Oil 0.09478*** 0.00535 -0.00316 -0.00219 0.00529*** 0.00052 -0.00305***
Tobacco -0.18798*** 0.08406*** -0.14213 0.05807*** -0.00226 0.00295*** 0.00279***
Prepared Foods 0.31390*** -0.05104** 0.14089 -0.08986*** -0.04928*** -0.00524*** -0.00582***
Spices 0.05390 0.00830 0.00010 -0.00840 -0.00356 -0.00099 0.00002

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice -0.00239 -0.00763*** 0.00092 -0.01184*** 0.00529*** -0.00226
Non Rice Staple 0.00436*** -0.00257*** -0.00129* -0.00137** 0.00052 0.00295***
Fish 0.00016 0.00034 -0.00184*** -0.00035 -0.00305*** 0.00279***
Meat 0.00122 -0.00141** 0.00153* 0.00051 -0.00011 -0.00226***
Eggs and Milk -0.00141** 0.01032*** 0.00206*** -0.00008 -0.00169*** 0.00341***
Legume 0.00153* 0.00206*** 0.00836*** -0.00269*** 0.00756*** -0.00208**
Fruit & Vegetable 0.00051 -0.00008 -0.00269*** 0.02224*** -0.00087 0.00156**
Edible Oil -0.00011 -0.00169*** 0.00756*** -0.00087 -0.00631*** 0.00734***
Tobacco -0.00226*** 0.00341*** -0.00208** 0.00156** 0.00734*** 0.00622***
Prepared Food -0.00157* -0.00088 -0.01436*** -0.00699*** -0.00666*** -0.01785***
Spices -0.00005 -0.00187 0.00183 -0.00013 -0.00202 0.00019

Commodity Prices Total Dummy Effect on Budget Share

House-hold
MODEL Prepared Budget
Spices Lower Middleb Higher Size
Foods Shares
Rice -0.04928*** -0.00356*** -0.04953*** -0.00827 -0.00398 -0.01536 0.07742***
Non Rice Staple -0.00524*** -0.00099*** -0.00351* -0.00492** -0.00019 -0.00131 0.00641***
Fish -0.00582*** 0.00002 -0.01916*** 0.00129 0.00529 0.01373 0.00548***
Meat -0.00157* -0.00005 -0.02846*** 0.01507*** -0.00237 0.00480 0.00738***
Eggs and Milk -0.00088 -0.00187*** -0.02104*** 0.00733* -0.00471 0.02691 0.00274*
Legumes -0.01436*** 0.00183*** -0.00305 -0.00282 0.00139 -0.00326 -0.00095
Fruit & Vegetable -0.00698*** -0.00013 -0.00387* 0.00078 0.00020 0.00444 -0.00632***
Edible Oil -0.00666*** -0.00201*** -0.01097*** -0.00159 0.00046 -0.00230 0.00335***
Tobacco -0.01785*** 0.00019 0.07087*** -0.01975*** -0.01533 -0.00513 -0.02316***
Prepared Foods 0.11454*** -0.00590*** 0.07411*** 0.01525** 0.01745 -0.02149*** -0.07266***
Spices -0.00590 0.01248 -0.00540 -0.00235 0.00178 -0.00105 0.00031
Note: *: p < .10; **: p < 0.05; ***: bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

90
 Table 6.9 Parameter Estimates for the LA/AIDS Model
Based on the 1999 - SUSENAS Micro Data: Rural East Java, Indonesia

Income Group Based- Dummy Commodity Prices

MODEL Intercept
Lower Middlea Higher Rice N.Rice.S Fish
Rice 0.30874*** -0.01973 0.13735 -0.11762*** 0.07671*** -0.01706*** -0.01192***
Non Rice Staple -0.13810*** 0.07470*** -0.20310 0.12840*** -0.01706*** 0.02499*** -0.00091
Fish 0.05717*** -0.00760 0.00953 -0.00193 -0.01192*** -0.00091 0.02022***
Meat 0.15254*** -0.07059*** 0.05898 0.01161 0.00812*** -0.00116 -0.00518***
Eggs and Milk 0.09281*** -0.04293*** 0.01904 0.02389** -0.00595*** -0.00157*** -0.00065**
Legume 0.05063*** -0.01447 0.00984 0.00463 -0.00734*** 0.00265*** 0.00213***
Fruit & Vegetable 0.05233*** -0.03274*** 0.03647 -0.00373 -0.00685*** -0.00077 -0.00115***
Edible Oil .15790*** 0.00079 0.02045 -0.02124* 0.03225*** 0.00336 0.00105**
Tobacco -0.11255*** 0.07752*** -0.08576 0.00824 -0.00197** 0.00041 0.00065
Prepared Foods 0.32408*** 0.03619** -0.00828 -0.02791 -0.05937*** -0.00975*** -0.00518***
Spices 0.05445 -0.00114 0.00548 -0.00434 -0.00663 -0.00019 0.00094

MODEL COMMODITY PPRICES

Meat Egg & Milk Legumes Fruit &V. Edible Oil Tobacco
Rice 0.00812*** -0.00595*** -0.00734*** -0.00685*** 0.03225*** -0.00197**
Non Rice Staple -0.00116 -0.00157*** 0.00265*** -0.00077 0.00336*** 0.00041
Fish -0.00518*** -0.00065** 0.00213*** -0.00115*** 0.00105** 0.00065
Meat 0.00171 0.00174*** 0.00201*** 0.00172** -0.00248*** 0.00010
Eggs and Milk 0.00174*** 0.00494*** 0.00025 -0.00026 0.00109*** 0.00064*
Legume 0.00201*** 0.00025 0.00984*** -0.00165*** 0.00435*** -0.00016
Fruit & Vegetable 0.00172*** -0.00026 -0.00165*** 0.01554*** -0.00307*** 0.00211***
Edible Oil -0.00248*** 0.00109*** 0.00435*** -0.00307*** -0.03502*** 0.00052
Tobacco 0.00010 0.00064* -0.00016 0.00211*** 0.00052 -0.00243***
Prepared Food -0.00524*** 0.00028 -0.01157*** -0.00577*** -0.00280*** -0.00016
Spices -0.00134 -0.00050 -0.00051 0.00016 0.00073 0.00028

Commodity Prices Total Dummy Effect on Budget Share

Household
MODEL Prepared Budget Size
Spices Lower Middleb Higher
Foods Shares
Rice -0.05937*** -0.00663*** -0.09448*** 0.00071 -0.02806 0.02734*** 0.08320***
Non Rice Staple -0.00975*** -0.00019 0.04206*** -0.00612 0.04264 -0.03652*** 0.02113***
Fish -0.00518*** 0.00095*** -0.00238 0.00146 -0.00165 0.00020 0.00291**
Meat -0.00524*** -0.00134*** -0.02403*** 0.01526*** -0.01567 0.00041 0.00077
Eggs and Milk 0.00028 -0.00050*** -0.01363*** 0.00959*** -0.00665 -0.00294 -0.00061
Legumes -0.01157*** -0.00051 -0.00564*** 0.00232 -0.00281 0.00049 -0.00926***
Fruit & Vegetable -0.00577*** 0.00016 -0.00285* 0.00689*** -0.00992 0.00303 -0.00810***
Edible Oil -0.00280*** 0.00073* -0.02179*** -0.00195 -0.00339 0.00533** 0.00054
Tobacco -0.00016 0.00028 0.05946*** -0.02012*** 0.02306 -0.00293 -0.03281***
Prepared Foods 0.10363*** -0.00408*** 0.06585*** -0.00823** 0.00339 0.00484 -0.05361***
Spices -0.00407 0.01115 -0.00257 0.00019 -0.00093 0.00074 -0.00417
Note:
* : p < .10 ; ** : p < 0.05; *** : bp < 0.01
a, b, c: parameter estimates are recovered by applying the model’s restrictions.

91
 Table 6.10 The Results Summary:
Number of statistically significant Estimates
Summarized from the estimation results with LA/AIDS Model:
SUSENAS Micro Data East Java, Indonesia
Number of statistically significant Estimates
   Dummy  TOTAL
TABLE
Expenditu Prices Income Household Number In
Intercept
re (out of Group Size (out of Percentage
(out of 10) (%)
(out of 30) 100) (out of 20) (out of 10) 170)
6.2 : Urban90 6 13 60 8 6 93 55
6.3 : Rural 90 9 23 76 14 10 132 78
6.4 : Urban93 10 18 69 10 7 114 67
6.5 : Rural 93 8 18 84 11 10 131 77
6.6 : Urban96 8 19 76 11 8 122 72
6.7 : Rural 96 9 18 87 10 7 131 77
6.8 : Urban99 10 15 78 10 9 122 72
6.9 : Rural 99 10 17 86 10 7 130 76

The result is displayed fully, with the intention to provide some ideas about the signs,
magnitudes of parameter estimates and the fitness of the model performance.
In the tables, one observes the coefficient of each parameter and their corresponding
statistical significance as indicated by p-values. The model was estimated with imposition
of homogeneity and symmetry restrictions.
As can be followed in the tables, own price, budget shares and household size determined
significantly the variation of food consumption of the households investigated with
significant level mostly 0.01. The presence of dummy for income groups indicated also
their significant influence, with exception for some food groups, in which their presence in
the model is not insignificant. The income group dummies work mostly in both manners:
shifting the curve and changing the slope of related demand curve for each food group.
Thus, households from different income groups will behave differently in their
consumption of the same food group.
In term of sign and magnitude, all own prices displayed satisfying performance. They
affects significantly different from zero, and move in the directions that conform theory
consistently.
Concerning the cross price parameters, not all of them display a significant influence
statistically. But this is not surprising. In the situation where any food group dominates the
share of household‘s budget, it is possible that the presence of the other not significant. The

92
model represents the real food consumption of households in East Java plausibly. This
holds for both areas under studied, urban and rural.

6.3 The Results of the Restrictions Test

Using likelihood ratio statistics described in Section 5.7, the study come to the test result
shown in Table 6. 11. As can be followed in the Table 6.11, the theoretical restrictions
tested are the model with homogeneity vs. unrestricted model, model with symmetry vs.
unrestricted model, and model homogeneity and symmetry simultaneously imposed vs.
unrestricted model. Since the specification of significance level for a large number of tests
is arbitrary, the critical levels presented in Table 6.12 displays a variety of possibilities.
One can evaluate the test statistics with reference of this table. The comparison between the
calculated Chi-squared test statistics in Table 6.11 and the corresponding critical level in
Table 6.12 indicates that in all cases, the calculated statistics are far greater than the
corresponding critical levels, meaning that the data rejected the validity of restrictions
imposed. It suggests that homogeneity and symmetry conditions either alone or jointly are
strongly violated by the data. It is evident that the test suffered from over rejection.
Table 6.11 Test Results
TEST STATISTICS
TYPES OF – 2 Log  Conclusion
RESTRICTIONS D.F
TEST 1990 1993 1996 1999
Rural Urban Rural Urban Rural Urban Rural Urban Rejected
Ho: Homogeneity
10 391.79 79.35 415.39 213.26 443.49 301.88 797.98 417.73 Rejected
H1: Unrestricted
Ho: Symmetry
20 1087.7 176.58 1457.3 425.87 1630.3 608.57 2102.9 990.34 Rejected
H1: Unrestricted
Ho: Symmetry
46 1186.9 201.19 1643.1 599.21 2013.8 715.94 2570.1 1268.7 Rejected
H1: Homogeneity
Ho: Homogeneity
Symmetry 56 695.98 97.23 1041.9 212.60 1186.8 306.69 1304.9 572.60 Rejected
H1: unrestricted
*) Degree of freedom

93
 Table 6.12 Critical Values of 2

Degree of Level of Significance

Freedom 0.100 0.050 0.025 0.010
10 15.99 18.30 20.48 23.21
20 28.41 31.41 34.17 37.57
40 51.81 55.76 59.34 63.69
50 63.17 67.50 71.42 76.15
60 74.40 79.09 83.30 88.38
70 85.53 90.53 95.02 100.42
Degree .of freedom for model_1 = 10; Model 2 = 20, Model 3 = 46; Model 4 = 56.
Source: Griffiths et.al (1993)

While the interpretation for that cannot be clearly given, the rejection of demand
restrictions by the data is actually a common finding in many previous studies of this type,
such as Deaton and Muellbauer (1980b, pp.68-73); Blanciforti and Green (1983, 19986);
Mergos and Donatos, (1989), Caps (1993), Cozzarin and Gilmour (1998) or Chang (2000).
The following arguments are proposed in the literature to explain this rejection
phenomenon:
1. Rejection due to data problems (Deaton and Muellbauer (1980a)
2. Rejection due to measurement errors and the use of proxy variables
3. Rejection because of model misspecification including the ignorance of dynamic
effects, variable omission, lagged dependent variables and price expectations (Chang,
2000).
4. Rejection due to the rigidity of the AIDS model itself (Syriopoulos and Sinclair (1993),
and the use of Stone index and associated problems
It does not necessarily mean however, that the theory is wrong; it may be rather the case,
that the data and model combined do not support the theory either because of data property,
and/or model specification (Chang, 2000).
Because of that reasons, in the estimation we imposed both of restrictions to assure that the
results are achieved in assurance of theoretical foundation.

94
6.4 Price Elasticities
Our concern is the demand elasticity of own- and cross prices for both neo-classical types:
Marshallian and Hicksian elasticities. Marshallian or ordinary price elasticity is defined
from the Marshallian demand function (Chapter 4): that is, a demand function obtained
from utility maximization subject to budget constraint. Hicksian or compensated price
elasticity is elasticity obtained through solving the dual problem of expenditure
minimization at a certain utility level. The compensated price elasticity measures a
response of consumer on the price change, given that their income be compensated, thus at
a constant purchasing power. The ordinary price elasticity indicates an overall response of
consumer on changing prices, -of own or other goods, without compensation on their
income.
Knowing both of them is important, as they reveal exhaustive information to describe
household‘s reaction on price changes. This delivers an important advantage on the use of
the study in policy making. The question of whether price or income should be used as a
policy instrument may be directed from the elasticities of both types.
In analyzing the elasticities, we pay always ourr attention on that food group with a relative
high budget share, thus rice and prepared foods. Rice may be, as indicated by the share
assessment above, our focus because of its ―strategic‖ status in as a political issue.
Table 6.13 Ordinary Own Price Elasticities
Based on the SUSENAS Micro Data 1990, 1993, 1996 and 1999:
Urban East Java, Indonesia
URBAN AREAS
Food Composites
1990 1993 1996 1999
Rice -0.4750 -0.3673 -0.3903 -0.5021
Non Rice Staple -0.7147 -0.8395 -0.7117 -0.6713
Fish -0.6910 -0.6660 -0.7762 -0.6031
Meat -0.8337 -0.7753 -0.7157 -0.9659
Eggs and Milk -1.1399 -0.9477 -1.0673 -0.7907
Legumes -0.7900 -0.8697 -0.8121 -0.8519
Fruit & Vegetable -0.6296 -0.5802 -0.4440 -0.4569
Edible Oil -1.1145 -0.9167 -1.1462 -1.1199
Tobacco -0.6597 -0.6780 -0.6393 -0.9802
Prepared Foods -0.7479 -0.8495 -0.7136 -0.7337
Spices -0.9955 -1.0046 -0.9852 -0.9787

95
 Table 6.14 Ordinary Own Price Elasticities
Based on the SUSENAS Micro Data 1990, 1993, 1996 and 1999:
Rural East Java, Indonesia
RURAL AREAS
Food Composites
1990 1993 1996 1999
Rice -0.6892 -0.7033 -0.4863 -0.6209
Non Rice Staple -0.7254 -0.9701 -0.4760 -0.6427
Fish -0.6928 -0.7325 -0.6759 -0.6179
Meat -0.9442 -0.8038 -0.7615 -0.9691
Eggs and Milk -0.9444 -0.9699 -0.9891 -0.8798
Legumes -0.9193 -0.8972 -0.8995 -0.8169
Fruit & Vegetable -0.7357 -0.6582 -0.8426 -0.6047
Edible Oil -1.1413 -1.2062 -1.4102 -1.5832
Tobacco -0.7906 -0.7043 -0.7041 -1.0636
Prepared Foods -0.7091 -0.8134 -0.6856 -0.6381
Spices -1.0023 -1.0028 -0.9871 -0.9872

6.4.1 Own Price Elasticities

The ordinary own price elasticities in all survey periods, are displayed in Table 6.13 (urban
sector) and Table 6.14 (rural sector), while the compensated own price elasticities are in
Table 6.15 (urban sector) and 6.16 (rural sector).
As can be observed in the tables, all own price elasticities of food composites in all survey
periods reveal conformity with theoretical postulate. They are all negative in sign. In term
of magnitude, most of food groups analyzed in this study are price inelastic, with
exceptions to Eggs and Milks, edible oil and spices. ―Egg and Milks‖ (urban sector in 1990
and 1996), edible oils (urban sector, in 1990, 1996 and 1999; rural sector in all survey
periods), and spice (urban: 1993; rural: 1990, 1993) they have own price elasticities
exceeding unity. Thus, food groups under investigation were, in general less responsive to
changes in price of food goods. This holds for all food groups in all period of surveys and
across areas under study: the urban and rural.
It is notable, that ordinary own price of rice group is less elastic compared to the other food
groups. This was consistent with general intuitive proposition for staple food with no close
substitute, which accounts for a dominant share of the consumer‘s budget, as was the case
of rice in this study. General phenomenon indicates that food good with such characteristics
has a low own price elasticity.

96
It is also of our interest to identify the effects of changing economy from 1996 to 1999 due
to the crisis. From own price elasticities we obtained, that here is no clear pattern of
changing in ordinary own price elasticities. Some food groups, -rice, fish, meat, fruits &
vegetables, tobacco, revealed an increase in magnitude, and some other a decline. So it is
fair to conclude, that, the changing economic condition affected each of the individual food
group differently.
Shown in Table 6.15 and 6.16, compensated elasticities indicate, that keeping purchasing
power unchanged makes demand for food goods more price-inelastic.
It may also be concluded, that own price elasticities in rural areas are in general higher than
that of urban sector.
Table 6.15 Compensated Own Price Elasticities
Based on SUSENAS Micro Data 1990, 1993, 1996 and 1999:
Urban East Java, Indonesia
URBAN AREAS
MODEL
1990 1993 1996 1999
Rice -0.3641 -0.2811 -0.2767 -0.3578
Non Rice Staple -0.6772 -0.8197 -0.6914 -0.6576
Fish -0.6400 -0.6213 -0.6898 -0.5561
Meat -0.7483 -0.7049 -0.6381 -0.9155
Eggs and Milk -1.0414 -0.8644 -0.9891 -0.7553
Legumes -0.6938 -0.8159 -0.7687 -0.7982
Fruit & Vegetable -0.5565 -0.5152 -0.3831 -0.4187
Edible Oil -1.0600 -0.8842 -1.1120 -1.0844
Tobacco -0.5522 -0.5379 -0.4980 -0.8323
Prepared Foods -0.4948 -0.4685 -0.3645 -0.3197
Spices -0.9563 -0.96957 -0.9713 -0.9675

97
 Table 6.16 Compensated Own Price Elasticities
Based on the SUSENAS Micro Data 1990, 1993, 1996 and 1999:
Rural East Java, Indonesia
RURAL AREAS
MODEL
1990 1993 1996 1999
Rice -0.4744 -0.5719 -0.3078 -0.4422
Non Rice Staple -0.6828 -0.9071 -0.4285 -0.5732
Fish -0.6359 -0.6649 -0.6304 -0.5656
Meat -0.8606 -0.7334 -0.6912 -0.9017
Eggs and Milk -0.8999 -0.9210 -0.9400 -0.8433
Legumes -0.8584 -0.8411 -0.8485 -0.7652
Fruit & Vegetable -0.6623 -0.6032 -0.7894 -0.5589
Edible Oil -1.0986 -1.1590 -1.3661 -1.5436
Tobacco -0.6875 -0.5663 -0.5814 -0.9376
Prepared Foods -0.4724 -0.5414 -0.3710 -0.3312
Spices -0.9593 -0.9531 -0.9609 -0.9589

6.4.2 Cross Price Elasticities

Estimation results of cross price elasticity, ordinary and compensated, for each of survey
period in urban and rural sectors are presented in Table 6.17 to 24 and Table 6.25 to 32
respectively.
The following pattern may be followed from the tables of cross price elasticities:
1. Change in prices of rice and prepared food groups have relatively considerable response
on the majority of food groups. Inverse of the relationship was not the case.
2. Without income compensation, the relationship between rice price and other food
groups is in general complementary. This holds also for the prepared food.
3. The relationship among majority of food groups may be complementary or competitive.
4. With income compensation when prices change, the cross response among food groups
may be changing: from complementary to competitive, and vice versa. This indicates
that income effect of price changes play an important role.
5. Seen from the magnitude of response, cross relationships among food groups are of less
importance.
6. The fact that change of rice price was considerably responsive to other food groups, and
not was the inverse, confirmed the prevalent phenomenon, that rice is the center of food
menu of Indonesia. Overall, the own price elasticities are larger than that of cross price
elasticities.

98
 Table 6.17 Ordinary Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1990 SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.00915 -0.01911 0.02598 0.04466 0.01784 -0.03615 0.09347 -0.06125 0.01133 0.0043
Non Rice Staple -0.28187 -0.06637 0.01224 0.11271 0.00621 -0.09219 0.01513 -0.08726 -0.04793 0.0035
Fish -0.24342 -0.03279 0.02851 0.0085 -0.03567 0.06807 -0.02024 -0.02059 -0.03991 0.0454
Meat -0.08535 0.00865 0.01382 -0.03676 0.05819 0.04693 -0.06067 0.06374 -0.17114 -0.0259
Eggs and Milk -0.16639 0.04682 -0.04163 -0.12808 -0.04814 -0.1148 -0.02335 -0.01932 -0.21237 -0.0078
Legumes -0.22143 -0.0079 -0.06012 0.03709 -0.01933 -0.00288 -0.01336 -0.24621 -0.15488 0.0005
Fruit & Vegetable -0.57916 -0.08632 0.04588 0.03675 -0.1285 -0.01616 -0.16091 0.01702 -0.17935 0.0181
Edible Oil 0.34157 0.01014 -0.03543 -0.11814 0.01097 0.00288 -0.12598 -0.00999 -0.09269 -0.0197
Tobacco -0.3413 -0.02439 -0.01688 0.04999 0.03284 -0.12604 0.03414 0.00058 -0.01413 0.0232
Prepared Foods -0.21388 -0.01056 -0.02693 -0.08698 -0.02148 -0.03271 -0.01918 -0.02505 -0.02657 -0.0193
Spices 0.0007 0.0009 0.0033 -0.0009 0.0027 0.0023 0.0026 0.0003 0.0045 0.0018

Table 6.18 Ordinary Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1990 SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.0205 -0.0250 0.0335 -0.0199 0.0148 0.0017 0.0724 -0.0351 -0.0883 0.0085
Non Rice Staple 0.0030 0.0056 0.0436 0.0429 0.0075 -0.0526 0.0325 0.0236 0.0995 0.0194
Fish -0.2007 -0.0350 -0.0782 -0.0065 -0.0257 0.0526 -0.0032 0.0394 -0.0968 0.0437
Meat 0.0468 0.0030 -0.0499 0.0179 0.0651 0.0037 -0.0598 0.0218 -0.0579 -0.0203
Eggs and Milk -0.1948 0.0402 -0.0070 0.0331 0.0654 -0.0090 0.0299 0.0097 -0.0094 0.0030
Legumes -0.0256 -0.0406 -0.0323 0.0905 0.0484 -0.0305 -0.0307 -0.0963 -0.0917 0.0224
Fruit & Vegetable -0.3493 -0.2580 0.0265 -0.0799 -0.0578 -0.0964 -0.1461 -0.0481 -0.2366 -0.0268
Edible Oil 0.3761 0.0261 0.0072 -0.0840 0.0333 -0.0161 -0.0565 0.0543 -0.0177 0.0032
Tobacco -0.2669 -0.0400 0.0132 -0.0008 -0.0067 -0.0720 0.0063 0.0119 -0.0763 -0.0262
Prepared Foods -0.3260 -0.0252 -0.0524 -0.0627 -0.0199 -0.0436 -0.0261 -0.0345 -0.0462 -0.0220
Spices -0.0217 -0.0066 -0.0019 -0.0084 -0.0028 -0.0022 -0.0025 -0.0039 -0.0072 -0.0138 -1.0023

99
 Table 6.19 Ordinary Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1993 SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.0071 -0.0350 0.0313 -0.0120 0.0390 -0.0419 0.0410 -0.0520 -0.0351 0.0156
Non Rice Staple -0.1766 -0.0151 0.1048 -0.0012 0.0063 -0.0052 0.0328 -0.0535 -0.0286 0.0303
Fish -0.1858 -0.0011 0.0642 0.0206 -0.0238 0.0793 -0.0353 0.0357 -0.0457 0.0101
Meat -0.0029 0.0278 0.0407 0.0297 0.0004 -0.0056 -0.0248 -0.0506 -0.0580 0.0332
Eggs and Milk -0.2193 -0.0079 -0.0139 -0.0045 0.0256 -0.0098 0.0129 0.0047 -0.1249 -0.0181
Legumes 0.0366 0.0034 -0.0317 -0.0084 0.0533 -0.0307 0.0670 -0.1084 -0.0380 0.0417
Fruit & Vegetable -0.3910 -0.0122 0.0645 -0.0673 -0.0206 -0.0730 -0.0588 -0.0953 -0.1194 -0.0582
Edible Oil 0.1233 0.0199 -0.0443 -0.0421 0.0552 0.0992 -0.0266 0.0073 -0.0253 0.0356
Tobacco -0.2563 -0.0155 -0.0096 -0.0787 0.0079 -0.0779 -0.0291 -0.0203 -0.0516 -0.0098
Prepared Foods -0.2438 -0.0135 -0.0544 -0.0823 -0.0417 -0.0450 -0.0240 -0.0394 -0.0523 -0.0291
Spices -0.0364 -0.0028 -0.0101 -0.0129 -0.0123 -0.0062 -0.0088 -0.0073 -0.0138 -0.0472 -1.0046

Table 6.20 Ordinary Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1993 SUSENAS - Micro Data: Rural -East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0377 -0.0147 0.0693 0.0159 0.0157 -0.0036 0.1423 -0.0618 -0.0113 -0.0015
Non Rice Staple 0.0266 -0.0295 -0.0182 0.0099 0.0230 -0.0192 0.0125 -0.0484 -0.0310 0.0201
Fish -0.2040 -0.0321 -0.0056 -0.0029 0.0018 0.0263 -0.0352 -0.0008 -0.0806 -0.0099
Meat 0.1282 -0.0004 0.0119 0.0220 -0.0052 -0.0436 -0.1216 0.0382 -0.0731 0.0273
Eggs and Milk -0.0538 0.0103 -0.0039 0.0199 0.0349 0.0134 0.0132 -0.0475 -0.0797 -0.0120
Legumes -0.0523 0.0268 0.0068 -0.0235 0.0316 -0.0352 0.0234 -0.0716 -0.0593 0.0507
Fruit & Vegetable -0.3303 -0.0782 0.0132 -0.1931 -0.0088 -0.1007 -0.1416 -0.0842 -0.1178 0.0087
Edible Oil 0.5600 0.0252 -0.0237 -0.1855 0.0215 0.0324 -0.0520 0.0184 -0.0546 0.0303
Tobacco -0.3861 -0.0529 -0.0205 -0.0160 -0.0363 -0.0628 -0.0177 -0.0212 -0.0437 -0.0321
Prepared Foods -0.2232 -0.0281 -0.0407 -0.0748 -0.0293 -0.0348 -0.0069 -0.0434 -0.0148 -0.0205
Spices -0.0295 -0.0043 -0.0060 -0.0057 -0.0050 -0.0027 -0.0016 -0.0042 -0.0118 -0.0200 -1.0028

100
 6.21 Ordinary Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1996 - SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0085 -0.0090 -0.0266 0.0161 0.0205 -0.0314 0.0984 -0.0875 -0.0425 -0.0077
Non Rice Staple -0.0073 -0.0626 0.1278 -0.0235 0.0784 -0.0137 -0.0205 -0.1237 -0.1676 0.0277
Fish -0.1407 -0.0178 0.0042 0.0204 -0.0315 0.0270 -0.0316 -0.0062 -0.0476 -0.0060
Meat -0.1567 0.0322 0.0155 -0.0028 -0.0339 -0.0003 -0.0340 0.0429 -0.0273 0.0099
Eggs and Milk -0.1475 -0.0191 -0.0055 -0.0497 -0.0058 -0.0193 -0.0028 0.0137 -0.1179 0.0044
Legumes 0.0231 0.0351 -0.0299 -0.0453 0.0330 -0.0282 0.1027 0.1118 -0.1447 0.0100
Fruit & Vegetable -0.3905 -0.0204 0.0046 -0.0591 -0.0350 -0.0748 -0.0814 -0.0865 -0.3571 -0.0243
Edible Oil 0.4386 -0.0036 -0.0262 -0.0420 0.0442 0.1249 -0.0381 0.0540 0.0031 -0.0149
Tobacco -0.3400 -0.0306 -0.0270 -0.0020 0.0156 0.0235 -0.0207 -0.0084 -0.2274 -0.0177
Prepared Foods -0.2199 -0.0221 -0.0417 -0.0467 -0.0209 -0.0575 -0.0485 -0.0313 -0.1029 -0.0188
Spices 0.0627 0.0067 0.0178 0.0250 0.0192 0.0185 0.0117 0.0142 0.0354 0.0795

Table 6.22 Ordinary Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1996 SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.0719 -0.0423 0.0047 0.0041 -0.0039 -0.00669 0.13977 -0.08015 -0.12156 -0.00881
Non Rice Staple -0.3485 -0.0408 0.0786 -0.0159 0.0667 0.00905 0.01718 -0.00439 -0.03469 -0.04218
Fish -0.2462 -0.0470 -0.0205 -0.0049 0.0315 0.06198 0.0264 0.06066 -0.03648 0.01924
Meat -0.0269 0.0534 -0.0131 -0.0047 0.0084 -0.0219 -0.08713 0.01476 0.0109 -0.00132
Eggs and Milk -0.0766 -0.0361 -0.0178 -0.0273 0.0171 0.00953 0.0384 0.02431 0.02041 -0.01274
Legumes -0.0983 0.0655 0.0253 0.0016 0.0190 0.02031 -0.00629 0.00717 -0.12884 0.02549
Fruit & Vegetable -0.3056 -0.0331 0.0613 -0.1242 -0.0129 -0.0014 -0.1459 0.05169 -0.30025 0.03424
Edible Oil 0.6377 0.0181 0.0279 -0.1317 0.0441 0.0028 -0.05946 0.08456 -0.01531 -0.00031
Tobacco -0.3776 -0.0314 0.0105 -0.0242 0.0016 -0.0118 0.02888 0.0226 -0.16523 -0.01601
Prepared Foods -0.3404 -0.0466 -0.0408 -0.0452 -0.0125 -0.0537 -0.03708 -0.03737 -0.08555 -0.02502
Spices 0.0299 0.0048 0.0085 0.0097 0.0046 0.0077 0.0055 0.0074 0.0120 0.0251

101
 Table 6.23 Ordinary Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1999 SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.0117 -0.0569 0.0083 -0.0208 0.0219 -0.0451 0.0399 0.0226 -0.1469 -0.0098
Non Rice Staple -0.0903 0.0324 0.2188 -0.0935 -0.0349 -0.0448 0.0415 0.1736 -0.1152 -0.0342
Fish -0.2147 0.0083 0.0167 0.0171 -0.0186 0.0028 -0.0410 0.0702 -0.0328 0.0056
Meat 0.0131 0.0713 0.0167 -0.0084 0.0369 0.0175 0.0097 -0.0075 0.0494 0.0051
Eggs and Milk -0.0707 -0.0399 0.0267 -0.0035 0.0579 0.0125 -0.0150 0.1012 0.0884 -0.0261
Legumes 0.0317 -0.0204 -0.0271 0.0314 0.0395 -0.0430 0.1336 -0.0271 -0.2231 0.0333
Fruit & Vegetable -0.2732 -0.0316 -0.0043 0.0171 0.0018 -0.0612 -0.0178 0.0458 -0.1482 -0.0014
Edible Oil 0.1629 0.0167 -0.0487 0.0145 -0.0217 0.1729 -0.0078 0.1819 -0.0616 -0.0358
Tobacco -0.0861 0.0191 0.0058 -0.0414 0.0132 -0.0371 0.0008 0.0502 -0.2575 -0.0063
Prepared Foods -0.2324 -0.0249 -0.0398 -0.0282 -0.0217 -0.0670 -0.0371 -0.0383 -0.0971 -0.0279
Spices 0.0888 0.0078 0.0264 0.0308 0.0201 0.0282 0.0175 0.0200 0.0486 0.1305

Table 6.24 Ordinary Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1999 SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.0506 -0.0342 0.0536 -0.0124 -0.0151 -0.0175 0.1472 0.0162 -0.1759 -0.0197
Non Rice Staple -0.2415 -0.0126 -0.0159 -0.0221 0.0383 -0.0107 0.0485 0.0066 -0.1372 -0.0025
Fish -0.2214 -0.0163 -0.0967 -0.0117 0.0408 -0.0212 0.0206 0.0135 -0.0943 0.0182
Meat 0.1340 -0.0076 -0.0625 0.0277 0.0324 0.0269 -0.0262 0.0112 -0.0421 -0.0147
Eggs and Milk -0.0970 -0.0248 -0.0065 0.0520 0.0145 0.0001 0.0342 0.0289 0.0456 -0.0072
Legumes -0.1217 0.0522 0.0417 0.0407 0.0069 -0.0281 0.0827 0.0017 -0.1992 -0.0080
Fruit & Vegetable -0.2224 -0.0329 -0.0396 0.0306 -0.0148 -0.0526 -0.0901 0.0382 -0.1942 -0.0009
Edible Oil 0.6317 0.0800 0.0348 -0.0185 0.0324 0.0920 -0.0405 0.0372 0.0292 0.0211
Tobacco -0.1220 -0.0239 -0.0143 -0.0298 -0.0106 -0.0240 0.0079 -0.0179 -0.1014 -0.0079
Prepared Foods -0.3059 -0.0579 -0.0348 -0.0409 -0.0100 -0.0614 -0.0335 -0.0263 -0.0235 -0.0237
Spices 0.0070 0.0036 0.0037 0.0030 0.0017 0.0022 0.0023 0.0040 0.0052 0.0091

102
 Table 6.25 Compensated Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1990 - SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0040 0.0030 0.0597 0.0658 -0.01864 0.11249 -0.01967 0.09412 0.0187 0.0187
Non Rice Staple 0.0340 -0.0039 0.1075 0.1726 -0.04263 0.06895 0.03039 0.18635 0.0443 0.0443
Fish 0.0150 -0.0023 0.1065 0.0575 0.10862 0.02378 0.07567 0.15178 0.0787 0.0787
Meat 0.1978 0.0420 0.0698 0.0169 0.09135 -0.01244 0.16919 0.03885 0.0106 0.0106
Eggs and Milk 0.3473 0.1074 0.0600 0.0269 -0.0342 0.06416 0.17203 0.16867 0.0585 0.0585
Legumes 0.1880 0.0404 0.0209 0.1606 0.0582 0.0564 -0.09369 0.14882 0.0534 0.0534
Fruit & Vegetable -0.1188 -0.0320 0.1370 0.1756 -0.0413 0.4426 0.18849 0.1621 0.0776 0.0776
Edible Oil 0.6603 0.0477 0.0276 -0.0220 0.0714 -0.07597 -0.06017 0.14371 0.0215 0.0215
Tobacco -0.0528 0.0096 0.0402 0.1370 0.0875 0.0794 0.04973 0.44779 0.0604 0.0604
Prepared Foods 0.1269 0.0296 0.0405 0.0158 0.0431 0.03429 0.03301 0.10036 0.50488 0.0248
Spices 0.2753 0.0303 0.0524 0.0579 0.0812 0.0611 0.0418 0.0493 0.1026 0.2077

Table 6.26 Compensated Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1990 - SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0443 0.0174 0.0977 0.0139 0.0558 0.0284 0.1117 0.0265 0.0407 0.0376
Non Rice Staple 0.1471 0.0340 0.0866 0.0656 0.0350 -0.0347 0.0588 0.0648 0.1859 0.0389
Fish 0.0882 0.0519 0.0081 0.0390 0.0294 0.0885 0.0496 0.1220 0.0765 0.0828
Meat 0.3273 0.0873 0.0053 0.0620 0.1185 0.0386 -0.0085 0.1020 0.1103 0.0177
Eggs and Milk 0.0885 0.1254 0.0487 0.1176 0.1194 0.0262 0.0817 0.0907 0.1604 0.0414
Legumes 0.2928 0.0552 0.0304 0.1855 0.0985 0.0090 0.0275 -0.0052 0.0992 0.0656
Fruit & Vegetable 0.2291 -0.0841 0.1403 0.0927 0.0332 0.0139 -0.0403 0.1173 0.1102 0.0516
Edible Oil 0.6109 0.0967 0.0534 -0.0139 0.0702 0.0287 -0.0274 0.1215 0.1231 0.0351
Tobacco 0.0926 0.0681 0.0840 0.1065 0.0499 -0.0035 0.0509 0.0777 0.1392 0.0225
Prepared Foods 0.0679 0.0932 0.0251 0.0549 0.0421 0.0316 0.0228 0.0376 0.0664 0.0314
Spices 0.2896 0.0900 0.0625 0.0882 0.0401 0.0515 0.0405 0.0498 0.0786 0.2116

103
 Table 6.27 Compensated Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1993 SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0018 -0.0095 0.0694 0.0149 0.0648 -0.0225 0.0604 -0.0034 0.0741 0.0303
Non Rice Staple 0.0175 0.0418 0.1899 0.0589 0.0639 0.0380 0.0762 0.0550 0.2152 0.0632
Fish -0.0323 0.0146 0.1315 0.0681 0.0218 0.1134 -0.0010 0.1216 0.1472 0.0361
Meat 0.1583 0.0443 0.0879 0.0796 0.0483 0.0303 0.0113 0.0395 0.1445 0.0606
Eggs and Milk 0.0482 0.0195 0.0645 0.1127 0.1050 0.0497 0.0727 0.1542 0.2111 0.0272
Legumes 0.2183 0.0220 0.0215 0.0713 0.1096 0.0098 0.1076 -0.0069 0.1902 0.0724
Fruit & Vegetable -0.1013 0.0175 0.1495 0.0597 0.0692 0.0130 0.0060 0.0667 0.2446 -0.0091
Edible Oil 0.2699 0.0349 -0.0013 0.0222 0.1006 0.1428 0.0060 0.0893 0.1590 0.0605
Tobacco -0.0061 0.0101 0.0638 0.0310 0.0854 -0.0036 0.0265 0.0357 0.2628 0.0326
Prepared Foods 0.0590 0.0175 0.0343 0.0504 0.0521 0.0449 0.0433 0.0283 0.1169 0.0223
Spices 0.2091 0.0206 0.0601 0.0923 0.0695 0.0640 0.0380 0.0512 0.1148 0.2568

Table 6.28 Compensated Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1993 - SUSENAS Micro Data: Rural- East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0694 0.0176 0.1137 0.0392 0.0446 0.01284 0.17158 -0.01098 0.09384 0.01929
Non Rice Staple 0.2893 0.0348 0.0701 0.0563 0.0805 0.01338 0.07074 0.05257 0.17811 0.06136
Fish 0.0719 0.0341 0.0871 0.0459 0.0621 0.06052 0.02595 0.10524 0.13889 0.03337
Meat 0.3385 0.0500 0.0634 0.0592 0.0408 -0.01748 -0.07493 0.11906 0.09426 0.06032
Eggs and Milk 0.2220 0.0764 0.0635 0.1126 0.0952 0.04766 0.07437 0.05847 0.1397 0.0313
Legumes 0.2042 0.0883 0.0694 0.0627 0.0769 -0.00333 0.08027 0.02693 0.14482 0.09094
Fruit & Vegetable 0.1035 0.0259 0.1192 -0.0473 0.0678 -0.0059 -0.04547 0.08252 0.2274 0.07675
Edible Oil 0.7740 0.0765 0.0286 -0.1136 0.0593 0.0792 -0.02545 0.10062 0.11565 0.06392
Tobacco -0.0286 0.0328 0.0669 0.1041 0.0269 0.0153 0.02664 0.05804 0.24073 0.02404
Prepared Foods 0.1179 0.0537 0.0427 0.0398 0.0310 0.0398 0.03546 0.03222 0.11627 0.0331
Spices 0.2939 0.0704 0.0686 0.0938 0.0571 0.0719 0.0357 0.0704 0.1126 0.2287

104
 Table 6.29 Compensated Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1996 - SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0185 0.0311 0.0134 0.0444 0.0448 -0.0105 0.1196 -0.0335 0.0837 0.0038
Non Rice Staple 0.1892 0.0148 0.2052 0.0311 0.1254 0.0267 0.0205 -0.0195 0.0760 0.0499
Fish 0.0791 0.0037 0.0907 0.0815 0.0211 0.0722 0.0142 0.1103 0.2249 0.0188
Meat 0.0341 0.0508 0.0907 0.0503 0.0118 0.0389 0.0058 0.1441 0.2093 0.0315
Eggs and Milk 0.1596 0.0109 0.1154 0.0712 0.0677 0.0439 0.0613 0.1766 0.2629 0.0391
Legumes 0.1872 0.0511 0.0347 0.0194 0.0786 0.0055 0.1369 0.1988 0.0587 0.0286
Fruit & Vegetable -0.0510 0.0127 0.1383 0.0745 0.0593 0.0064 -0.0105 0.0935 0.0638 0.0141
Edible Oil 0.5733 0.0096 0.0268 0.0110 0.0816 0.1571 -0.0104 0.1254 0.1701 0.0004
Tobacco -0.0632 -0.0036 0.0819 0.1070 0.0925 0.0898 0.0362 0.0493 0.1158 0.0137
Prepared Foods 0.0675 0.0060 0.0714 0.0665 0.0589 0.0113 0.0106 0.0286 0.0495 0.0137
Spices 0.2154 0.0205 0.0595 0.0806 0.0609 0.0602 0.0395 0.0489 0.1187 0.2670

Table 6.30 Compensated Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1996 - SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice -0.03139 -0.00529 0.06182 0.035591 0.03154 0.01524 0.17694 -0.01504 0.02849 0.00969
Non Rice Staple -0.13854 0.00275 0.14577 0.021064 0.10841 0.03482 0.06085 0.07211 0.1416 -0.02045
Fish -0.02551 0.00301 0.05015 0.033948 0.07528 0.08906 0.07231 0.14107 0.14883 0.04208
Meat 0.19325 0.10325 0.03248 0.034059 0.05207 0.00511 -0.04134 0.09496 0.19573 0.02146
Eggs and Milk 0.20213 0.02711 0.03995 0.06188 0.07245 0.04373 0.09638 0.12587 0.25445 0.01611
Legumes 0.15887 0.12373 0.07856 0.08391 0.064251 0.05187 0.04719 0.10086 0.08707 0.05211
Fruit & Vegetable 0.12416 0.06429 0.15037 0.01333 0.062749 0.08391 -0.0565 0.20827 0.06059 0.07872
Edible Oil 0.85061 0.06629 0.07202 -0.06358 0.081586 0.04505 -0.03334 0.16212 0.16344 0.02172
Tobacco -0.04127 0.04485 0.08022 0.08338 0.060831 0.05496 0.07015 0.09256 0.11713 0.01879
Prepared Foods 0.03393 0.03821 0.03672 0.07458 0.053361 0.02059 0.00886 0.04049 0.05083 0.01372
Spices 0.2653 0.0571 0.0608 0.0795 0.0395 0.0513 0.0317 0.0597 0.0207 0.2169

105
 Table 6.31 Compensated Cross Price Elasticities
Derived from the LA/AIDS Model for Food
Based on the 1999 - SUSENAS Micro Data: Urban East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0041 -0.0150 0.0547 0.0171 0.0627 -0.01622 0.0734 0.10105 0.06812 0.00772
Non Rice Staple 0.0372 0.0693 0.2597 -0.0601 0.0011 -0.01936 0.07101 0.24277 0.07448 -0.01875
Fish -0.0518 0.0261 0.0690 0.0598 0.0273 0.03534 -0.00325 0.15864 0.20951 0.02531
Meat 0.1703 0.0884 0.0622 0.0328 0.0813 0.0489 0.04613 0.07787 0.28329 0.02417
Eggs and Milk 0.0653 -0.0251 0.0661 0.0402 0.0963 0.03967 0.01652 0.17508 0.2907 -0.00966
Legumes 0.2222 0.0004 0.0281 0.0925 0.0894 -0.005 0.17769 0.07633 0.06021 0.05638
Fruit & Vegetable -0.0813 -0.0106 0.0513 0.0787 0.0521 -0.0071 0.0266 0.14992 0.13724 0.02184
Edible Oil 0.3170 0.0335 -0.0041 0.0639 0.0187 0.2164 0.02291 0.26555 0.16757 -0.01717
Tobacco 0.1862 0.0489 0.0846 0.0460 0.0845 0.0397 0.05509 0.11329 0.14755 0.02666
Prepared Foods 0.0458 0.0055 0.0408 0.0611 0.0512 0.0114 0.01841 0.02609 0.05385 0.00577
Spices 0.2064 0.0190 0.0600 0.0700 0.0481 0.0618 0.0399 0.0480 0.1102 0.3041

Table 6.32 Compensated Cross Price Elasticities

Derived from the LA/AIDS Model for Food
Based on the 1999 - SUSENAS Micro Data: Rural East Java, Indonesia
MODEL -U93 RICE N.RICE S. FISH MEAT EG & MILK LEGUMES FRUIT-VEG E. OIL TOBACCO PRE.FOOD SPICES
Rice 0.0005 0.0045 0.1090 0.0194 0.0247 0.0107 0.1897 0.0813 0.0022 0.0001
Non Rice Staple 0.0018 0.0400 0.0594 0.0212 0.0924 0.0276 0.1062 0.0952 0.1050 0.0244
Fish 0.0206 0.0528 -0.0218 0.0313 0.0947 0.0169 0.0780 0.1016 0.1465 0.0450
Meat 0.3523 0.0548 -0.0153 0.0666 0.0810 0.0613 0.0256 0.0907 0.1752 0.0095
Eggs and Milk 0.1091 0.0341 0.0381 0.1157 0.0604 0.0326 0.0831 0.1040 0.2507 0.0156
Legumes 0.1109 0.1186 0.0920 0.1126 0.0482 0.0085 0.1379 0.0864 0.0323 0.0177
Fruit & Vegetable 0.0679 0.0500 0.0232 0.1204 0.0368 0.0120 -0.0212 0.1439 0.0948 0.0312
Edible Oil 0.7995 0.1279 0.0710 0.0334 0.0623 0.1294 -0.0141 0.0983 0.1961 0.0396
Tobacco 0.2233 0.0747 0.0603 0.0771 0.0508 0.0528 0.0622 0.0641 0.2423 0.0303
Prepared Foods 0.0022 0.0301 0.0318 0.0545 0.0448 0.0072 0.0150 0.0468 0.0886 0.0104
Spices 0.2434 0.0698 0.0509 0.0787 0.0395 0.0495 0.0402 0.0607 0.0903 0.2359

106
6.5 Expenditure Elasticities
Expenditure elasticities on food groups were calculated at sample mean. The results were
presented in Table 6.33. In this table each of food group was cross tabulated across income
groups and areas. The point estimates imply that increased expenditure on food by one per
cent per household per week was associated with an increase of budget share indicated by
elasticity coefficient of each food group. The coefficient estimates of total food
expenditure lent support for a strong income or wealth effect on changing budget share.
This finding reinforces the view of the World Bank saying that raising income as the
critical factor in improving food and health status in poor countries.
As can be followed in the table, some food groups indicated a clear type of expenditure
elasticity, while the other groups were found to have a mixed one. In this group were
1. Rice, meat, edible oil, egg and milk, and legume which tended to belong to necessities,
irrespective of the income groups and, survey periods, and survey areas;
2. Tobacco and prepared food were luxurious.
3. The rest of food groups, i.e., fish, non-rice staple, fruits and vegetables, and spices, are
found to have mixed expenditure elasticities depends on income groups, survey
periods, and survey areas.
It is our interest, to know how the expenditure elasticities of each food groups change over
the time of survey, across space and income groups: Is there any pattern to follow? One by
one food group observations indicated that it is hard to draw a unique pattern. However, we
have tried to group the pattern according to a variation over time, across income group and
across areas.

107
 Table 6.33 Expenditure Elasticities of Food Demand
Across areas and Income Groups
Based on the SUSENAS Data: 1990, 1993, 1996 and 1999 East Java, Indonesia
RURAL AREA URBAN AREA
FOOD GROUPS Income Groups Income Groups
Lower Middle Higher Lower Middle Higher
1990 0.3985 -0.3295 0.5553 0.3969 0.6602 0.5377
1993 0.4630 0.6397 0.6250 0.4215 0.5797 0.4983
Rice
1996 0.6036 0.6389 0.7242 0.5833 0.8985 0.6462
1999 0.6166 0.4990 0.7255 0.7191 0.8188 0.7399
1990 1.5311 4.3742 1.3932 1.5541 1.9894 1.1514
Non Rice 1993 1.5567 2.4509 1.2062 1.1200 1.1952 0.8906
Staple 1996 0.8428 0.8224 0.7507 0.9818 1.0262 0.9615
1999 1.5184 2.2216 1.0799 0.6154 1.0729 0.8315
1990 1.3501 2.1696 1.2394 0.8699 0.4002 0.7300
1993 1.2608 1.5033 1.1071 0.8640 0.8797 0.6900
Fish
1996 1.0309 1.1369 0.7809 0.8804 0.5654 0.9090
1999 0.9823 0.9228 0.9582 0.6977 0.5647 0.7654
1990 0.8537 -0.0807 0.5661 0.9794 0.1568 0.6324
1993 0.7762 0.4386 0.6679 0.6939 0.4007 0.6797
Meat
1996 0.9049 0.5533 0.6132 0.8804 0.5654 0.9090
1999 0.8834 0.4718 0.6858 0.7959 0.3726 0.5301
1990 0.8928 0.5610 0.8619 1.1148 -0.2309 1.5522
Eggs and 1993 0.9321 0.5382 0.9536 1.0797 0.7833 1.2753
Milks 1996 0.9329 0.6066 0.9464 0.9733 0.5177 1.2181
1999 0.9057 0.5273 0.6138 0.7433 0.5572 0.5182
1990 0.9679 0.3770 0.8837 0.9141 0.0288 1.3180
1993 0.9595 0.8181 0.9493 0.8029 0.6273 0.8345
Legume
1996 0.9984 1.0735 0.9905 0.8733 1.1001 0.9232
1999 0.9383 0.8432 0.9043 0.8980 0.9719 0.9711
1990 2.0028 1.0565 1.6053 1.6985 1.9817 1.7858
Fruits and 1993 1.6015 1.1900 1.5857 1.4438 1.1565 1.2685
Vegetables 1996 1.3548 0.4919 1.3413 0.8733 1.1001 0.9232
1999 1.1068 0.6629 1.0049 0.9238 0.8801 0.9094
1990 0.8694 1.1383 0.9144 0.9062 0.2821 0.9611
1993 0.7793 0.7372 0.8335 0.7047 0.8381 0.7745
Edible Oil
1996 0.7365 0.5061 0.7151 0.7840 0.9584 0.8222
1999 0.5869 0.5618 0.7136 0.7336 0.7914 0.7771
1990 1.3665 1.1735 1.0992 1.3210 1.5230 1.0030
1993 1.4305 1.1702 1.2550 1.4733 1.4660 1.0917
Tobacco
1996 1.3846 1.3187 1.1779 1.2563 1.1015 1.1242
1999 1.4441 1.9316 1.6382 1.4590 1.9513 1.4986
1990 1.3115 1.6262 1.5579 1.4109 1.9341 1.4018
Prepared 1993 1.2594 1.2015 1.3402 1.4051 1.4426 1.5320
Foods 1996 1.4093 1.7066 1.5607 1.3617 1.6383 1.3909
1999 1.2367 1.2844 1.2904 1.2917 1.1352 1.2989
1990 1.4005 2.0695 1.2445 1.3468 1.9463 1.0975
1993 1.2865 1.1863 1.0808 0.7817 0.8390 1.3913
Spices
1996 0.9677 1.2033 0.9420 0.7470 0.9195 0.7545
1999 0.9207 0.8833 0.9390 0.6125 0.8190 0.6775

108
 Variation over Time
The following pattern is identified:
1. Patter A: Expenditure elasticities rise over time of survey.
In rural area, food group following this patter is rice in lower and higher income
groups. Those belong to the middle income group did not follow this pattern.
In urban area, this pattern was found among households of lower income group.
2. Pattern B: Expenditure elasticities decline over time of survey.
In rural area, food groups for which the expenditure elasticities performed this
pattern are fish in lower and middle income groups, meat in lower income group,
Eggs and Milk in middle income group, ―fruits and vegetables‖ and ―edible oil‖ in
lower and higher income, and spices in all of income groups.
In urban area, food groups for which the expenditure elasticities follows this
pattern are eggs and milk in lower and higher income groups, fruits and vegetables
in middle and higher income groups, edible oil in lower income group, prepared
foods in lower income group, and spices in all income groups.
3. Pattern C: the expenditure elasticities first decline and then rise:
In rural area: food groups for which the expenditure elasticities performed this
pattern are non-rice staple in all income groups, fish in higher income group,
In urban area, the food group for which the expenditure elasticities follow this
pattern is non-rice staple in middle income group.
4. Pattern D: the expenditure elasticities first rise and the decline.
In rural area: food groups for which the expenditure elasticities performed this
pattern are legume in middle income groups, meat in middle income group;
In urban area, food groups for which the expenditure elasticities follow this pattern
are legume in middle income group, edible oil in middle income group.

 Variation across Income Groups

A similar pattern is performed across income groups. The pattern of expenditure
elasticities might be characterized as:
1. Pattern A: expenditure elasticities rise across income groups.
In rural area: only rice in 1996 performs this pattern of development.
In urban area: only spice in 1993 performs this pattern of development.
109
2. Pattern B: expenditure elasticities decline across income groups:
In rural area: non-rice staples in 1996 and prepared food in 1999 perform this
pattern of development.
In urban area: eggs and milk in 1999, tobacco in 1993, and prepared foods in 1993
perform this pattern of development.
3. Pattern C: expenditure elasticities fall and then rise, as income group moves from the
lower, middle to the higher.
In rural area, expenditure elasticities with this pattern were found in food groups
of meat, egg and milk, Legume in 1990, 1996 and 1999, fruits and vegetables,
tobacco in 1993 and 1996.
In urban area, those to which expenditure elasticities performed this pattern are,
fish in 1990, 1996, and 1999, meat in all years, Eggs and milk in 1990 to 1996,
legume in 1990 and 1993, fruits and vegetables in 1993 and 1999.
4. Pattern D: expenditure elasticities rise, and then fall, as income group move from the
lower, middle to the higher group.
In rural area, food groups of which the expenditure elasticities followed this
pattern are rice, in 1993, fish in 1990, 1993, 1996, Legume in 1996, Tobacco, in
1999, prepared food in 1990 and 1996, and spice in 1990 and 1996.
In urban area, food groups of which the expenditure elasticities revealed this
pattern are, rice in all years of survey, non-rice staple, in all years of survey, fish in
1993, legume in 1996, Edible oil in 1993 and 1996, tobacco in 1990 and 1999,
prepared food in 1990 and 1999, and spices in 1990, 1996 and 1999.
The other pattern, namely a steady increase, a steady decrease are also performed but not so
frequent.
What we could delineate from observing the pattern is that the development in expenditure
elasticities is not regular. There is no grand pattern representing a large section of the
observation units.
This irregularity in the pattern may be explained in term of both quantitative and qualitative
changes in the household consumption baskets. For a given, quality of a commodity, the
immediate concern of household is to consume that commodity up to a certain minimum
desired level. If the households are not consuming the preferred commodity in the desired

110
minimum amount then expenditure on that commodity increases with an increase in the
level of income. Once households have achieved that desired level, given the quality of
commodity, the expenditure share of that commodity in total household expenditure
declines as income increases. However, as income continue to increase, households may
switch to better quality of the commodity and thus expenditure on the commodity starts to
increase again. This pattern is repeated as incomes continue to increase even further. In
other words, as long as there are various qualities available in the market, the irregularity
will exist.
Pattern A in which expenditure elasticities rise along the increase of income level may
represent a situation in which the households did not consume the goods in the desired
minimum amount. For a ggiven quality, the household desire to first consume a minimum
amount of good. As long as this desired minimum amount of good is not yet achieved, the
expenditure elasticity will rise as income level of the household increases. The reflection is
increasing expenditure elasticity across income level.
Pattern B, the decreasing expenditure elasticity, may represent a well-known Engel‘s Law.
The household in this situation starts to consume the food good in the desired minimum
amount, for a given quality. As the income level increases, the expenditure elasticity on that
food decreases.
Pattern C, in which expenditure elasticity first falls and then rises, as income group moves
to a higher level, may represent a situation, in which a household has started consuming
food good at minimum amount. So it started first to reduce the consumption of food good
of a given quality. Therefore, the expenditure elasticity declines. As the income level
further rises, the household switchs to the food group of higher quality
Based on the above estimates assessment, we suggest that households in different income
groups performed different patterns and that in general, they alter their consumption
bundles both quantitatively and qualitatively in response to changes in income.

111
6.6 Family Size Elasticity
"Family" includes the members of the household who normally ate from the same kitchen
(sharing the same pot). The family size thus, the number of those who normally (almost
every day) eat from household‘s common kitchen. As in this study the size is measured
only by the number, irrespective of age and gender of the member, it is implicitly assumed
that the individual member in the household has an equal demographical character. This
approach may be a crude one.
In the literature (Brown and Deaton, 1972; Deaton and Muellbauer, 1980a), one introduces
what is called adult equivalent scale or the similar scale to take into account the variations
in demographic aspects of family member (mostly age and sex). Because of statistical
problems in estimation and availability of data, however, few studies have attempted to
estimate the demand function with adult equivalent scales.
Prais and Houthaker (1955) emphasized the need of including family size in the
specification of demand function on the grounds that households‘ total expenditure and
household size are positively correlated and exclusion of the latter may bias the results. In
addition, variations in household size have comparatively larger effects on the consumption
of certain commodities than variations in the total expenditure. For a given expenditure,
larger households tend to spend a higher proportion of their total expenditure on staple food
compared to smaller households.
The coefficient of household size captures the effect of economies of scale in consumption
among larger households. According to Houthakker (1957), the coefficient of household
size represents two effects determining demand, namely, the specific effect and the income
effect. The first effect might be related to the need of diversifying commodities, as the size
of family i
The income effect refers to a reducing real purchasing power household as family size
increases. If specific effect is dominating the income effect, the gross effect of increasing
family size is positive. Otherwise, the effect will be negative.
Table 6.34 reports the estimation result of the elasticities of eleven food groups from the
LA/AIDS model with respect to household size.

112
The performance of demand elasticities with respect to household‘s size may be grouped
into four categories.
7. positive elastic:
8. positive inelastic
9. negative elastic
10. negative inelastic

Table 6.34 Demand Elasticities of Food Items on Household Size

Estimated from the LA/AIDS Model
Based on the 1990 –SUSENAS Data, Urban East Java, Indonesia

RURAL AREAS URBAN AREAS

MODEL
1990 1993 1996 1999 1990 1993 1996 1999
Rice 0.6137 0.9884 0.7077 0.6118 1.2113 1.2464 0.9732 0.6766
Non Rice Staple 0.8741 0.2032 0.7565 0.3131 -0.0837 0.0359 0.1104 0.6854
Fish -0.1752 -0.2610 0.1900 0.0696 0.1319 0.3152 0.0125 0.3055
Meat 0.0702 0.2557 0.1604 0.1213 -0.1831 0.2806 0.0985 0.3527
Eggs and Milk -0.0955 -0.1977 -0.1422 0.1485 -1.1515 -0.5039 -0.4378 0.3961
Legumes -0.2965 -0.2105 -0.1159 -0.1192 -0.5626 0.2374 0.2261 0.0614
Fruit & Vegetable -1.6481 -1.2125 -1.0404 -0.4007 -1.2943 -0.8529 -0.6866 -0.0846
Edible Oil 0.0920 0.1079 0.1861 0.3298 -0.1552 0.2901 0.2974 0.3277
Tobacco -0.6349 -0.8030 -0.6322 -0.7815 -0.3907 -0.5046 -0.4235 -0.5292
Prepared Foods -0.6277 -0.5695 -0.7180 -0.4768 -0.5540 -0.7978 -0.5174 -0.5859
Spice -0.0866 -0.1971 0.1238 0.0505 0.0083 -0.1748 0.3047 0.4403

Elasticities variation over time of each food groups is shown in the following pattern.
In rural area:
(i) There is no food group performing category 1, positive elastic;
(ii) Rice, non-rice staple, meat, edible oil, overall years of survey, fish in 1996 and 1999,
spices in 1996 and 1999, performed the second category of household ‗s elasticity:
positive inelastic;
(iii) Category 3 of household‘s elasticity (negative elastic) is performed by food group
fruits and vegetables;
(iv) Otherwise: fish in 1990, 1993, eggs and milk in 1990, 1993, 1996, legume, tobacco
prepared foods, and spices in 1990, 1993, performed negative inelastic.

In Urban area:
(i) Rice revealed a positive elastic household‘s size elasticities in 1990 and 1993.
113
(ii) Non rice staple, in 1993, 1996, 1999, fish in all periods of survey, meat, legume,
edible oil, in 1993, 1996, 1999, exhibited positive inelastic;
(iii) Eggs and Milks in 1990, Fruits and Vegetables, in 1990, exhibited a negative elastic
household‘s elasticity.
(iv) Otherwise: non rice staple, in 1990, meat in 1990, eggs and milk, fruits and
vegetables, tobacco, prepared food and spices in 1993, these groups exhibited
negative inelastic.
These patterns indicate that for rice, non-rice staple, meat, edible oil, fish, spices, both in
rural or urban areas, the specific effects was dominating the income effects. Since, the
magnitudes were less than unity, it follows, and that for these food goods economic of scale
was there, as the number of household‘s member increases. Varying magnitudes of
household‘s size elasticities over periods of survey may also suggest that the degree of
economic of scale in consumption are not only different across commodities but may also
be different over time.
For food goods with a negative inelastic elasticity of household‘s size, such as Eggs and
Milks in 1990, Fruits and Vegetables, in 1990 etc., this implies that an increase in family
size, holding price and income unchanged, makes the family poorer, meaning that
reallocating the expenditure for additional necessities, cannot but spend less on these food
goods.

114
 CHAPTER 7. POLICY EXERCISE

This chapter is specified to demonstrate the usefulness of demand study for policy analysis.
Parameter estimates derived from the demand system may be treated into a simulated
scenario to answer policy issues under investigation. There are four ways for demand study
to address policy issues (Rauniker and Huang, 1987). The first is by providing a demand
system specified for a certain policy issues. The second is by adaptation of the estimated
demand model, to permit development of an empirical framework so the policy issue can
be addressed. The third is by providing elasticities matrix to answer issues related with
quantity dependent perspective. The fourth is by providing flexibility matrix for a price
dependent issue. The policy assessment in this chapter follows the second type.
The policy issues and corresponding policy instruments are numerous, but one which is
relevant to this study is pricing policy. Price change introduced via government
intervention may affect wellbeing of private households. This may be beneficial or adverse.
To know how this works, one needs to measure and assess this welfare change. Estimation
results of empirical study may help to answer typical questions commonly addressed in
welfare analysis i.e. Given price policy, who is the gainer, the looser and how big is the
magnitude of gain or loss? In what follows the above question was addressed.

7.1 Background
The immediate impact of economic crisis on Indonesia‘s food and agricultural sector is the
fast market liberalization of the strategic commodities like wheat, cooking oil, sugar,
soybean, cloves. This deregulation is an unavoidable choice because of
(1) concern on budget of central authority,
(2) Indonesia‘s economic commitment with lending institutions (the letter of Intent) and
trading partner countries -bilateral as well as multilateral, mainly in the frame of
WTO, AFTA and ASEAN,
(3) Consumerism movement, i.e. domestic nongovernmental movements demanding a
liberalization of domestic market from protection and monopoly, which rises along
with an increasing degree of democratization and freedom to organize in the country.

115
Previously, import licensing, tariffs, export regulation (ban, tax, and licensing control) were
typical policy instruments in international trade of this country. Administered price system
for some food commodities combined with its marketing monopoly is instruments typically
applied for domestic food market.
Due to letter of Intent, these all have been deregulated in favor of free market mechanism,
except for rice, - a commodity recognized as being sensitive to the country. Thus, to think
of policy intervention that matches this situation, two policy proposals was considered in
the scenarios.

7.2 The Scenarios

The proposed policy scenarios are designed to represent the situation as realistic as
possible. However, their nature is illustrative; the scenarios are not representing the
government position. These scenarios are forseen and are deemed compatible with the
existing intervention space available for the government under the existing political
framewoks.
Three pricing policy scenarios are considered to be implemented in this exercise. The
objective is to see how these scenarios affect the purchasing power of households. It must
be noted

Scenario 1
The government phases out the import duty for rice that leads to a decrease of rice price by
30 per cent.
Import tariff by this moment is the only intervention that can be exercised by the
government of Indonesia, in accordance with WTO agreement and the letter of Intent. If the
import duty on rice is eliminated the market prices of rice is expected to decrease to match
the competitive equilibrium prices. According to some studies (Choudori, 2000) the fall of
rice price may range from 20 per cent to 40 per cent, due to the tariff elimination. In this
exercise, we considered to place a percentage of 30-price reduction as a result of it.

Scenario 2
This is a combination of imposition of a new tobacco tax and an elimination of the import
duty on rice, which induce the price decrease of 30 per cent. Imposition of tobacco tax

116
maybe called a ―sin‖ tax policy, in which government justifies this policy on the argument
that tobacco and related products (cigarettes, cigar, etc.) are ―sin‖ luxurious goods. Its
budget share is relatively big (see Table 7.1). Many proofs indicated the danger of tobacco
products on human health. But infectivity of law enforcement failed to protect non-
smokers from the danger of negative externalities of smoking. Imposition of tax on tobacco
is considered as a way to reduce its consumption, to increase government revenue, and to
facilitate a healthy environment. Re-allocation of this budget share to other food groups or
may contribute a better diet and health. This may justify the imposition.
In this exercise the tobacco tax is assumed to induce a price increase of 40 per cent, ceteris
paribus.

Scenario 3
The policy makers impose simultaneously a 50 % import duty on rice and imposition of
tobacco tax. The imposition of 50 % represents a contra situation from market liberalism as
a result of some reasons: (i) there is an increasing movement advocating for rice farmers to
get protection from global market. (ii) There is still a room for this imposition as far as rice
concerned. In the World Trade Organization agreement, Indonesia is actually allowed to
impose the tariff level of 110 per cent until 2003. According to the agreement on Indonesia,
from 2004 to 2010 the tariff may be maintained to 95 %. Also, with some conditions the
imposition is still possible. So imposing of more 50 per cent than the old tariff level on rice
is possible to occur given a domestic political situation.

Reference situation
1. The year of reference is 1999, meaning that the household‘s food consumption behavior
of rural and urban East Java is represented by the estimated AIDS model of the year
1999.
2. The equilibrium level of domestic rice price is determined under influence of ceiling
price instrument and import duty of 30 per cent of world price.
3. Domestic market for all other food commodities is liberalized. So the domestic price
level is determined by the supply and demand at the border. The level follows the
world‘s price equilibrium.

117
4. As the base line, the household budget share on food groups of the reference situation is
presented in table 7.1.
Table 7.1 the Weekly Average of the Household's Budget Share on Food Groups
Across Income Groups and Areas:
The Observed Data 1999 - East Java, Indonesia
URBAN AREA RURAL AREA
FOOD GROUPS Income Groups Income Groups
Lower Middle Higher Lower Middle Higher
Rice 0.23 0.21 0.17 0.25 0.25 0.22
Non Rice Staple 0.03 0.02 0.02 0.09 0.07 0.05
Fish 0.05 0.06 0.07 0.05 0.05 0.06
Meat 0.06 0.06 0.07 0.07 0.08 0.08
Eggs and Milk 0.05 0.05 0.06 0.04 0.04 0.05
Legumes 0.06 0.06 0.05 0.05 0.05 0.06
Fruits and Vegetables 0.04 0.04 0.05 0.04 0.04 0.04
Edible oils and Fat 0.05 0.05 0.04 0.06 0.06 0.05
Tobacco and Betel 0.10 0.12 0.12 0.08 0.09 0.10
Prepared Food 0.30 0.30 0.32 0.24 0.24 0.26
Spices 0.03 0.03 0.02 0.03 0.03 0.03
Source: Own Calculation based on the SUSENAS data: East Java, 1999.

7.3 Welfare Measurement Revisited

Expenditure function or cost function is a natural way for measuring welfare change. If the
cost of affording a bundle of food goods for someone changes, then changes also the
wellbeing of him/her. A relevant question to that is ―how much money do we need (or can
we afford to give up) in order to get back to the reference level of wellbeing?‖
As already discussed in two previous chapters, some approaches are available for
evaluating welfare effects of any policy regime based on demand study. Three of them are
consumer surplus, compensating variation, equivalent variation, and living index. In this
exercise compensating variation concept is used as a tool of our analysis.

118
Compensating Variation (CV)
If p0 is the initial reference situation and p1 is the final one, and W or C (u0i, pi0) is the
minimum (maximum) amount of money that has to be given to (taken away from) an
individual to make them as well off as before the price rise (fall).
Rewriting (5.22) compensating variation is expressed as
CVi  c (ui0 , pi1 )  c (ui0 , pi0 ) (7.1)

 c (ui0 , pi1 )  c (ui1 , pi1 )

This is a welfare cost of price change from the reference situation to the final one. Because
compensating variation is money metric, its expression is dependent on an absolute
expression in term of country‘s currency unite. This is less comparative. To avoid this, one
can transform it in relative term, by using for example, price index, which is metric free.
So (7.1) can be expressed algebraically in term of price index as

C ( pi1 , ui0 ) C ( pi1 , ui0 )

 (7.2)
C ( pi0 , U i0 ) C ( pi1 , U i1 )

In this transformation, welfare change took place by households is interpreted as a changing

purchasing power. Because the level of utility is not observable, the price index is
approximated17. To approximate the price index we applied Fischer Ideal Price Index,
which is a geometric means of Laspeyres- (PL) price index
p1
PL =  wi0
i P0

and the Paasche (PP) price index

 p 
Pp = 1 /   wi1 0  .
i P1 

So, Fischer ideal price index is = PL .PP ,

where
w0i = Budget share at preference condition based on observed data. For the calculation, this
share is represented by mean value of observed data.
w1i = food group‘s budget share after price change to be estimated from the AIDS model.

17
See further Deaton and Muellbauer (1980a) in Grings (p.128)

119
The ideal Fischer Ideal Index = ( PL .PP ,) represents a changing purchasing power as an
approximation of welfare change.

7.4 The Welfare Effects of the Scenarios

Table 7.2 and 7.3 present the estimated welfare change of the policy scenarios in term of
price indices. The price index lower than 100 indicates a cheaper price, and thus a welfare
gain for the Households well being. The index 100 indicates an unchanged well being.
Otherwise, it indicates a welfare loss for corresponding households. Based on these indices,
it can be assessed as follows.

Scenario 1
1. The elimination of import duty on the imported rice has relatively similar welfare
effects across different income groups of household in rural as well as urban areas, both
in the direction and magnitude of change. That is, it will increase the purchasing power
of all households across income groups and areas with an increase ranging from 10 per
cent (by the rural households in the higher income group) to 19 per cent (by the rural
households in the middle-income group). By referring to table 7.1, it appears that the
magnitude of welfare impact on the households correlates with the budget share of the
corresponding food group at the base line situation. For example, the rural households
in the middle-income group have a weekly average of budget share on rice as much as
25 per cent, the biggest budget share among the observed household groups. The
welfare effect corresponding to the scenario 1 on those of this group is the highest,
namely, 19 per cent gain of purchasing power. Since the household with the higher
income consumes rice less percentage of their food budget, the impact on them is also
less.
2. A comparison between the urban and the rural areas gives no conclusive picture with
respect of the welfare effect induced by this scenario. Among the higher income
households, the gain from this policy scenario is greater in rural area than that of urban
area. Meanwhile for those belong to the lower and the middle income groups the gain
is higher in rural than in urban areas The difference of the welfare effect among income

120
 groups is more extremely seen in the rural area, than in the urban. In all, an elimination
of duty on imported rice brings about gain to households.
3. While the effect is positive for the households, this policy however, will likely be
objective by rice farmers, the likely looser of this policy. But, if this policy is
accompanied by a well managed direct or indirect income transfer to the farmers to
protect them from declining rice price, this combination of policy may be beneficial for
all. Compared to the current general support price (which tends to create an urban bias),
a well managed income transfer may be less costly to the government in term of fiscal
burden. Income transfer to the farmers, though previously not commonly implemented,
is at the moment being thought of as an alternative measure, especially to mitigate the
crisis impact on the farmers. Thus, its implementation is likely allowable and accepted
by domestic politicians.

Scenario 2:
1. Tables 7.2 and 7.3 exhibits gains of purchasing power for all income groups. It is
exhibited that the urban households gain better than their counterparts in the rural. This
may be caused by the fact that the budget shares for tobacco and betel of those who are
in urban area are larger than the budget share of tobacco and betel of those who are in
rural area.
2. Compared to the welfare change induced by the scenario 1, the welfare change induced
by this scenario is not obviously different. At the other side, this scenario may bring
revenue to the government. So, besides the gain enjoyed by consumers, this scenario
bring also gain to the government. The elimination import duty and imposition of ―sin
tax‖ for tobacco and betel group brings mostly gain, with small loss suffered by
households. Thus, it may be a good policy option to implement.

Scenario 3
For this scenario, welfare change exhibits losses for most of households across income
groups and areas. The highest loss is suffered by households of rural- low income group,
urban-low income group, and rural middle income group who should afford 19 %, 15 %
and 14% more budget respectively for them to stay at the same well being as before the

121
imposition. It is shown, that the imposition of the tobacco tax does not help much, at least
from the perspective of consumer households.

6 Concluding Words
It is shown that demand estimates derived from the LA/AIDS model in combination with
price index concept may be used to measure a welfare change of pricing policy option. The
results are useful for policy makers, policy analyst and consumer interests. Three policy
scenarios exercised in this assessment are,
1. the elimination of duty on imported rice leading to a decrease of rice price by 30 per
cent,
2. the above option is combined with the imposition of tobacco‘s ―sin tax‖ that lead to rice
price decrease by 30 per cent and an increase of tobacco and betel price by 40 per cent,
3. The imposition of import duty on imported rice combined with an imposition of
tobacco‘s ―sin tax‖ that lead to 40 per cent price increase.
If the first scenario is implemented all private households across the income groups and
areas may receive benefits from the decreased market price of rice.
If the second scenario is implemented most of households gain benefits. Households in
urban area will benefit better than those in rural area.
If the combination of import duty and the tobacco tax is implemented the larger loss of
purchasing power will incur to the household from lower income group in rural area.

7 Notice
In this assessment, the demand models of different income groups are assumed to have the
same pattern of consumption, as they are represented only by one model. This may be
misleading, because each income group might have a unique consumption behavior that
requires a different treatment for each. However, this results support the proposition, that
tarification for households to pay more for the same good of the same quality.

122
 CHAPTER 8. CONCLUSIONS

The descriptive statistics, the results of estimation and their corresponding discussions has
lead to the following conclusions:
1. The demand systems that we specified and estimated take the form of budget share of
eleven food group as being independent on the own price and ten prices of other food
groups in this system, the total expenditure on food, income groups where the
household belongs, and the number of household‘s member (household size) of each
household. The eleven food groups are the groups of rice, non-rice staples, fish, meat,
eggs and milk, legumes, Fruits and Vegetables, Edible Oils, tobacco and betel, prepared
food, and spices and miscellaneous. Rice has the highest share of total food
expenditure.
2. As clearly shown in the model, the price which was taken out from the cross sectional
data could sufficiently estimate the coefficients necessary for computing the price
elasticities of demand.
3. Estimated own price elasticities for the LA/AIDS model based on micro data suggest
that food groups, with exception on the edible oil and Eggs and Milks, are generally
price inelastic. All estimated own price elasticities are negative. The difference in
magnitudes between the Hicksian compensated own price elasticities and that of
ordinary own price elasticities suggest the presence of income effects in each of price
change. The existence of cross price elasticities confirmed that the demand for food
commodities is responsive to the relative prices change. The response is however weak.
These cross price elasticities are lower compare with own price elasticities. Thus,
consumer demand for particular food groups in general were more sensitive to the
change in own price than other prices. The cross effect of rice price to the other food
groups are in general bigger than otherwise. This again suggests the prevalence of rice
as a centre food commodity in Indonesia.
4. The coefficient estimates of total food expenditure lent support for a strong income or
wealth effect on changing budget share. This finding reinforces the view that raising
income, instead of just a pricing policy as the critical instrument in the improvement of
food and health status in poor countries. Across commodity, one may draw a pattern

123
 that includes rice, meat, edible oil, egg and milk, and legume into a one group of
necessities, irrespective of the income groups, survey periods, and survey areas.
Tobacco and prepared food tended to be luxurious. The rest of food groups, i.e., fish,
non rice staple, fruits and vegetables, and spices, have in general a mixed expenditure
elasticitities, depends on the income groups, survey periods, and survey areas. More
general indications are that there is no general systematic pattern. This irregularity in
expenditure elasticities may be due to the effects of quality changes in consumer‘s
spending.
5. As evident from the signs of the elasticities estimates, household size has positive
effects for rice and non rice staples and edible oil, and negative effects on most other
food groups. So, food consumption for households of with big family member
consumes merely carbohydrate rich diets.
6. As indicated in the last chapter, the use of results from this study for policy analysis has
shown a reasonable result. Furthermore, by application of micro data in demand
models, the economic view is widened and the frame work of micro analysis is
maintained.
7. As this study did not employed exhaustively the existing methods available for study
like this type, we cannot compare directly which methods conveys the most reliable
results. Therefore, the estimates should be used with caution and are perhaps best
regarded as providing orders of magnitudes.

124
 REFERENCES

Alderman, H. and C. Peter Timmer.1980. Food Policy and Food Demand in Indonesia.
Bulletin of Indonesian economic studies Vol. XVI, No.3, November 1980, p.83-93
.Department of Economics, Research School of Pacific Studies, and Australian
National University.

Barten, A.P.1969. Maximum Likelihood Estimation of a Complete System of Demand

Equations. European Economic Review 1:7-73.

Blanciforti, L. and Green, R.1983. The Almost Ideal Demand System: A Comparison and
Application to Food Crops. Agricultural Economic Research 35:1-10.

Blanciforti, L. and Green, R.1986. US Consumer Behaviour over the Post-war Period: An
Almost Ideal Demand System Analysis, Giannini Foundation Monograph Nr.40,
Berkeley, California.

Blundell, R.W. Pashardes, P. and Weber 1993. What Do We Learn about Consumers
Demand Pattern from Micro data? American Economic Review 83:570.97.

Burdett, K. and Judd, K.L. 1983. Equilibrium Price Dispersion. Econometrica. 51:955-59.

Chang H.S (Christie). 2000. An econometric Analysis of the Competitive Position of

Australia Cotton in the Japanese Market, Working Paper Series in Agricultural and
Resource Economics, University of New England.

Debreu, G. 1959 Theory of Value: An axiomatic analysis of economic equilibrium. Cowles

Monograph. No.17. New York: Wiley.

Caves D.W and L. Christensen. 1980. Global Properties of Flexible Functional Forms.
American Economic Review. (1980):422-432

Cowling, K. and Raynor, A. J. 1970. Price, Quality, and Market Share. Journal of Political
Economy. 78: 1292-309.

Cox, T. L. and Wohlgenant, M.T. 1986. Price and Quality Effects in Cross-sectional
Demand Analysis. American Journal of Agricultural Economics. 68:908-19

Currie, 1972. The Analysis of family budget in The Demand for Food, p.19, Manchester
University Press.

Deaton, A. and J. Muellbauer. 1980. Am Almost Ideal Demand System. American

Economic Review 70:312-326

125
____1986. Demand Analysis. in Handbook of Econometrics, Vol. III, ed.Z. Griliches and
M.D. Intriligator, 1768-1829. New York: Elsevier Science Publisher.

____1987. Estimation of Own-and Price Elasticities from Household Survey Data. in

Journal of Econometrics 36 (1987):7-30

____1988. Quality, Quantity, and Spatial Variation of Price. American Economic Review:
78:418-30

Dillon, H.S. 1992. The Indonesian Strategy in the Field of Agriculture. Indonesian
Quarterly. 20: 438-46.

Eales, J.S. and L.Unnevehr. 1988.Demand for Beef and Chicken Products: Separability and
Structural Change. American Journal of Agricultural economics 70:521-53

George, P.S. and King, G.A. 1971. Consumer Demand for Food Commodities in the US
with Projections for 1980. Giannini Foundation Monograph No.26. University of
California, Berkeley

Goldman, F. and Grossman, M. 1978. The Demand for Paediatric Care: A Hedonic
Approach. Journal of Political Economy. 86 : 259 - 80.

Green, R. and Alston, J.M. 1990. Elasticities in AIDS Models‘. American Journal of
Agricultural Economics.72:41-52

Griffiths et al, 1993: Learning and Practicing Econometrics John Wiley and Sohn Inc,
New York, 1993.

Grings, M. (1993): Die Nachfrage nach Nahrungsmitteln in Ost- und Westdeutschland.

Vergleichende Analyse auf der Grundlage eines ökonometrischen Modells,
unpublished habilitation thesis, Göttingen.

Hausman, J. A. 1978. Specification Tests in Econometrics. Econometrica 46:1251-1271

Houthakker,.H.S. and L.D. Taylor. 1992. Consumer Demand in the United States: Analysis
and Projections, Second Edition. Harvard University Press, Cambridge,
Massachusetts.

Jayasuriya, S. and Manning, C. 1988. ―Survey of Recent Development―. Bulletin of

Indonesian Economic Studies. 24 : 3 - 44.

Judge, G.G., Griffiths, W.E., Hill, R.E. Luthkepohl, H. and Lee, T.C. Introduction to the
Theory and Practice of Econometrics. 2nd. Edition. New York, John Wiley & Sons.

Just, R.E, D.E. Hueth and A. Schmitz. 1982. Applied Welfare Economics and Public
Policy. Prentice Hall, Inc., Engelwood Cliffs, New Jersey

126
Kmenta, J. 1986. Elements of Econometrics. 2nd. Edition. New York, Macmillan Co.

Kyrk, H.1924.A Theory of Consumption. London. Sir Isaac Pitman & Sons, Ltd.

Leontief, W. 1993 „ Can Economics be Reconstructed as an Empirical Science?

American Journal of Agricultural Economics. 75th Anniversary Issue. 75 : 2 – 5

McKenzie G.W. 1983, Measuring Economic Welfare. New Methods. Cambridge

University Press, Cambridge.

Mergos G.J., G.S. Donatos. Demand For Food in Greece: An Almost Ideal Demand System
Analysis, 1989. Journal of Agricultural Economics 40: 983 – 993.

Nasution, A.The Adjustment Program in Indonesian Economy Since the 1980s. In Hill, H.
(ed). Indonesia Assessment 1991. Political and Social Change Monograph 13.
Research School of Pacific Studies. Australian National University, Canberra.

Parker, Survey of Recent Developments in Indonesia. Bulletin of Indonesian Economic

Studies, Vol.27 No.1, April 1991

Pollak, R.A. and Wales, T.J. 1978. Estimation of Complete Demand System from
Household Budget Data:
The Linear and Quadratic Expenditure Systems. American Economic Review
68:348-59

_____1980. Comparison of the Quadratic Expenditure System with Alternative

Specification of Demographic Effects Econometrica 48:595-612.

_____ 1981. Demographic Variables in Demand Analysis.Econometrica: 49:1533-1551

_____ 1992. Demand System Specification 2nd Estimation. New York, Oxford University
Press

Prais and Houthakker, H.S. 1955. The Analysis of Family Budgets. Second Impression and
Abridged. Cambridge, Cambridge University Press.

Pratt, J.W., Wise, D.A. and Zeckerhauser, R. 1979. Price Differences in Almost
Competitive Markets. Quarterly Journal of Economics.XCIII:189-211

Raunikar, R. and Huang C.H.. Food Demand Analysis, problems, Issues, and Empirical
Evidence, Iowa State University Press/Ames, 1987.

Reinganum,.J.F. 1979.A Simple Model of Price Dispersion. Journal of Political Economy.

87:851-58.

Salop, S. and Stiglitz, J.E. 1977. Bargains and Rippoffs: A Model of Monopolistically
Competitive Price Dispersion. Review of Economic Studies. 44: 493 – 510

127
Teklu T. and Johnson, S. R. 1986. A Review of Consumer Demand Theory and Food
Demand Studies on Indonesia. Staff Report 86-SR7. Centre for Agricultural and
Rural Development, Iowa State University, Ames.

_____ 1987. Demand System from Cross Section Data: An Experiment for Indonesia.
Working Paper 87-WP. Centre for Agricultural and Rural Development, Iowa
State University, Ames

_____1988. Demand System from Cross-Section Data: An Application to Indonesia.

Canadian Journal of Agricultural Economics. 36:83-101.

Theil, H. 1975. Theory and Measurement of Consumer Demand. North-Holland Publishing

Company. Amsterdam.

Thomas,. R. L. 1987. Applied Demand Analysis. New York, Longman Inc.

Thorbecke, Erik at al. Gen. Ed. Morrison Christian.1992.Adjustment and Equity in

Indonesia OECD, Paris

Wang, Q.B., Halbrendt, C. and Johnson, S.R. A non-tested test of the AIDS vs. The
Translog Demand System. Economic Letters 1996, 52, 139 – 143.

Zellner, A. 1962. An Efficient Method of Estimating Seemingly Unrelated Regression and

Test for Aggregation Bias. Journal of American Statistical Association 57:348-68

128