Citation
Demand for food commodities in the Dominican Republic and some implications for nutritional policies

Material Information

Title:
Demand for food commodities in the Dominican Republic and some implications for nutritional policies
Creator:
Chanlatte, Marino, 1949-
Publication Date:
Language:
English
Physical Description:
vii, 151 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Agriculture ( jstor )
Commodities ( jstor )
Elasticity of demand ( jstor )
Food ( jstor )
Imports ( jstor )
Income elasticity of demand ( jstor )
Income estimates ( jstor )
Price elasticity ( jstor )
Rice ( jstor )
Seas ( jstor )
Dissertations, Academic -- Food and Resource Economics -- UF
Food and Resource Economics thesis Ph. D
Food consumption ( fast )
Nutrition policy ( fast )
Dominican Republic ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1983.
Bibliography:
Includes bibliographical references (leaves 146-149).
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Marino Chanlatte.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
029289703 ( ALEPH )
09880982 ( OCLC )

Downloads

This item has the following downloads:


Full Text














DEMAND FOR FOOD COMMODITIES IN THE DOMINICAN REPUBLIC
AND SOME IMPLICATIONS FOR NUTRITIONAL POLICIES











BY

MARINO CHANLATTE











A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY


University of Florida

1983


















ACKNOWLEDGMENTS



The author wishes to express appreciation and

gratitude to Dr. Max R. Langham, chairman of his supervisory

committee, for the inspiration, encouragement, and

contributions throughout work on this dissertation. Special

thanks are due to Dr. W. W. McPherson, a member of the

supervisory committee, for his guidance during the earlier

period of the author's Ph. D. program.

The author also wishes to thank Dr. Carlton G. Davis

and Dr. Gustavo Antonini, members of his supervisory

committee, for their help and contributions.

Special thanks are due to Mr. Rom Alderman for his

assistance in computer programming, and to Mr. Laurent Ross

for cleaning the data and preparing the computer tape used

in this study.

Dr. Arturo Martinez Moya and Gumersindo del Rosario of

the Centrai Bank of the Dominican Republic, and David Jessee

of USAID were very helpful in providing data. Dr. Mervin

Yetley of USAID, Dr. Per Pinstrup-Andersen, Dr. Philip

Musgrove and Mr. Ivan Castelar provided fruitful comments.



ii











Mrs. Pat Smart and Dr. Rekha Mehra edited portions of this

work.

Special gratitude to Dr. Ruben Nunez, of the

Secretariat of State for Agriculture of the Dominican

Republic and to Felipe Manteiga, of USAID, for their

encouragement, moral support, and contributions.

The author is deeply indebted to the Secretariat of

State for Agriculture of the Dominican Republic for

providing financial support without which this endeavor

would not have been possible.

Finally, the author wishes to extend appreciation and

gratitude to his wife Nellie for her encouragement,

sacrifice, and labor and to his daughters Michelle and Lena

and his son Marino for their love and understanding.


























iii














TABLE OF CONTENTS

PAGE

ACKNOWLEDGMENTS........................................ ...... ii


ABSTRACT.................. .... ...... .... ......... ..........vi


CHAPTER

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

Problem Setting and Objectives.......................
Characteristics of the Dominican Economy.............5
Plan of Dissertation................................ 19

II THEORETICAL FRAMEWORK AND REVIEW OF LITERATURE........21

Household Budget Surveys............................ 22
Engel Curves........................................ 23
An Alternative Method of Estimating Engel Curves....29
Application of the Frisch Method.................... 35
Price Elasticities from Cross-Sectional Data........40
Demand Estimates in the Dominican Republic..........43
The Limited Dependent Variable Model................45


III METHODOLOGY........................................... 48

The Restrictions of Consumer Theory................. 48
The Demand Models ............................. 49
Use of the Tobit Model.............................52
The Elasticities Matrix............................ 53
The Effect of Price and Income Changes on Calorie
and Protein Intake................................ 59
The Data ............. ............................ .. .61
Note .................................... ........... 63

IV EMPIRICAL RESULTS..................................... 64

Tobit Model Estimates..... ......................... 64
The Full Matrix of Demand Elasticities..............67
The Family-Size Elasticities........................78
The New Method of Estimating Engel Elasticities.....78


iv












V IMPLICATIONS FOR POLICY.................................81

The Effect of Changes in Prices on the Calorie and
Protein Intake...................................82
The Effect of Changes in Income on the Calorie and
Protein Intake...................................86
The Effect of a Change in Income and Prices on the
Consumption and Imports of Selected Food
Commodities................................... ...... 89

VI SUMMARY, CONCLUSION, LIMITATIONS AND RECOMMENDATIONS ....99

Summary and Conclusions.............................99
Limitations of the Study............................ 101
Suggestions for Further Research................... 103

APPENDIX 1. COMPOSITION OF FOOD GROUPS.................... 106

APPENDIX 2. REGRESSION COEFFICIENTS FROM TOBIT MODEL....... 111

APPENDIX 3. ROTATION OF AXES ............................... 142

APPENDIX 4. RESULTS FROM THE NEW METHOD OF ESTIMATING
ENGEL ELASTICITIES........................... 145

REFERENCES................................................... 146

BIOGRAPHICAL SKETCH........................................ 150























v














Abstract of Dissertation Presented to the
Graduate Council of the University of Florida in
Partial Fulfillment of the Requirements for
the Degree of Doctor of Philosophy






DEMAND FOR FOOD COMMODITIES IN THE DOMINICAN REPUBLIC
AND SOME IMPLICATIONS FOR NUTRITIONAL POLICIES

By

Marino Chanlatte

April 1983



Chairmam: Max R. Langham
Major Department: Food and Resource Economics

The nutritional problem has been identified as one of

the most important development issues in the Dominican

Republic. The nutritional status and other development

problems associated with food consumption projections and

the effect of food demand on trade are of major concern for

the Dominican policymakers. This study developed a full

matrix of demand elasticities for food commodities at a

disaggregated level. The primary concern of this study was

to develop these estimates and to demonstrate their

usefulness in evaluating nutrition and food consumption

policies.




vi











Coefficients estimated from 30 demand equations were

used to estimate the elasticities matrix. These demand

equations were estimated using the tobit model and data from

a national household consumption survey. This matrix was

comprised of 996 price and income elasticities for 30 food

commodities and groups of commodities, the all-food

aggregate, and one nonfood group. Family size elasticities

were also estimated. In general, the estimated cross and

direct price elasticities for most of the commodities were

within a logical range.

The effects of changes in prices and income on calorie

and protein intake were estimated using the elasticities.

An increase of 10 percent in the price of rice, and

also in income, were postulated, and the change in calorie

and protein intake were estimated.

An analysis of the effect of a change in income and

prices on the consumption and import goals of the

Secretariat of State for Agriculture (SEA) crop programs

was undertaken for rice, red beans, and cassava. The

conclusions were that, in the three cases, the projections

of the SEA's operative plan could be improved by making use

of price and income elasticities.









vii













CHAPTER I

INTRODUCTION



Problem Setting and Objectives

One of the main objectives of the governmental

development plans in the Dominican Republic is to improve

the nutritional status of the lower income groups of the

population (Secretaria de Estado de Agricultura,SEA, 1976;

SEA, 1979; Oficina Nacional de Planificacion,ONAPLAN,

198Ua).

To make explicit the importance that the government

has placed on this objective, and the means to achieve it,

ONAPLAN published a detailed study on the nutritional

situation of the country (ONAPLAN, 1978) and later presented

a paper giving basic suggestions for a national policy on

food and nutrition (ONAPLAN, 1980b). This paper led to

discussions and to the current formulation of food and

nutrition plans.

Some of the proposed instruments to cope with the

nutrition problem are increased income per capita, the

redistribution of income, and increased food production.

Along with these instruments, the plans being formulated

also call for greater self sufficiency (fewer imports) of

basic food crops.


1




2




Some measures have been taken to increase the income of

the low-income groups. The minimum wage for agricultural

workers was increased from $2.50 for 8 hours of work to

$3.50 (Law 25 of 1979, Quezada 1981, p. 39). In other

sectors of the economy, minimum wages were raised to $125.00

per month (Law 45 of May 25, 1979). Also, an increase of 10

percent was established in the wages and salaries of all the

workers who were earning $300.00 or less. Income taxes were

also modified to favor the low-income groups (Law 39, May,

1979, ONAPLAN 198Ua, p. 32).

Regarding production goals, the SEA's Operative Plan

for 1982 contains specific production and consumption goals

for 22 products along with a package of projects and

programs to reach these production goals.

ONAPLAN has pointed out the need to estimate the

effect of changes in income among income groups on food

consumption and the impact of increased income on the

nutritional situation of each income group. This agency

also called for the study of the effect of food policy on

production and trade (ONAPLAN 1980b, p. 13).

Some work has been done to estimate income

elasticities (ONAPLAN 198Ub). The range of income was

divided into four income strata for 22 food commodities.

Using data from the 1969-70 Central Bank survey on household

income and expenditure, the change in expenditures on food




3





between two income groups was used to calculate a gross

income elasticity. With this arc elasticity and three

postulated income distributions, ONAPLAN projected the

consumption of food products and compared consumption with

the production goals of food crops from SEA's Operative Plan

to analize the need for food imports (ONAPLAN 1980b, p.

38-39).

A more sophisticated measure of the effect of income

on food expenditure is needed to improve projections for

policy purposes. The effects of prices on the food

consumption and family budgets were not measured. Estimates

of the direct and cross price elasticities are needed to

study the effect of the change in the price of one product

on the consumption of that product and on the other products

being considered.

Thus, estimates of income and price elasticities are

necessary to analyze the effect of income and price changes

on the consumption of food commodities and, through the

nutrient characteristics of each commodity, to calculate the

impact on the nutrition levels of different income groups.

It is also possible to use such estimates to analyze the

effect of government policies aimed at redistributing or

increasing incomes on the food price level and food imports.





4





The general objective of this study was to analyze the

effectiveness and consistency of government food policies

concerning income and price changes, and basic food imports.

The study was partial in the sense that it focused

mainly on the demand side. The impact on food imports

(production deficits) was based on various assumptions about

the effect of domestic policies on the supply side.

Specific objectives of the study were as follows:

1--to estimate a full matrix of demand elasticities of

disaggregated food commodities from cross sectional data,

2--to estimate expenditure elasticities of selected

food commodities using a new method that provides more

flexible coetficients,

3--to estimate the effect of household size on

consumption of food in the Dominican Republic,

4--to estimate the potential effects of changes in

prices and income on the nutrition level of the population

of the Dominican Republic, and

5--to analyze food and nutritional policies with the

information generated and to determine the consistency or

inconsistency of the food policy instruments proposed.





5





Characteristics of the Dominican Economy

The Dominican Republic occupies the eastern two-thirds

of the Island of Santo Domingo (Figure 1.1). It encompasses

an area of 18,900 square miles and had an estimated

population of 5.4 millions in 1979, a density of about 2b6

people per square mile (ONAPLAN, 1980a, p. 1). It has a

1,000-mile coast line, and a 193-mile common border with

Haiti.

Its relatively high population density (compared with

other Latin American countries) is of concern since only 44

percent of the total area is agricultural land (22.4 percent

crop land and 21.6 percent pastures). Forest land and

reserves comprise 54.4 percent and the rest consists of

lakes and- non-usable land (ONAPLAN, 1980a, p. 1).

Even though productivity is low, the agricultural

sector has played, and must continue to play, an important

role in the economic development of the country.

In 1970, 55.5 percent of the labor force was engaged

in agriculture; its average productivity (value added per

person employed) was only $506. In comparison, 7.6 percent

of the labor force was engaged in manufacturing and had an

average productivity of $2,646 (Table 1.1).

The shares for agriculture, manufacturing, and

commerce, which account for more than half of the GDP,

declined during the period 1976-79. Government,






100o 90 ... 70 s oO

300 *Gainesvill 300




N Atlantic Ocean
Gulf of Mexico An "e
6



20e orninican Republic 20c















0 200 600 Km.

o o o o \ \
100too 90 80 70 < 6O

Figure 1.1. Location of the Dominican Republic













Table 1.1. Employment, value added and value added per person employed in
the Dominican Republic, by sectors, 1970

----- -------------------------------------------------------------------

Employment Value added
Sector Number Share

----- -------------------------------------------------------------------

percent millions of percent
1962 DR $

Agriculture 594,924 55.5 310.2 23.3
Mining 927 -- 18.1 1.4
Manufacturing/Industry 81,619 7.6 216.0 16.7
Power 2,398 0.2 19.7 1.5
Construction 30,130a/, 2.8 64.2 5.0
Government 68,725a/ 6.4 137.4 10.6
Services 293,784 27.4 533.4 41.3
Total ---------------------------,072,784 100.0 --------------------,290.0 100.0

----- ------------------------------------------------------------------












Table 1.1. Continued

----- ------------------------------------------------------
Sector Value Added per
Person Employed
----- ------------------------------------------------------
1962 DR $

Agriculture 506
Mining 19,525
Manufacturing/Industry 2,646
Power 8,215
Construction 2,131
Government 1,999
Services 1,186

Total 1,203
----- ------------------------------------------------------

a/ Persons employed in public works excluded from government and
included in construction.

Source: World Bank. Dominican Republic_--It Main Economic
Development roblems. Washington, D.C., December 1978, p. 114.





9




construction, and housing showed the fastest growth rate per

year (7.5, 6.3, and 5.9 percent, respectively) since 1976

(Table 1.2). Food production has lagged behind the

population growth rate. Since most of the cultivable land

is already being used, increased agricultural output will

have to come from improved yields (World Bank, 1978, p. 14).

Real income per capita in the Dominican Republic

decreased from $489 in 1976 to $481 in 1979 (Banco Central,

Dec. 1981, p. 199), and income distribution in 1976-77

showed that 50 percent of families with the lowest income

receive only 18.5 percent of the total income. Income

distributions for the country and for the urban and rural

areas are depicted by the Lorenz curves presented in Figure

1.2.

In 1976-77, 50 percent of the population had a per

capita income below the poverty level, and the percentage

was higher in the rural areas than in the cities (Quezada,

1981, p. 3).

The decrease in income per capita along with the

uneven distribution of income has worsened the nutritional

conaition of the population. Consumption of calories per

person decreased 9.2 percent from 1973 to 1977, and overall

about 75 percent of the population consumed less food than

is needed to fulfill the minimum nutritional requirements

(ONAPLAN, 198Ub, p. 2).





10







Table 1.2. Gross domestic product in the Dominican Republic
in 1976 and 1979, the change between these two years, and
the equivalent compound rate of growth by sectors


Value Shares
Sector
1976 1979 1976 1979

millions of 1970 DR $ Percent

Agriculture 429.2 461.7 17.6 16.8
Mining 146.7 146.5 6.0 5.3
Manufacturing 457.4 504.8 18.6 1d.4
Construction 153.2 184.0 6.3 6.7
Commerce 414.0 455.5 16.9 16.6
Transport 166.7 195.8 6.8 7.1
Housing 156.8 186.0 6.4 6.8
Government 189.9 236.1 7.8 8.6
Other / 329.0 375.7 13.5 13.7

Total 2,442.9 2,746.1 100.0 100.0













Table 1.2. Continued


Growth 1976-1979

Sector RD$ millions Percent of Rate of Compound Rate
of 1970 total growth growth of growth per
annum (percent)


Agriculture 32.5 10.7 7.6 2.5
Mining -0.2 -- --
Manufacturing 47.4 15.6 10.4 3.3
Construction 30.8 10.2 20.1 6.3
Commerce 41.5 13.7 10.0 3.3
Transport 29.1 9.1 17.5 5.5
Housing 29.2 9.6 18.6 5.9
Government 46.2 15.2 24.3 7.5
Other-/ 46.7 15.4 14.2 4.5

Total 303.2 100.0 12.4 4.0
----------------------------------------------------------

S/ Includes electricity, finance, communications and other services.

Source: Banco Central de la Republica Dominicana. Boletin Mensual,
Dec. 1981. Santo Domingo.





12


















I 7
o












20 C
i /






















o# Urban
S* Rural


0 10 20 30 '40 50 50 7 0
P CNT C '. ,, -6 -


Figure 1.2. Family income distribution, Dominican Republic,
1976-77





13





The pattern of development of the Dominican economy is

characterized by the dominant role of the foreign sector.

Agricultural products and some minerals are exported, and

manufactures, oil, and capital goods are imported. In 1979,

the value of exports represented 16 percent and imports 21

percent of the GDP (Banco Central, Dec. 1981).

Although the share declined from 91.2 percent in 1966

to 70.7 percent in 1976, agricultural exports still

represent the bulk of Dominican exports. The share lost by

agriculture has been gained by mineral products (Table 1.3).

In 198u, for example, the four leading minerals (bauxite,

ferronickel, gold, and silver) accounted for 39.4 percent of

the total exports, and the four leading agricultural exports

(sugar, coffee, cocoa, and tobacco and their by-products)

accounted for more than 50 percent (Banco Central, Dec.

1981, p. 129).

Intermediate goods have been dominant among imports

(29.8 percent in 1974), followed by consumer goods (26.8

percent) and capital goods (23.9 percent). Fuels and

lubricants showed the fastest growth rate among all imports

--a 68.3 percent average annual growth from 1970 to 1974

(Table 1.4)

It is interesting to observe the increase of the

imports of foodstuffs and crude oil from 1973 to 1977. The

increase in crude oil imports was estimated at $145.68










Table 1.3. Proportion of exports from the Dominican Republic by commodity, 1966-7b

---- ----------------------------------------------------------------------------

1966 1967 1968 1969 1970 1971
---- ----------------------------------------------------------------------------
percent
Agricultural Products (Processed and
Unprocessed) 91.1 90.2 89.4 89.0 89.6 90.1
Sugar and By-products 58.9 60.3 56.1 53.7 54.3 60.2
Raw Sugar (51.4) (52.4) (50.7) (47.9) (48.5) (54.8)
Furfural ( 2.7) ( 2.6) ( 1.8) ( 2.5) ( 2.4) ( 2.7)
Molasses ( 2.9) ( 3.5) ( 2.9) ( 2.6) ( 3.2) ( 2.7)
Other ( 1.8) ( 1.8) ( 0.7) ( 0.7) ( 0.2) ( --)
Cocoa and By-products 8.1 7.7 8.5 10.9 9.2 5.4
Cocoa Beans ( 7.9) ( 7.4) ( 8.3) (10.7) ( 9.0) ( 5.2)
Chocolate ( 0.2) ( 0.3) ( 0.2) ( 0.2) ( 0.2) ( 0.2)
Other ( -) ( -) ( -) ( _-) ( -) ( _) -
Coffee 15.4 10.9 10.9 11.6 13.5 9.9
Tobacco and By-products 4.8 6.7 6.9 6.9 6.7 8.6
Bananas 0.6 -- 0.1 0.1 0.1 --
Meat -- 0.2 2.2 2.2 1.6 1.2
Other 3.4 4.4 4.7 3.6 4.2 4.8
Mineral Products (Processed and
Unprocessed) 8.0 8.6 8.4 8.5 7.6 7.3
Bauxite 7.5 8.2 7.7 7.9 7.1 6.6
Ferronickel -- -- -- -- -- 0.2
Other 0.5 0.4 0.7 0.6 0.5 0.5

Manufactures, Re-exports and Other 0.8 1.2 2.2 2.5 2.8 2.6
Total Commodity Exports 100.0 100.0 100.0 100.0 100.0 100.0
---- ----------------------------------------------------------------------------








Table 1.3. Continued



1972 1973 1974 1975 1976

percent
Agricultural Products (Processed and
Unprocessed) 78.1 75.3 80.0 81.2 70.7
Sugar and By-products 50.7 46.6 54.7 66.5 40.5
Raw Sugar (45.9) (42.3) (50.9) (62.7) (35.5)
Furfural ( 2.5) ( 1.8) ( 1.3) ( 2.0) ( 2.9)
Molasses ( 2.1) ( 2.3) ( 2.0) ( 1.6) ( 1.8)
Other ( 0.2) ( 0.2) ( 0.5) ( 0.2) ( 0.3)
Cocoa and By-products 5.3 5.5 7.5 3.2 7.0
Cocoa Beans ( 4.6) ( 4.4) ( 6.9) ( 2.8) ( 6.3)
Chocolate ( 0.1) ( 0.1) ( 0.2) ( 0.1) ( 0.3)
Other ( 0.6) ( 1.0) ( 0.4) ( 0.3) ( 0.4)
Coffee 8.6 10.5 7.2 4.8 14.1
Tobacco and By-products 8.3 6.9 6.2 4.0 5.6
Bananas 0.3 0.3 0.2 0.2 0.3
Meat 1.9 2.3 1.4 0.5 1.1
Other 3.0 3.2 2.8 2.0 2.1
Mineral Products (Processed and
Unprocessed) 17.9 22.4 17.6 16.4 25.5
Bauxite 4.3 3.3 2.8 1.9 2.2
Ferronickel 13.5 18.9 14.6 11.4 15.5
Other 0.1 0.2 0.2 3.1 7.8
Manufactures, Re-exports and Other 4.0 2.3 2.4 2.4 3.8
Total Commodity Exports 100.0 100.0 100.0 100.0 100.0


Source: World Bank. Dominican Republic--Its Main Economic Development Problems.
Washington, D.C., 1978.




16








Table 1.4. Imports into the Dominican Republic by major categories,
1967-74


1967 1968 1969 1970 1971


millions of US $
Total Import Goodsa/ 171.3 200.8 221.9 295.0 340.1
Consumer Goods 48.6 62.8 59.8 76.9 88.7
Food 31.5 44.8 38.5 44.3 53.6
Other 17.1 18.0 21.3 32.6 35.1
Fuels and Lubricants 12.8 13.8 18.1 19.3 24.8
Intermediate Goods 59.8 71.3 77.0 104.6 113.0
Capital Goods 50.1 52.9 67.0 94.2 113.6

percent of total import goods

Total Import Goods 100.0 100.0 100.0 100.0 100.0
Consumer Goods 28.4 31.3 27.0 26.1 26.1
Food 18.4 22.3 17.4 15.0 15.8
Other 10.0 9.0 9.6 11.1 10.3
Fuels and Lubricants 7.5 6.9 8.1 6.5 7.3
Intermediate Goods 34.9 35.5 34.7 35.5 33.2
Capital Goods 29.2 26.3 30.2 31.9 33.4





17






Table 1.4. Continued



1972 1973 1974 Average Annual
Growth 70-74 %

millions of US $
Total Import Goods-/ 382.0 448.5 796.4 28.2
Consumer Goods 104.3 124.9 213.6 29.1
Food 59.2 74.2 133.7 31.8
Other 45.1 50.7 79.9 25.1
Fuels and Lubricants 46.7 48.5 155.0 68.3
Intermediate Goods 120.3 145.9 237.5 22.8
Capital Goods 110.7 129.2 190.3 19.2

percent of total import goods

Total Import Goods 100.0 100.0 100.0
Consumer Goods 27.3 27.9 26.8
Food 15.5 16.6 16.8
Other 11.8 11.3 10.0
Fuels and Lubricants 12.2 10.8 19.5
Intermediate Goods 31.5 32.5 29.8
Capital Goods 29.0 28.8 23.9


a/ Based on UN Trade Publications. Includes OECD countries and
Venezuela. This source is adjusted for petroleum, transport vehicles
and wood. These totals will not necessarily coincide with total
imports used in balance of payments which are based on Direction of
Trade imports from trading partners.

Source: World Bank. Dominican Republic-Its Main Economic Development
Problems. Washington D.C., 1978.





18




millions. For the same period, the increase in foodstuff

imported was estimated at $40.43 millions (World Bank, 1978,

p. 146). These two items added considerably to the trade

deficit of $557.8 millions in 1980--10 percent of the GDP

(Banco Central, Dec. 1981, p. 129).

Between 1978 and 1980 prices were inflationary. The

consumer price index increased 7.1 percent in 1978, 9.2

percent in 1979, and 16.7 percent in 1980 (Banco Central,

Dec. 1981, p. 155). After a jump in the rate of inflation

between 1979-80, prices were still increasing in 1981 but at

a more moderate rate. During the 12-month period from

November 1980 to November 1981 the consumer price index

increased 6.3 percent and this slower rate continued through

1982.

The impact of the increase in prices has been

different in the rural and urban areas. In 1980, the urban

price index increased by 18.4 percent, that is, 1.6 percent

more than the national increase. The rural price index

increased by 15.4 (Banco Central, Dec. 1981, p. 155).

Looking at the increase in the consumer price index by

income strata, it is observed that in 1980 the index for the

$300 to $400 income stratum showed the smallest increment

(13.8 percent). The index for all other strata (above and

below) grew relatively equally.





19




The change in the comsumer price index by groups of

commodities was 15.4 percent for food, alcohol, and tobacco;

10.2 percent for housing; 20.4 percent for clothing; and

31.8 percent for others (Banco Central, Dec. 1981, p. 157).



Plan of Dissertation

Chapter II is used to present the traditional uses of

household budget surveys and their applications to

estimating Engel curves. This chapter is also used to

review the literature within the theoretical framework

applied in this study. This review included consumer theory

and demand analysis, and applications of cross-sectional

data in demand estimations.

The methodology applied in this study is described in

Chapter III, including the use of the restrictions of

consumer theory, the models applied, and the estimation

procedure. Steps followed to calculate the elasticities

matrix are also explained in this chapter. In the last

part, the method used to calculate the effect of prices an

income changes on calorie and protein intake is discussed.

The coefficients estimated are discussed in Chapter IV

along with the presentation and discussion of the full

matrix of demand elasticities and family-size elasticities.

A final comment is made about the results of the new method

of estimating Engel elasticities.





20




The implications for policy are discussed in Chapter

V, showing the effects of changes in prices and income on

the calorie and protein intake. The last part of this

chapter presents the implications of changes in income and

prices for consumption and import goals.

Finally, Chapter VI gives the summary and conclusions.

Limitations of this work are identified and some suggestions

for further research are offered.

















CHAPTER II

THEORETICAL FRAMEWORK AND REVIEW OF LITERATURE



In applied demand analysis there has been concern

about the theorical plausibility of empirical demand

estimates. One way to proceed in order to ensure the

theoretical validity of the estimated demand model is to

specify a utility function. The demand functions to be

estimated can then be derived from the first order

conditions obtained when maximizing the utility function

subject to the budget constraint of the consumer. A

disadvantage of this approach is the requirement of

specifying well-behaved utility functions and the derived

demand equations which result may be too complicated to be

estimated due to the number of parameters or to

nonlinearities. The other approach is more pragmatic and

leads to the direct specification of demand functions with

causal variables that are believed to capture the behavior

of the consumer.

In reviewing the literature, an attempt was made to

fill the gap between the theory of consumer behavior and

empirical demand analysis. Emphasis was upon those studies



21





22




using cross sectional data to estimate Engel curves and

income and price elasticities of demand.



Household Budget Surveys

Household budget surveys have been widely used to

estimate income-consumption relationships and some attempts

have been made to estimate price elasticities of demand.

This kind of survey provides data indicating all

expenditures on consumer goods and services made by

individual families. A basic objective of the budget survey

is to obtain the percentage budget share of all consumption

items to be used as weights in the official cost of living

index (Phlips 1974, p. 101-102). These cross-sectional data

can be used to estimate Engel curves, which express the

expenditure on a good as a function of income, i.e.,

Ei = piqi = fi (y) ; i=1,...,n

where p represents price of good i; q represents quantity

of good i; y represents income and n is number of goods.

The fact that prices are (almost) constant is used to

isolate the influence of income and the simultaneous

variation of both prices and income encountered in time

series analysis is avoided. However a cross section does

not automatically satisfy the assumption of a single given

utility function, because the preference ordering may change

from family to family (Phlips 1974, p. 103). To overcome





23





this problem it is necessary to take a collection of

families that have different incomes which are as homogeneus

as possible with respect to such things as social class and

profession.

Phlips adds that to ensure the theoretical

plausibility of estimated Engel curves the general

restrictions of demand theory must be satisfied. All of the

price derivative restrictions of demand theory (i.e.,

homogeneity, symmetry, negativity of the own substitution

effect) disappear because prices are constant (Phlips 1974,

p. 105). The only restriction that remains is the adding-up

condition, i.e., that



Efi(y) = y

or similarly

2 (dEi/dy) = 1 ,

the sum of the marginal propensities to consume (or the

marginal budget shares) has to be equal to one at all income

levels.



Engel Curves

The first attempt to analyze cross-sectional data

based on a theoretical model was the work by Allen and

Bowley (1935). They estimated linear equations which

automatically satisfy the adding-up restriction but were





24





very restrictive in term of the underlying utility function

that was implied. In addition, the implications of

linearity in terms of the income elasticities were

unsatisfactory from an economic point of view (luxury goods

resulted with negative dE./dy), and the statistical fit was

very poor (Phlips, 1974, p. 109).

Pollak (1971) thoroughly analyzed the implications of

linear Engel curves. He defined a demand system by



{h (p,u), ... h (p,u)}

where p represents the price vector (p, ..., p n) and


n
u = Ei= y
i =1


He characterized the demand functions in the following way:

if the demand functions are of the form



h (p,u) = xi(p) + z (p)u



in some region of the price-income space, the demand

functions are locally linear in income; if



h (p,u) =zi(p)u





25







the demand functions exhibit expenditure proportionality,

this is a especial case of linearity.

The demand functions are linear in income if and only

if the marginal budget shares of the ith good



mi (p,u) = (Dpihi(p,u))/ 3u



are functions of prices alone, independent of income. The

income-consumption curves are straight lines in a region of

the commodity space if and only if the demand functions are

linear; they are straight lines radiating from the origin if

and only if the demand functions exhibit expenditure

proportionality.

Pollak proved a theorem stating that additive utility

functions yield demand functions which are locally linear in

income if and only if they are of the form


n
U = k ak log( qk bk ), a >0, (qi b ) > 0, I ak =1 ;
k=

n c
U = ak ( Bk 6k k )
k=l
or
n -B
U ate k qk ak Bk> 0.
k-k





26





He pointed out that in demand analysis the assumption

that all income elasticities are unity has little merit and

that the additivity assumption is defensible if the

arguments of the utility function are taken to be broad

aggregates of goods rather than individual commodities.

Prais and Houthakker (1955) analyzed non-linear

functions to estimate Engel curves more adapted to income

consumption patterns. In fitting income-consumption

equations they took into account a priori economic as well

as statistical considerations. They incorporated two

properties in the algebraic formulations: a) that there is

an initial income below which a commodity is not purchased,

and b) that there is a satiety level, that is, a maximum to

the quantity of the commodity consumed which is not exceeded

however high income may rise. Based on these considerations

they chose the following models:

log Ei = ai + bilog y; double-logarithmic,



log Ei = ai (bi /y); log reciprocal



Ei = ai + bilog y; semi-logarithmic



Ei = ai (bi/y); hyperbolic.





27





Prais and Houthakker concluded that none of the

functions were uniformly satisfactory. The double-

logarithmic form gave fairly satisfactory results for most

commodities except for the difficulty of treating zero

expenditures. The semi-logarithmic form gave a satisfactory

description of expenditures on most foodstuffs, and all

non-linear forms proved to give better estimates than that

given by the linear form.

With respect to the work of Prais and Houthakker,

Phlips pointed out that, while much has been gained with

this pragmatic approach in terms of descriptive power, much

has been lost in terms of theoretical plausibility.

Non-linear specifications are not compatible with utility

maximization, as they do not satisfy the adding-up

criterion. He makes the same criticism of the work of

Aitchison and Brown (1957), Wold and Jureen (1953) and

Champernowne (1969).

Leser (1963) investigated the properties of various

forms of Engel functions satisfying the additivity criterion

using nine commodity groups. He took into consideration

several factors in choosing among various functional forms.

These factors included the close connection with a direct or

indirect utility function, the validity of the function for

all positive values of total outlay within a wide range, and

the plausibility of the income elasticities besides the





28




statistical goodness of fit. He concluded that the

equations used by Working (1943) of the form

wi = ai + bi log y + ei

where wi = Ei/y describe a behavior which would seem

acceptable, and that the goodness of fit criterion tends to

reinforce the theoretical considerations in favor of this

model.

The above model was rejected by Prais (1953). In that

work he comments that Stuvel and James in their study of

households expenditure on food in Holland showed that

neither the linear nor the exponential forms were

appropriate for the whole range of budgets in that

collection. In his conclusion, Prais suggests that a

typical Engel curve may be described by the equation



( Ei/N ) = a + b log ( y/N )



where N is household size in equivalent adults. The above

equation means that the income elasticity is inversely

proportional to the expenditure on the commodity. To

specify the behavior of the Engel curve for quantities he

takes into account the change in quality measured by the

average price per physical unit of the commodity and

describes it by the equation

p = h + k log ( y/N )





29






where p is average price paid per unit of commodity.

Dividing expenditures by price he obtained the quantity

consumed



q a + b log (y/N)

N h + k log (y/N)

i.e.

q B + log (y/N)

n C + log (y/N)



where A, B, and C are constants respectively equal to b/k,

a/b, and h/k (Prais, 195j, p. 101-102).

Prais pointed out that this equation for the Engel

curve for quantities seems very satisfactory, giving an

asymptote equal to A, and consumption falling to zero at a

positive income.



An Alternative Method of Estimating Engel Curves

A new method of estimating Engel elasticities was

proposed by Kakwani (1978). The new equation of the Engel

function was based on a coordinate system for the Lorenz

curve developed by Kakwani and Podder (1976). The equation

of the Lorenz curve was fitted to the expenditure data and

then the expenditure elasticities were derived in terms of





30





the estimated parameters of the Lorenz curve for total

expenditure and expenditure on each commodity.

The income elasticities derived from the Engel

functions investigated above were either constant or

monotonically increasing or decreasing. The actual data

indicated that the elasticities may not be either constant

or monotonic (Kakwani, 1978, p. 103).

To estimate the Lorenz curve, Kakwani and Podder

(1976) defined x as per capita total expenditure or income

of a famiy; x is a random variable with mean M and the

probability distribution function F(x). If F(x) is the

proportion of families having income less than or equal to

x, then



F (x) = (1/M) fxx g(x) dx
1 o


is the first moment distribution function, where g(x) is the

density function of x. The Lorenz curve is the relationship

between F(x) and F (x). This relationship is shown in

Figure 2.1. Let P be any point on the curve with

coordinates F and F Then, by introducing a new

coordinate system (see Appendix 3),



S= (F + FI)/V2- and n = (F F )/V2-,







31

































F


Figure 2.1. Kakwani and Podder coordinates for estimating
a Lorenz curve





32




the equation of the Lorenz curve in terms of n and a can be

written as n= f( n ). The derivatives of n with respect to -r

are



f'( 7 ) = (M x)/(M + x) (1)

and

f' ( r ) = -(2 V 2 M2)/g(x)(M+x)3 < 0 (2)

The model fitted was

n = an d (V2 n )B; a > 0, a > 0, and B > 0.

Kakwani (1978, p. 104) denoted v.(x) as the Engel function
1
of the ith commodity which has the mean per capita

expenditure M.; then the function
1


F1[v.(x)] = (1/M.) fx v.(x) g(x) dx
11 0 1


is interpreted as the proportion of per capita expenditure

on the ith commodity of the families having a total per

capita expenditure less than or equal to x. The

concentration curve of the ith commodity is given by the

relationship F(x) and F [vi(x)] (Kakwani 1978, p. 104; 1977,

p. 720). In terms of n. and r the result is
1 i


i = (1/2)[F(x) + F (v (x))]



n. = (1/2)[F(x) F (v.(x))] .
1 i i





33






The concentration curve can be written as n. =f.( ), and
1 1 1
the derivatives of ni with respect to Ti are

Mi vi(x) (3) ,
M. + v.(x)


2
2 V2 M. v' .(x)
1 3
-g(x)[Mi+ vi(x)] (4)

Combining (1), (2), (3), and (4) the Engel elasticity of the

ith commodity is obtained as



avi(x) x v.(x)x f ( i)[l + f' ( )]2 [I f' (T)

3x vi(x) vi(x) f" ( )[1 + f. (T )]2 [1 f (-T ) (5)



Kakwani (1978, p. 105) pointed out that if the data on

different commodities are available in grouped form arranged

according to total per capita expenditure or income, this

equation can be used to compute expenditure elasticities at

the end point of each expenditure class. To compute

elasticities at any income level the following relationship

can be used:

T + f(?) = 7. + f.(T .) .

The function fitted was the same assumed by Kakwani and

Podder (1976)





34






n. = a i (V7 )Bi
1 1 1 1
and was estimated using ordinary least squares method.

Blaylock and Smallwood (1982) applied the method

discussed above to estimate expenditure elasticities among

three urbanizations for several food groups. They

introduced a generalized functional form to describe the

Lorenz and concentration curves. The model developed was



n A*+ a ) + B*(V V) + u



where x s A* and B* are parameters and u is the error

term. This equation approaches the function used by Kakwani

and Podder as all transformation parameters approach zero.

The equation was estimated using maximum likelihood

technique since it is nonlinear in the parameters. Blaylock

and Smallwood concluded that, as income increases,

elasticity patterns can be nonmonotonic and the pattern

found in their analysis could not be obtained using

traditional methods of analysis. This added flexibility for

estimating Engel parameters is important for policy analysis

involving the impact of income transfer programs on

population subgroups located at extremes of the income

distribution (Blaylock and Smallwood, 1982, p. 139).





35





Application of the Fri Method

In an effort to compute all direct and cross price

elasticities, Frisch (1959) made the assumption of want

independence and suggested a method of obtaining the price

elasticities with knowledge of the budget shares, the income

elasticities, and the flexibility of the marginal utility of

money. In his work, he defined marginal utilities as a

function of the quantities

ui = ui(qj, ..., qn) ; i = 1, 2, ..., n

and the inverse function as

qi = qi(ul, ..., un) ; i = 1, 2, ..., n;

he also defined




ik ui (q ,qn) k ; i=1, 2, ..., n
Ui -j ~----------
aq u .(q ...,q ) ; i=1, 2, ..., n
k i 1 n


as the utility acceleration which is the elasticity of the

marginal utility with respect to qk And he defined



aq (ul, ..., u ) u
1 n k
i- a
ikk q (u ..., u) ik
k1 1' n


as the want elasticity. This elasticity tells the

proportionate change in the want for a good with respect to

the change in quantity of other goods. Frisch defined good




36





i as want independent of all other goods if qik = 0 for all

kfi, or equivalently q ki = 0 for all k i i. Another

equivalent definition he gave was that good i is want

independent of all other goods if uik = 0 or u ki = 0 for all

k ii; uik = 0 means that the marginal utility of good i

depends only on the quantity of good i and not on any other

quantity. These concepts can be applied to broad

aggregates; want independence will not hold with single

commodities.

The marginal utility of money is WU/3y. From the

first order conditions of constrained utility maximization

X = (u.)/(Pi)

and the elasticity of the marginal utility of money with

respect to income is

( X / y)(y/) ) =

and is called the flexibility of the marginal utility of

money or money flexibility.

Using these results Frish developed the relationship

for the price elasticity as



eij = i wjeiy (1/ )wj e



where eijis the cross price elasticity of good i with

respect to the price of good j; wj is the budget share of

good j; eiy is the income elasticity of good i; and 4 and a .
11





37





are as defined above. This relationship is equivalent to

the classical Slutsky relation

e.. = e..(w./w.) + w.(e. e. )
ij 31 J 1 J Jy ly
and

e. = Ya.. ; i,j = 1,2, ...,-n
ly 1j
which can be compared with the classical relation that

states the restriction that

Seij = -e (homogeneity restriction) .
J
Now if a good i is want independent of all other goods,o =

0 for all i i j and the cross price and income elasticities

can be simplified to the relation



eij = jeiy (/ jejeiy

eij = eiy w [1 + (e. / )] (cross price

elasticity)

ely = ii (income elasticity)

so

i.i = (eiy/ )

thus

eii = w .e (1/4 )wi.e. e

eii = (eiy/ P ) wieiy (1/ )wieiyeiy

eii = wieiy + eiy[(-wiei )/ ]

eii = eiy[Wi (1-wieiy)/ y )]

from which can be obtained, i. e.

eii+ 4eiyw = ey weiyey




38





d (e.. +e. w.) = e. w.e. e.
11 ly 1 ly 1 Ty Ty


e. w.e. e.
iy 1 iy iy
eii+ eiywi

(George and King, 1971, p. 22-23).

The disadvantages of using this approach are that want

independence imposes complete additivity in the utility

function (which implies that the marginal utilities of i and

j are unaffected), and that the ..i are not invariant for

any monotonic transformation of the utility function.

Frisch pointed out that his development can be

extended to the case where there are groups of commodities

in which there may be dependence within each group, but want

independence between a good in a given group and all the

goods in the other groups. In this case, in addition to the

budget share, wi, the income elasticity, ey and the money

flexibility, p the cross price elasticities for the

commodities within the groups are needed.

George and King (1971) made a full application of this

method, estimating a demand system for food commodities in

the United States. They used most of the theoretical

restrictions on demand systems and made full application of

the Frish relations developed above.

In order to apply the Frish method, they imposed the

assumption of want independence for all commodities outside




39





a given group to obtain all the cross price elasticities of

commodities between different groups. The direct and cross

price elasticities of commodities within each group were

estimated directly using time-series data.

They estimated a demand system for 49 commodities

grouped in 15 separable groups. The demand equations were

specified with quantities as dependent variables. The

independent variables included income, the price of all

commodities belonging to the same group, and the price

indices of other groups.

To obtain the income elasticities from cross-sectional

data, they used both ordinary and weighted regressions. The

regression equation obtained was

log q = a + b log y

where q is per capita .quantity consumed and y is per capita

income. They also evaluated the effect of household size on

consumption in the cross-sectional analysis. Expenditure

and quantity were used as dependent variables. The

regression equations were of the form

log q = a + b log y + c log N

and

log E = a' + b' log y + c' log N

where E represents expenditures, and N household size. The

quality elasticity was also analyzed, with and without

household size. This measure was obtained from the





40




decomposition of the expenditure elasticity into quantity

and quality elasticities. Combining the coefficients

calculated in the ways mentioned above with the restrictions

of the theory of consumer behavior, George and King

calculated all the remaining coefficients.



Price Elasticities From Cross-Sectional Data

Several attempts have been made to estimate price

elasticities from household budgetary data. The approaches

to this problem can be classified as three kinds. First,

estimates of direct and cross price elasticities have been

obtained making strong assumptions about the underlying

utility function. The second approach has been based on

pooling cross-sectional and time series data to overcome the

restriction on the number of observations that usually

exists when using only time series. The third kind is aimed

at capturing the price variations during the time period

covered by the household budget surveys, which in most cases

requires enough time to observe some price variation.

Friedman (1935) discussed a method suggested by Pigou,

as early as 1910, of using budgetary data to obtain the

ratio of the elasticity of demand for one commodity to that

for another. Pigou's method is based upon the assumptions

that the marginal utility of a commodity to an individual is

a function solely of the quantity of that commodity




41





possessed, and that the marginal utility curve for a

commodity is the same for each individual. He then makes

the utility elasticity equal to the price elasticity.

Friedman shows that that assumption would hold only if there

is a very large number of commodities purchased or if at

least one commodity has a very large utility elasticity.

Leser (1941-42) reviewed the assumptions that have to

be made in order to estimate price elasticities from budget

data. He pointed out that the assumption made by Marshack

--that the difference between the income and price

elasticities of demand for a commodity is negligible if the

amount spent on it is small compared with total expenditure

--was rejected by Frisch who showed that the observed income

elasticities can not be interpreted as price elasticities.

In his work to obtain price elasticities, Leser assumed that

the cross elasticities of demand depended on the nature of

the good whose price changes, and not on the nature of the

good for which the effect was studied. This assumption

excluded a high degree of substitution or complementarity,

and thus would be valid only for broad aggregates and not

for single commodities.

Hassan and Johnson (1976) estimated consumption

functions and Engel curves based on cross-sectional data

from a family budget in Canada. They also examined

hypotheses of possible structural change related to six





42






socioeconomic characteristics. These included city or urban

community in which the family resided, class of tenure, age

of family head, family life cycle, family income, and

education of family head. In this study the Engel curve was

approximated with a linear function. To justify this

functional form they pointed out that the aggregate level of

the commodity groups should not involve a substantive

misspecification error; that is, the commodities as they

were grouped were neither luxuries nor inferior goods.

In a new study using cross-sectional data from an

urban family food expenditure survey in Canada, Hassan and

Johnson (1977) estimated demand parameters for 122 food

items and computed, from these estimates, elasticities for

expenditure, income, quality, family size, and price. The

price elasticities were computed by using implicit prices

derived from the expenditure and quantity data. This time

they fitted a semi-logarithmic function to approximate the

Engel functions, which implied that elasticities vary

inversely with the level of consumption. Among the reasons

they gave for choosing this functional form were that the

semi-logarithmic function gives a satisfactory description

of expenditures for most food commodities; that this form

allowed observations of zero for the dependent variable

(which occurs frequently when microdata are used); and that

even though this function does not satisfy the adding-up





43





criterion, it provides an appropriate representation of

Engel curves for the range in which the income elasticity is

less than one.

To estimate the price elasticities, Hassan and Johnson

(1977) used the price variations within the period over

which the survey data were collected. In the survey which

provided the source of their data the period was one year.

They pointed out that during such a period, price variations

are likely to be small and statistical measurment cannot be

overly ambitious. So the estimated models must include a

large amount of prior information. They included only own

price elasticities in the demand functions.

They concluded that the price variations that occurred

within the sample period were sufficient for estimating the

coefficient necessary to compute the elasticities, and that

the estimated parameters were consistent with theory and

with available elasticity estimates from time series data.



Demand Estimates in the Dominican Republic

A few attempts have been made in the Dominican

Republic to estimate income and price elasticities. All of

them faced the problem of inadequate data; time series data

did not exist at all for some crops or food commodities and

when they existed they were very inconsistent (USDA, 1977,

p. 103). Cross-sectional data first became available in




44




1971, from a household income and expenditure survey (Banco

Central, 1971), but only for the city of Santo Domingo.

Later, in 1976-77 the Central Bank conducted a new household

income and expenditure survey (this time at a national

level), whose final results were available in May 1982.

Erichson, House, and Nunez estimated income and price

elasticities for several crops (USDA, 1971). In their work,

they rejected the use of time series because of the lack of

meaningful data which led to very poor results. They used

the Frish method (discussed earlier) to estimate price

elasticities. Estimates of expenditure elasticities and

budget shares were derived from the 1969 household income

and expenditure survey. The money flexibility coefficient

was taken from cross-country studies done by de Janvry,

Bieri and Nunez (1972), Lluch and Williams (1975), and Le-Si

and Pomareda (1976). In this way they calculated direct

price elasticities for nine crops.

Another study using data from the 1969 household

survey was done by ONAPLAN to estimate gross income

elasticities between two income groups (see discussion on

pages 2 and 3).

In a preliminary study using data from the 1976-77

household survey, Del Rosario and Musgrove (1980) estimated

consumption functions for 13 groups comprised of 67 single

commodities. They worked with data grouped by 10 income




45




classes. The expenditure on each commodity, Ei, and the

quantity of each commodity, qj, were regressed separately

against total expenditure, y. The models fitted were as

follows:

log qi = ai+ bi/y + ci log y and

log Ei = a' + bi/y + cilog y

where a' = ai + log pi .

The authors suggested extension of their work by

including more explanatory variables such as family size and

geography location. They also suggested that this kind of

estimation be done with single household data (Del Rosario

and Musgrove, 1980, p. 28).

Quezada (1981, p. 162, 219) estimated demand equations

for 6 commodities using data from a 14-years time series on

prices and expenditure. He fitted a linear model deflating

all prices and consumption expenditures by the consumer

price index. In the model he included per capita

consumption of the product as a function of its own price

and of prices of some substitutes or complements, and per

capita consumption expenditures.



The Limited Dependent Variable Model

When working with cross-sectional data from a

household survey many of the observations on the consumption

of specific goods will be zero, since at the time of the





46




survey the consumer did not buy that good, so the

consumption variable is not observed over its entire range

(Tobin, 1958 p. 24). As described by Maddala (1977, p.

162), if

q = bx + e

and q is observed only if it is greater than zero, then



q = bx + e if bx + e > 0 or e > -bx or

q = 0, i.e., e = -bx otherwise.



Ordinary least squares cannot be used to estimate the

regression equations using only the observations for which

q> 0 because the residuals do not satisfy the assumption

that E(e) = 0, if only those residuals for which e > -bx

were used. Making some specific assumption about the

distribution of the error term, e, the maximum likelihood

method can be used to estimate the parameters. If, for

instance, it was assumed that e has a normal distribution,

with mean zero and variance a the joint distribution of

the observations is



L = T f((qi bx. )/ a ) F F(-bx./o)
1 1 2 J


where f((qi bxi )/a ) is the standard normal density, and

represents the observations for which qi was observed; and




47





F(-bx / c ) is the cummulative normal density, and represents

the observations at the limit, i.e. where e = -bx when the

limit is 0. The ML estimates are obtained by maximizing L

with respect to b ando. (Maddala, 1977, p. 136). Tobin

(1958) pointed out that if only the probability of limit and

nonlimit responses were to be estimated, the probit analysis

would be appropriate, but that it would be inefficient to

throw away information on the value of the dependent

variable when it is available. So he developed what he

called a hybrid of probit analysis and multiple regression,

that was later to be known as the tobit model.


















CHAPTER III

METHODOLOGY



The Restrictions of Consumer Theory

In this study, a demand system for food commodities

was estimated using the general restrictions of consumer

theory. The restrictions also serve to reduce the number of

parameters to be estimated. Some price elasticities were

estimated directly. In such cases the variation in prices

over the period in which the data were collected was used.

The general restrictions on demand coefficients

comprise the following:



se.. = -e (Homogenity Condition)
J ly
i,j = 1 ,...,n

Zw. e = 1 (Engel Aggregation)

i= 1 ,...,n

e = (w./w .)e.. w. (e. e. ) (Symmetry
ii 3 1 j1 j ly Jy
Relation)

i,j = 1 ,...,n
Ewieij = -w. (Cournot Aggregation)




48




49




i,j = 1,... ,n

where e. represents the elasticity of good i with respect
1j
to the price of good j, e. represents the income elasticity
ly
of good i, w. = p.q./Ep. q. is the budget share of good i,

and n is the number of goods.



The Demand Models

A semi-logarithmic model was hypothesized for the

demand functions. The equations were of the form



q. = a'. +a. log p +...+a. log p +b.log y +c.log m + u.



where qi represents quantity consumed by the household of

good i, for i=1,...,n; pj represents prices of good j, for

j=n-k and k=0,...,n-1; n is the number of commodities (some

are single commodities and some represent groups of

commodities); y represents the household total expenditure

(which is also referred to as income in this study); m is

the number of members in the family, adjusted for adult

equivalents; uj is the error term; and a'. = log aio.

There exists an extensive literature on the

measurement of household size. Brown and Deaton (1972, p.

1178-1186) reviewed many of the proposed scales to adjust

the size of the family for adult equivalent; they pointed

out that such scales, including the nutritional scales, are





50




based on normative judgments rather than on market behavior.

Nevertheless, for this study it was considered that from a

practical point of view it would be better to make some

adjustment in the number of members, since all of them do

not consume the same quantities of the different food

commodities. The scale used was to consider any family

member 14 years old or less as 0.5 of an adult equivalent.

Using the proposed semilog models, the estimation

procedure was as follows: the first equation had n prices as

independent variables plus total expenditure and number of

adult equivalent household members, the second equation had

n-1 prices as independent variables plus total expenditure

and number of adult equivalent household members, and so on

up to the 30th equation which had only its own price along

with total expenditure and number of members. The

estimation procedure is shown in Table 3.1.

A few exceptions were made to include the prices of

some products that were known, a priori, to have an

important effect on the respective demand equations, and

that would not be included if the procedure had been

followed. These exceptions were made for cassava, where the

price of plantain was included; for other-meats, where the

prices of beef and pork were included; for other-roots,

where the price of cassava was included; for other-milk,

where the price of fresh-milk was included; and for banana,













Table 3.1. Estimation procedure using the tobit model

------ ------------------------------------------------------------------------



q] = a',0 +a,11 lo +al,2log pp+a10 P2+...+al,29 log P29+al,30 og P30+b log y+c]log m+u]

q2= a'2,0 +a2,2log P2+...+a2,29 log P29+a2,3010og P30+b21og y+21og m+u 2







q29 =a'29,0 +a29,29 log P29+a29,30 log P30 +b29log y+c291og m+u29

30=a'30,0 +a30,30 log p30+b30log y+c30log m+u30






a'--------------------------------------------------j = log j------------------------




52




where the price of other-fruits was included. However, the

coetficients of these prices were not considered in the

calculations of the price elasticities; they were included

just to lessen specification errors for the estimated

parameters used in the calculations. (See estimated

coefficients of regressions in Apendix 2.)



Use fl ie Tobit Model

To estimate the demand equations the tobit model was

applied to the semilog forms for each single equation. The

usefulness of this estimation method is twofold; it provides

a way to use all observations including those which are

concentrated at the zero limit value; it also allows the use

of all the observations in the estimation of each equation;

this was important, since otherwise each equation would have

had a different number of observations and it would not have

been possible to establish the relationships between the

elasticities of different goods consistently with the same

data set for each equation.

To overcome the remaining problem of missing values in

the prices of some goods in each equation, the average

prices of the respective goods were used in place of the

missing values (Maddala, 1977, p. 202).

Since the purpose of the estimation of these equations

was to establish a functional form for the demand functions




53




and to obtain the respective regression coefficients, the

observations were not weighted. A similar procedure was

followed by Tobin (1958, p. 31) who stated that it may well

be that the sample gives umbiased estimates of the

parameters of the relationship, even though it gives biased

estimates of the separate frequency distributions of the

variables.

The estimations of the mean quantities consumed and of

the mean prices, as well as that of the budget shares, were

made using weighted observations.



The Elasticities Matrix

The elasticities matrix was comprised of direct and

cross price elasticities plus income elasticities for 30

food groups, the all food aggregate, and one nonfood group.

The 30 food groups and the variable names used for each of

them are listed in Figure 3.1.

Since the estimated values of the elasticities would

not be invariant to the ordering of the commodity groups

within the matrix, the groups were arrayed in descendent

order according to the budget shares. In this way the

group, or commodity with the largest share of the consumer

expenditure would be in the first row. As can be seen in

Table 3.1, the item in the first row is the one with more

coefficients estimated directly. The number of coefficients





54











RIC Rice
FAT Fat and oils
BEF Beef
FMK Fresh milk
PTN Plantain
RBN Red beans
BEV Non-alcoholic beverages
POL Poultry
FDW Food away from home
FVG Fresh vegetables and soups
SPC Spices and miscelaneous
POR Pork
SUG Sugar
OCE Other cereals
FIS Fish
BRE Bread
CAS Cassava
OFT Other fruits
OMT Other meats
EGG Eggs
ALC Alcoholic beverages
ORT Other roots
OMK Oter milk and milk products
BAN Banana
CHE Cheese
OLG Other dried legumes
CVG Canned vegetables and soups
POT Potato
NUT Nuts
OBK Other bakery products
ALL All food commodities
NON Non-food commodities


Figure 3.1. Definition of consumtion roups----------------------
Figure 3.1. Definition of consumption groups





55




estimated was decreased by one for each additional row down

the matrix. The strategy was to attempt to minimize

specification errors for the most important items in the

budget.

The elasticities were calculated using the

coefficients estimated with the tobit model.

Own price and cross price elasticities, e with
iJ
respect to those prices which were included in the

estimation equations (see Table 3.1) were calculated

directly. The estimated regression coefficients were

divided by the weighted mean of the dependent variable (the

quantity consumed of each commodity or group) to obtain the

estimated elasticities. To see the reason for using this

procedure consider the semi-logarithmic model,


k
i = a'io + na log p. + b log y + ci log m



where k = the number of prices included as independent

variables in the ith equation. Here,



aq. / p. = a.. /p. and



ei = ( q. /3p ).(p /q. ) = (a.. /p. ).(p./q.) = a.. /q
1 3 3 1 13 ej J 3 ij i


the income elasticity e. and the elasticity with respect to





56





the number of members, e i were calculated following the

same procedure, so

eiy = bi /qi and eim = ci/qi



Nelson (1976) has derived elasticity estimates for the

tobit model based on the expected quantities consumed given

values of the independent variables (in this study price,

income, and family size). His derivation leads to the

formula



ei = [BiPiF(B'P/o)]/LB'PF(B'P/o) + of(B'P/G)] .



For the semi-logarithmic model specification used in this

study the Pi would cancel and the formula would reduce to



e i= [BiF(B'P/l )]/[B'PF(B'P/l ) + of(B'P/ I)]



Bi

B'P + [cf(B'P/c)/F(B'P/c)1 .



This value for the elasticity is approximately equal

to (aij/qi) for values of (B'P/a) over 2.5. There is some

overestimation of the elasticities at the mean quantities

versus expected quantities for purposes of national policy.

However, the overestimation with the method used was minor.-1/




57





The rest of the elasticities for the food commodities

and groups were calculated using the symmetry restriction.

For example, the element in the second row and first column,

e21 of the elasticity matrix in Table 4.2, was estimated

as

e2,1 = (w /w2)el,2 w1(e2,y -e,y ).



The rest of the elasticities in that row were obtained using

the estimated coefficients as explained above.

The elasticities of each food group with respect to

the prices of the all-food aggregate, elf were obtained as

the sum of all the price elasticities in the ith row, or

eif = Eeij *
J
The elasticity of the all-food aggregate with respect

to the prices of each food group, efj was calculated using

the symmetry relation, from the price elasticities of the

food groups with respect to all-food, e .

The own price elasticity of the all-food aggregate,

eyf was obtained as the sum of all the price elasticities

of the all-food aggregate with respect to the price of each

food group, efi or e = ref..

The price elasticities of each food group with respect

to the price of the non-food group, ei,nf were calculated

using the homogeneity restriction




58




eij + einf = -eiy but
J


eij = eif then



einf = -eiy -eif


The elasticities of the non-food group with respect to

the price of each food group, e nf,j ,were calculated using

the symmetry relation and the price elasticities of each

food group with respect to the nonfood group, ei,nf .

The income elasticity of the all-food aggregate, efy,

was obtained as the weighted sum of the income elasticities

of each food group, using the budget shares as weights, or



e = : w.e. /zw. = zw.e. /w
fy i I y i i1 y f


The income elasticity of the nonfood group was

obtained using the Engel aggregation, or

wieiy + Wnf e =1 i.e.
1iy nf,y


e nfy = (1 wi w.e.y )/wnf
if1y 1 iy nf


The cross price elasticity of the nonfood group with

respect to the price of the all-food aggregate was

calculated using the homogeneity restriction or




59





efj + efnf = -ey but
J


efj eff then



ef,nf = -ey -eff


The cross price elasticity showing the effect of the

all-food aggregate on the nonfood group, enf ,f was

calculated from the symmetry relation based on efnf .

The own price elasticity for the nonfood group,

enf,nf was obtained from the homogeneity condition as



enf,nf enfj enfy



The Effect aof Price 3n Income Changes on Calorie and
Protein Intake


The elasticity matrix was used to estimate the effect

of changes in prices on the intake of calories and protein.

The procedure, in general, was as follows: the contents of

calories and protein in one pound of each commodity, ci and

pi respectively, were estimated from tables on food

composition from INCAP-ICNND (1961), ONAPLAN (1980b), and

Whitney and Hamilton (1977); and the percentage change in

the average consumption of each food commodity or food

group, per capita, per day, qi when the price pj changed 1





60





percent was estimated from the price elasticities, eij. The

equation for estimating the change in consumption was



A-qi = (e..ij /100)q. .



Thus, the estimated change in calories consumed from the

good i when p. changed 1 percent was
J


c j = Aqij ci



and the estimated total change in calories consumed from all

food commodities, when pj changed 1 percent was



ACij = = = Aclj + Ac2j + ... + Ac30,j

where i,j = 1,...,30.



The change in protein intake as a result of a change

in price and changes in calories and protein intake as a

result of a change in incomes were determined in an

analogous way. In the policy analysis some assumptions on

price and income changes will be evaluated and compared with

the goals of nutrition in regard with income changes and

food imports.




61






The Data

Data from the "First National Survey on Households

Income and Expenditure" (SHIE), collected by the Central

Bank of the Dominican Republic and the National Office of

Statistics, were used. The first objective of the survey

was to obtain the information necessary to calculate the

official cost of living index. The data were collected

during the period from May 1, 1976, to April 30, 1977. They

were based on a probabilistic sample by areas in multiple

stages. The probability of one household being selected in

the sample was 0.005 or one in each 200 households. The

final sample used in this study was comprised of 4,027

households. The urban and rural distribution and a

comparison with the 1970 census are shown in Table 3.2.

In the data there were 477 consumption items, of which

222 were food. The 477 items were grouped into 30 food

groups (some groups were single commodities) and one nonfood

group. The 30 food groups are shown in Figure 3.1.

(Consumption items in the food groups are given in Appendix

1.) In the equations estimated (Appendix 2) the logarithms

of the prices used as independent variables were designated

with the three letters given in figure 3.1 preceded by the

letter L. For example, the logarithm of the price of fats

and oils is designated LFAT in Appendix 2.














Table 3.2. Sample size and urban and rural distribution of the "First National
Survey on Household Income and Expenditure" in the Dominican Republic with
comparisons to the 1970 census

----- ------------------------------------------------------------------

Sample 1970 Census
No. of Households Percent No. of Households Percent

----- ------------------------------------------------------------------

Total 4,027 100.0 779,920 100.0
Urban Area 1,681 41.7 309,447 39.7
Rural Area 2,346 55.3 470,473 60.3

----- -------------------------------------------------------------------




63









Note

1. For example, estimating the elasticities for rice at the
expected quantities consumed led to the following estimates
-.46, .01, -.12, -.22, -.05, .08, .01, -.00, -.07, .05,
-.05, -.25, -.22, .05, -.00, -.01, -.07, -.07, -.07, .06,
.02, -.06, .02, -.01, .03, -.00, .02, .15, -.04, -.00, .24,
.56. When respectively compared to the first row of Table
4.2 it can be seen that the overestimation was slight.
















CHAPTER IV

EMPIRICAL RESULTS



Tobit Model Estimates

The tobit model was used to estimate the demand

equations following the procedure described in the

methodological section. These coefficients and the t values

associated with their implicit standard errors are shown in

Appendix 2.

The coefficient for the own price in each demand

equation, except in the demand for canned-vegetables

-and-soups (CVG), had the expected negative sign. This sign

implies an inverse relationship between the price of a good

and the quantity demanded of it.

Also the absolute value of the coefficient of the own

price in each demand equation, with the exception of CVG,

was greater than the coefficients of all other prices.

This result shows that, as expected, the effect of the own

price is more important than the effect of other prices on

the demand for a good.

The coefficient of total expenditures for all the

commodities had a positive sign implying that there were no

inferior goods among the food groups.


64




65





The t statistics reported in Appendix 2 clearly

indicate that the elasticities of demand with respect to own

price and the elasticities of demand with respect to income

were estimated most accurately relative to their standard

errors. With cross-sectional survey data the variation in

income is greater than in prices and it would be expected a

greater accuracy in measuring income effects. The high t

ratios for own price effects suggest that there was enough

price variation to consistently measure price effects. The

t ratios for the number of household members was also in

general indicative of an effect of household size on the

purchases of food items.

The coefficients showing the effect of the size of the

family on the quantities consumed of food commodities, in

general, had positive signs. The exceptions were a few

commodities (BEF, FDW, OMT, ALC, OFT, OMK, CHE, and OBK)

which might be considered luxury goods within the food

groups due to their high prices (see Table 4.1). This

conclusion is reinforced by the estimated income

elasticities for those commodities, which were all greater

than one. The point is that when the family size increases

with a given income, cheaper foods are substituted for more

expensive foods.

All the coefficients from the regression in one demand

equation were divided by the mean quantities consumed of the




66






Table 4.1. Average expenditure proportions, quantities
consumed, and weighted average prices paid by households in
the "First National Survey on Household Income and
Expenditure", by food group


Food Expenditure Average Quantity Weighted
Group Proportions Consumed Average


pounds/month $/pounds

RIC 0.0728 63.9910 0.25
FAT 0.0490 12.0711 0.97
BEF 0.0303 8.3404 0.78
FMK 0.0257 4.5411 1.25
PTN 0.0256 47.2419 0.14
RBN 0.0233 12.4066 0.43
BEV 0.0228 13.2299 0.79
POL 0.0224 7.9530 0.64
FDW 0.0209 1.9729 2.89
FVG 0.0200 20.7630 0.30
SPC 0.0176 19.6044 0.24
POR 0.0168 5.0203 0.77
SUG 0.0160 20.3194 0.18
OCE 0.0145 14.2361 0.32
FIS 0.0137 3.7067 1.02
BRE 0.0137 9.5707 0.40
CAS 0.0135 39.3553 0.09
OFT 0.0119 16.5314 0.25
OMT 0.0107 2.2902 1.04
EGG 0.0092 30.6472 0.07
ALC 0.0089 1.3867 7.00
ORT 0.0081 24.3326 0.14
OMK 0.0075 1.9638 1.29
BAN 0.0067 15.6197 0.11
CHE 0.0052 0.8674 1.31
OLG 0.0044 3.5431 0.34
CVG 0.0025 1.0745 1.01
POT 0.0023 3.5716 0.15
NUT 0.0019 3.8800 0.20
OBK 0.0019 0.2641 5.09
- ---------------------------------------------------




67





respective commodity (see Table 4.1) to obtain the price,

income, and family size elasticities in the diagonal and the

upper off diagonal part of the price elasticity matrix.



The Full Matrix of Demand Elasticities

The calculated elasticity matrix (Table 4.2) had 900

price elasticities plus 30 income elasticities for the 30

food groups, 32 price elasticities plus one income

elasticity for the all-food aggregate, and 32 price

elasticities plus one income elasticity for the nonfood

group, for a total of 996 prices and income elasticities.

The 32 groups and their variable names were given in Figure

3.1. The 30 family-size elasticities will be discussed in a

separate section.

Applying the general restrictions of the consumer

theory, only 495 elasticities or



(n 2 + n) ((n 2 + n + 2)/2)



where n is equal to 31 (30 food groups plus one nonfood

group) were calculated directly making use of the estimated

regression coefficients. The rest were derived as explained

in Chapter III.

All the coefficients presented in the elasticity

matrix in Table 4.2 were the result of the applied




68













Table 4.2. Elasticity matrix for food and nonfood commodities


---R-- IC----- FAT------BEF------FMK------PT------RBN------ BEV---
RIC -.468678 .0150396 -.121812 -.224391 -.048324 .0764311 .0090591
FAT .0013273 -1.37831 .0399466 -.021191 .0139341 .1068585 -.010728
BEF -.391032 .0125062 -1.70453 .3880998 .2422791 .4621612 .0178289
FMK -.704585 -.072877 .4695052 -1.73966 .0900230 .1765227 -.041158
PTN -.190008 .0053786 .3057456 .0962717 -1.62926 .4988961 -.035854
RBN .2329542 .2348904 .6392313 .2171300 .5644442 -1.28724 .0110183
BEV -.001672 -.029456 .0519074 -.032684 -.032460 .0032815 -.774118
POL -.079073 .0521671 -.210296 -.017451 -.260845 .1398433 -.141376
FDW -.654877 -.489411 -.397863 -.119285 -.193705 -.197665 -.133329
FVG .1568708 .0777268 .1764044 -.038167 -.054107 .0351349 -.130922
SPC -.249605 .0030094 .2962188 .1424885 -.022344 -.132987 .0255896
POR -1.20349 -.345123 .5686421 .3979459 -.513988 -.424806 .0943125
SUG -1.02641 -.443872 .7838838 .2092158 .1700612 -.476323 .2366396
OCE .2106366 .0198445 -.021178 .2769948 -.345445 -.003008 .0291195
FIS -.050510 .0570831 .3696716 .3022375 -.136728 .0887535 .0381152
BRE -.089414 -.022468 -.106919 -.131639 .0052685 .0421949 -.003413
CAS -.408111 -.419758 .8980362 .5091213 .5273834 -.239844 -.159968
OFT -.504434 -.196100 .0771819 .0442710 -.377728 -.069468 -.059093
OMT -.621277 .1314742 .6459895 -.025551 -.520553 -.516326 -.165390
EGG .4077358 .4468458 1.595364 -.401605 -.099060 .4926756 .0705563
ALC -.521719 -.680581 .0814897 -.096825 -.211606 -.114294 -.585757
ORT -.574183 -.073812 .2093247 .5150462 -.324768 -.195058 .1239140
OMK .0178859 -.239134 -.415854 .5440894 -.378302 .0174882 -.206608
BAN -.147904 -.402379 1.614299 .4270427 -.263817 -.107713 -.159442
CHE .1107597 -1.39507 -2.17848 -1.23576 -1.83773 -.924305 1.762128
OLG -.086048 1.646997 4.586966 3.544175 -.790643 .0617924 -.067987
CVG .5123973 1.360553 -.648218 .0240597 1.325407 .8198891 -.039534
POT 4.905634 4.761158 -2.64277 -.341124 5.547962 1.557051 -.425352
NUT -1.73880 2.475468 1.651586 .6731463 -2.51944 -1.45126 -.245286
OBK -.406634 .0247874 1.607111 -.062502 -.639433 .1291453 -.068498
ALL -.263497 -.154201 .1014618 -.009737 -.125162 -.026405 -.048393
NON -.008912 .0328441 -.038649 .3928913 .0272895 .1391381 .0378603




69














Table 4.2. Continued


---POL------FDW----FVG ------ SPC------POR------SUG------OCE---
RIC -.002116 -.074839 .0494663 -.054681 -.257391 -.222875 .0474739
FAT .0396898 -.101656 .0323500 .0015989 -.102426 -.146714 .0072156
BEF -.163877 -.189692 .0956791 .1536741 .3143862 .3959892 -.024196
FMK -.014336 -.003545 -.042193 .0862794 .2657309 .1182323 .1478950
PTN -.222641 -.060000 -.050174 -.022455 -.329182 .0979258 -.200498
RBN .1551674 -.065610 .0349733 -.096038 -.287113 -.325544 .0025309
BEV -.126207 -.017574 -.116448 .0179215 .0835985 .1625031 .0179366
POL -1.72640 -.135949 -.091148 .1170500 .3168737 .0438199 -.007318
FDW -.245214 -6.84919 -.584827 -.196007 2.618950 1.893060 -.129860
FVG -.087474 -.506591 -1.30310 -.027915 .1904397 .1900881 -.049925
SPC .1646086 -.126602 -.031611 -.567362 -.121687 -.036109 .0121044
POR .4165246 3.342894 .2071881 -.143399 -3.16503 -.073182 -.368103
SUG .0793315 2.582815 .2394031 -.037401 -.059087 -.457666 -.067468
OCE .0023722 -.082038 -.069930 .0135673 -.413074 -.077615 -2.40712
FIS .0025164 .7293255 -.038098 -.040031 -.206233 -.004602 .0039534
BRE .1070434 .0679063 .0638405 -.013524 -.125610 -.063524 .0366223
CAS .6312724 1.591905 -.258442 -.066166 -1.00354 -.198893 -.426279
OFT -.001356 .7512477 -.043646 -.007968 -.481341 -.104555 -.408852
OMT -.821357 3.554726 .2976448 -.127213 .4339475 -.087508 -.166115
EGG .6032040 -.030757 -.061632 .2897114 1.274996 .0019316 -.338014
ALC -.486680 .6610031 -.293090 -.263621 -.461080 -.190489 -.134547
ORT -.408086 2.314882 -.135975 .0159748 .0363135 -.025087 .0914517
OMK -.364127 -1.10712 -.025113 -.212570 -.008941 -.142061 -.191409
BAN .5122970 .6602959 .0325937 .2311630 -.208516 -.098515 -.316049
CHE -.414248 7.714869 .3238960 -.285560 -1.74674 .0702598 -.290836
CLG .7914259 -1.02578 .8370777 -.523126 1.809225 -.200454 .2520830
CVG .6067594 1.519959 -.812643 -.229105 1.140329 .1418944 .4111551
POT -.683898 -20.2892 2.747935 .7909645 6.325170 .1229049 1.21a449
NUT -1.34475 5.218851 -.188226 -.436674 1.106062 -.519088 .5771732
OBK -.400466 -6.51834 -.006278 -.059811 -.203133 -.245502 -.360782
ALL -.090387 .0160301 -.052093 -.040872 -.023864 .0271916 -.117495
NON .0387742 .2331353 .0200807 .0210751 .0734059 -.016122 .0031538





70













Table 4.2. Continued


---FIS------BRE------CAS------CFT------C oT-----EGG------ALC---
RIC -.001616 -.007526 -.074492 -.068217 -.070040 .0593708 .0234377
FAT .0198739 -9.44E-4 -.118365 -.036749 .0471208 .0889397 -.038762
BEF .1563478 -.057491 .3831360 .0286797 .2354929 .4777240 .0988804
FMK .1557346 -.073793 .2557332 .0236068 5.946E-4 -.143623 .0451657
PTN -.075092 .0022014 .2695404 -.170116 -.204217 -.033944 .0072965
RBN .0589122 .0329744 -.138741 -.022037 -.216772 .2008286 .0428401
BEV .0249889 .0014891 -.099094 -.021459 -.060726 .0325475 -.144362
POL -.004162 .0609455 .3676344 .0018735 -.381390 .2450016 -.114095
FDW .4104132 -.020528 .9546401 .3776693 1.783323 -.054995 .3189737
FVG -.022656 .0486153 -.177951 -.015499 .1778163 -.023128 -.046082
SPC -.027642 -.005601 -.053911 .0052233 -.059538 .1560619 -.048841
POR -.176821 -.109873 -.820205 -.341312 .2842243 .6896559 -.166921
SUG .0013140 -.047590 -.168976 -.065681 -.039150 .0074313 -.020138
OCE .0063922 .0386272 -.401030 -.326501 -.105485 -.208884 .0012925
FIS -2.26643 .0865737 .0767535 .0988758 .0220144 .3681472 -.003480
BRE .0851989 -1.65064 .7229140 .1992851 -.148714 .1104829 -.022381
CAS .0844492 .7405521 -2.46890 -.167235 -.158218 -.817234 -.052326
CFT .1047978 .2213606 -.204471 -1.91782 -.003799 -.533802 -.304935
OMT .0085992 -.207905 -.225103 -.013667 -5.54957 1.011021 -.070650
EGG .5461401 .1623003 -1.21607 -.691475 1.195802 -.896023 -.034385
ALC -.131739 -.159447 -.210890 -.512510 -.169077 -.117199 -9.00939
ORT -.309221 .3689959 .1517886 -.003153 .2702793 -.058410 1.631070
OMK -.099549 -.014837 -.066880 -.527333 -.129997 -.106480 -1.11082
BAN -.678602 .3803565 .3286830 .3250292 -.488552 -.166960 -.061887
CHE -.115647 -.455627 -.531897 -.166492 .4436546 -.404729 2.087747
OLG 2.553686 2.041350 -3.02329 .8955413 3.287324 .2433079 8.316117
CVG .0083157 .3797364 .2558534 -.706085 -.172108 .1193439 -6.11157
POT 1.435144 -1.53408 2.076496 -2.96326 .6919679 .4707097 -2.87398
NUT -5.60772 1.099851 2.512835 -.826205 -1.17740 -.141964 -3.61483
GBK -.211946 .1436658 -.387002 -1.46915 .4510727 -.150557 -4.13735
ALL -.028171 -.00531 -.068436 -.116379 -.02553 .0550306 -.137937
NON .0287064 .861019 -.001164 .0154369 .0446424 -.003228 .0972331














E6D000"- S06O00o 8 ZqLS000 9080E 62E200oo t*ELEEEo0 2L2SO'- NON
E8L2O2- 9LLEO- 2oo6 tO ELO'- L6O'- L tESL'- S2620" TIT
L8632O'- 8?tS2'- Zt?2t2 t6208l'- E06t?2' 2L.2Z't- 99L$tW'- liO
9L692ZLl- S259- tEOtr9"- ELL6LEL LtLE69L" 8O91'L 09LEi9"- IfnN
Z6SWl7- OS899"- 4ILLl4E6"- 8L6EL'L- S F28S9" 91 0 E' L09662'L JOd
9tLLS'L- 8tr89FtiW" tit76L" 964y029"- 7IS61?'- 986 1SS*- ZI9986 '- DAD
Lt28 Zt0o8L 29?Z9?L- L8LS9L*E El.ZZP- tLLAOS*- 6080291 CliO
0oL 8'- E6E'- 0oLqgL9" ZL.9Z'S- t790086" 9 080"o 06L6'6- SRO
19SL9Z$" LLLL9l- ZEO'LL- OE90E" 6tOOL'E- 92LESLL 58tr0o4- wE
OS689*- 91t76L'- 7S860'- S699t?90" 80ELO'L 617S0"S- LtSS6F'- X)NO
62t4696P 28tE8t- 4186-o 090oS- LeLE- Oft4rlS- LEEP luo
6969L'- 328L'L- Z0!OL Lt L9S6L'L t7060L'- 91686"- 06SZLt'L D IV
L9LLLL S.Z90" tELOLW S8S60O'- L991F'- 6n.LLO'- 6ZaL'o- D003
OLEtIIV 92A1to- 680SPl o968z zE80V- 9080- LJt?9o6L* IO
99ESIS'- I4L9tL'- L6LtOEP" 09tISO0- trS6SLL* I06LL'- t trLOLO'- 1^O
EOLO2S 9060SO" tOE9966- 6Sfr0v'- 066SS9L" 90LSLOC- 99trS60' 3VO
9E96SE'- E90ZO' L098099" L06LSL'- tlOt?81PL" E2Lt600" 689SL" SH
6&OLLt7" lt900" 8EE89" C9EF0"'- LEEE'- LSL9O'- 01992L'- sI
L26?6" L 6682LO- LLE92LO" Sq9Lt0'- 99Li L'- S916ZO-- 8O500SO' Soo
9S0LLO" ESOSO0' 006ZS0*- IT9t29t70 OE0L60*EO- 020StO- 9StiLO*- DfiS
t29E898* O9L289" t9EEEZt" titEOS'- E6L80'- 02!tE600" 920O6000 HOd
StL9EOL* 6L9S60- 26 6'- L6tL90"- 6EZL620O 2OZ0L0O- S9t9900' D9S
h7LELL 960260'- ZOESL2L' ZLEOLOL" Zt790LO"0 OL O" 289SO9 DA9
9LIt?22- t7tO00L-L OLILS'-- 2L0606"L I908LL LLftrLt'- 69LOLZS' McI
266LLO'- 6?Z0890" 9L OLSS 61* 909LO- I2LLH61" OS90L-- 090tSL- liod
t6no0- 69LLOO0- 9L2LLO'- {4902t7* 6S290"- SLE40"-S8SZt0" ASa
2LLtStl 2oo0060o 08LL2LO 2021 8820- t2220LL SE2990'- NS9
6LE96r 2O2OOEL" E9S9EL'- 4O0LS'- 9t4OZ0'- L2tE60"- G6LLOL'- id
6E2EO- fE2EOS0" LELE609" 9t4662'- S08080L 4EIZ4 LL" 8EOSL" )w1a
2iELEO2- 16L2SO*- 9LLS299* L2ZSS- tSZ2OS 2060- 29EOL70 ;[
600I22* SOS9LLO" St?980SrL OL2tZL'- L6SESO'- SS09L0t- 8S2!L0- lV2
9t0095L" St0020' E06L0O- 98SEZE0" 9620O0"- S94i O" EL90"- OIE
-O----Ld--- OAD ----3... ----3--O O --- NV ------)N1DO------ O--


panuTquoo "'-t GTqe,




72












Table 4.2. Concluded


---NUT------ 0BK---ALLFOOD-N CN-F OOD--IN COME--
RIC -.045119 -.003569 -1.30598 .1886992 .2464378
FAT .0961801 .0074558 -1.21402 .4408520 .5352039
BEF .1017460 .1052350 1.501769 -1.06439 1.598464
FMK .0485129 6.166E-4 -.161558 7.418284 1.198487
PTN -.188523 -.041770 -2.30136 .4213434 .9696625
RBN -.118389 .0174181 -.101468 3.196350 .3243274
BEV -.020869 5.442E-4 -.766214 .8697111 .6663845
POL -.115667 -.028731 -2.00511 .6177352 1.238625
FDW .4652566 -.594711 -1.82417 3.107933 5.669399
FVG -.018172 .0058325 -.962248 .5916515 .5693156
SPC -.047545 -2.55E-5 -.817862 .6905149 .5647457
POR .1232190 -.018246 -.830175 1.804281 1.497870
SUG -.061700 -.022427 1.255585 -.349708 .4390878
OCE .0755965 -.040966 -3.73929 .1679450 .6279950
FIS -.781780 -.023471 -.813257 1.011188 .8223554
BRE .1523711 .0257034 -.029459 31.39321 .9248112
CAS .3550119 -.047577 -2.07493 .1619863 .3361097
OFT -.133957 -.229291 -4.97955 .2924275 1.456156
OMT -.212780 .0831820 -1.68737 1.332207 2.247922
EGG -.030828 -.025878 3.066789 -.368595 1.130403
ALC -.795114 -.897658 -12.1737 .8261623 10.05747
ORT -.150757 -.105821 .5699425 -.830109 .4731143
OMK .3204962 -1.20296 -11.2678 .2906487 3.274978
BAN .0477090 -.206579 -3.02762 .2442520 .7395020
CHE .4979151 -.067903 -3.18660 1.557967 4.964625
OLG -1.22408 -.219353 16.39067 -.060932 .9987187
CVG -4.88981 -.190790 -7.83437 .2150543 1.684815
POT -1.47873 -.055998 -5.63910 .0867209 .4890273
NUT -1.35568 -.047835 -14.6953 .0316775 .4655107
OBK -.054649 -6.99516 -27.0397 .1462476 3.954494
ALL -.057541 -.097627 -1.41957 .1655520 1.254023
NCN -.000414 .0066494 .4192827 -2.80366 .7461540





73




methodology. Some of these coefficients are out of a

logical range; however they were not intuitively adjusted as

in George and King (1971, p. 39) or prespecified as in

Hassan and Johnson (1976, p. 41). This was done in order

not to obscure the results obtained with the data and method

used.

As indicated earlier, the results of the estimation

procedure were affected by the ordering of the commodity

groups. The groups were arrayed in a descending order by

the budget shares. As a consequence, any error build-up

associated with the procedure will fall on those food groups

that are least important in household budgets. Elasticity

coefficients associated with these minor items in the budget

are the ones closer to the bottom of the matrix. Problems

with errors build-up were by and large associated with

estimated coeificients corresponding to commodity groups

with an expenditure proportion equal to or less than .006

(see Table 4.1). These are cheese, other dried legumes,

canned-vegetables-and-soups, potatoes, nuts-and-other-

oily-seeds, and other bakery products. All of them together

represent 1.8 percent of the average total expenditures of

households in the sample.

Looking at the matrix in general, it can be observed

that most of the coefficients have a negative sign. For the

own price elasticity this result was expected. For the




74





cross price elasticities a negative sign means that the

commodities are complements. Many of these commodities with

negative cross price elasticities were expected to be

substitute or independent groups. In many cases the

unexpected negative sign can be explained because of the

general substitution effect. Even though two goods i and j

were independent, a change in the price of j does not leave

the demand for i unaffected because there is an overall

effect due to the fact that all goods compete for the

consumers' income (See Phlips, 1974, p. 63). In a complete

demand system this fact becomes obvious.

Observing closer the results for own price

elasticities, the inelastic commodities, those with own

price elasticities greater than -1, were rice (-0.47),

non-alcoholic beverages (-0.77), spices and miscellaneous

(-0.57), sugar (-0.46), eggs (-0.90), and potatoes (-0.48).

The expenditure on those commodities will increase if their

prices increase. Ten commodities were in the elastic range

of -1 to -2, and include fat-and-oils (-1.38), beef (-1.70),

fresh milk (-1.74), plantain (-1.63), red beans (-1.29),

poultry (-1.73), fresh vegetables (-1.30), bread (-1.65),

other fruits (-1.92) and nuts-and- oily-seeds (-1.36). Seven

commodities had own price elasticities less than -2. These

coefficients seem too high in absolute value and included

food away from home, alcoholic beverages, other-milk-and-




75




milk-products, cheese, other dried legumes, and other bakery

products.

Price elasticities for beef and pork were estimated by

the USDA (1978) for Mexico and Central America. The own

price elasticity of the demand for beef was estimated to be

-0.4; elasticity of the demand with respect to the price of

pork, 0.1; income elasticity of beef, 0.7; own price

elasticity of the demand for pork, -0.3; elasticity of the

demand for pork with respect to the price of beef, 0.1; and

income elasticity of pork, 0.6.

The signs were consistent, but in each case the

absolute values of the elasticities in this study were

larger (by factors of 2.3 to 10.7) than the USDA estimates

for Mexico and Central America. Also, in this study, some

of the cross price elasticities were excessively high--for

example, potatoes, other dried legumes, canned vegetables,

nuts-and-oily-seeds, and other bakery products. Also the

elasticities showing the effect of the price of food away

from home on the consumption of several commodities were

believed to be too high.

In general, the estimated cross and direct price

elasticites for most of the commodities were within a

logical range. For instance, on consulting the elasticities

table it follows that for a 10 percent increase in the price

of rice the consumption of rice will decrease by 4.7




76




percent. The cross price elasticity of beef with respect to

the price of rice (row 3, colunm 1) shows that for the same

increase of 10 percent in the price of rice the consumption

of beef was reduced by 3.9 percent. Also a 10 percent

increase in the price of nonfood decreased the consumption

of rice by 1.9 percent. An increase of 10 percent in income

increased the consumption of rice by 2.5 percent, and so on.

All-food has an estimated income elasticity of 1.3

which is reasonable for the country. The average total

expenditure level was $220 per month, of which 50 percent

was spent on food. For comparison Del Rosario and Musgrove

(1980) estimated expenditure elasticities for the food,

beverages, and tobacco group of 1.6, 1.0, 0.6, and 0.5 for

expenditure levels of $100, $200, $400, and $800

respectively.. The estimated income elasticity for food was

greater than the nonfood income elasticity. This result is

reasonable in a country with a low total expenditure per

household and implies that with a given increase in income,

the consumption of food will increase more than the

consumption of nonfood items. This result has implications

for the nutritional policy discussed in the next chapter.

The income elasticities for alcoholic beverages and

food away from home are probably too high. The problem

might be related to their high price/weight ratio since

these elasticities are the result of dividing the




77





coefficient for total expenditures from the regression

equation, bi, by the mean quantity consumed measured in

pounds per month.

Most of the income elasticities are within a logical

range. For comparison FAO (1971) estimated income

elasticities for the Dominican Republic for rice (0.60),

starchy roots (0.29), sugar and sugar products (0.60),

oilseeds (0.50), vegetables (0.40), fruits (0.26), beef and

veal (0.80), poultry (1.0), pig meat (0.70), eggs (1.0),

fish (0.60), milk (0.70), cheese (1.0), and fat and oils

(0.69).

The own price elasticity of all-food was -1.4. As

indicated before, this estimated elasticity was obtained by

summing the price elasticities of all-food with respect to

each food group, e fj. One would expect the absolute value

of this elasticity to be lower. However, in comparison with

the own price elasticity for nonfood (-2.8), all-food was

more inelastic than nonfood, as was expected.

Further interpretation of this table may be

interesting for specific commodities and in general, cross

price elasticities in a demand system are important. The

usefulness of this matrix and some implications for policy

analysis will be discussed in the next chapter.





78





The Family-Size Elasticities

A family-size elasticity may be interpreted as the

percentage change in the consumption of a good as the result

of a 1 percent change in the size of the family. Most of

the elasticities were positive and within a range of 0 to 1

(Table 4.3). This implies that the demand for those goods

will increase when the family size increases, but in a

smaller percentage. The commodities with a negative

elasticity, as indicated earlier, were beef, food away from

home, other fruits, other meats, alcohol, other milks and

milk products, cheese, and other bakery products. Except

for other fruits, all were most expensive products (see

Table 4.1) and can be considered as "luxury" foods.



The New Method of Estimatin EnEel Elasticites

The results of the new method of estimating Engel

elasticities were applied to the consumption of rice. The

estimated values of the elasticities were disappointing and

excessively high. The results are presented in Appendix 4.

Estimated expenditure elasticities used in the policy

analysis were the ones obtained from the elasticity matrix.

These coefficients are consistent with the price

elasticities since they belong to the same demand system.

The elasticities obtained using the new method which

was described in the theoretical framework section though




79










Table 4.3. Estimated family-size elasticities


F AM. SIZE
RIC .5728634
FAT .1864453
BEF -.150928
FM.K .0938988
PTN .1399309
RBN .4891751
BEV .1264484
POL .0058217
FDW -1.96352
FVG .1371189
SPC .1010487
POR .0600757
SUG .2340128
OCE .2897563
FIS .1296042
BRE .3040631
CAS .7693678
OFT -.176609
OMT0 -.394077
EGG .0015564
ALC -4.05590
ORT .8337375
OM1 -1.38754
BAN .6814984
CHE -1.87372
OLG .4978635
CVG 2.606377
POT .7955863
NUT .7251987
CBK -.100358




80





seemingly overestimated were logical in relationship to

total expenditures. The elasticities for rice increased as

total expenditures increased up to $200 per month where they

reached the maximum. At total expenditure level over $200

the expenditure elasticities for rice began to decrease.















CHAPTER V

IMPLICATIONS FOR POLICY



The data used in this research recorded the amount and

the value of food items purchased by sampled households. In

order to derive nutritional implications, it was assumed

that the food purchased was consumed by the household. In

the Dominican Republic this assumption is close to reality,

especially for low-income households. Since the data do not

provide knowledge of the distribution of food among family

members, it was further assumed that food was distributed in

proportion to the needs of each household member.

Using the results of the elasticity matrix, the

nutritional implications of the demand system were analyzed

from the point of view of the potential impact of income

changes on calories and protein intake. Income and price

elasticities were used to analyze how a change in income

affected nutrient intake.

Before going further it is necessary to clarify the

meaning of nutrition as this concept is used in this study.

To analyze nutritional policies, some indicator about

nutrition has to be used. But to choose an indicator is a

controversial matter. In this study the indicators were



81





82




determined by the kinds of data available to be calories and

protein intake per person per day. These are flow

measurements, but are also commonly taken as shorthand

indicators of current or likely future nutritional status

(Dowler et al., 1982, p. 103). Indicators of state are also

commonly used. These indicators are based on anthropometry

or on biochemical nutrition parameters. For a discussion of

state indicators see Moussie (1981, p. 10); and Davis

(1982).



The Effect o1 Changes in Prices on the Calorie and Protein
Intake


Following the method described in chapter IV, the

change in calorie and protein intake as the result of a

change in the price of a commodity was estimated. The

estimates were based on consumption per capita per day of

each commodity, the price elasticities (Table 4.2), and the

calorie and protein contents of each commodity (Table 5.1).

The estimated effect of a 10 percent increase in the

price of rice on calorie and protein intake was a reduction

of 70 calories and 0.8 grams of protein in the overall per

capita intake per day (Table 5.2). It should be pointed out

that such estimates based on discrete changes can only be

approximated by the point elasticities given in Table 4.2

since point elasticities are based on infinitesimal changes.





83





Table 5.1. Calorie and protein in one pound of edible
portion by food commodity group



Commodity Calories grams of protein



RIC 1645 33.0
FAT 3676 0.9
BEF 1107 84.8
FMK 277 15.9
PTN 597 5.4
RBN 1526 99.0
BEC 843 18.0
POL 771 84.0
FDW -- --
FVG 222 2.5
SPC 590 5.4
POR 980 70.0
SUG 1678 0.9
OCE 1674 47.9
FIS 453 94.0
BRE 1424 46.7
CAS 598 4.5
OFT 295 3.7
OMT 907 80.0
EGG 671 51.3
ALC 616 --
ORT 526 7.9
OMK 1002 27.2
BAN 498 5.4
CHE 1007 108.9
OLG 1526 102.1
CVG 216 8.5
POT 412 13.7
NUT 1890 52.8
OBK 1769 33.0
---------------------------------------------------

Source: Calculated by the author based on average for the
groups from information about the food composition for
single commodities from: INCAP-ICNND, 1961; Whitney and
Hamilton, 1977; and ONAPLAN, 1980b.





84






Table 5.2. Effect of a 10 percent increase in the price of
rice on calorie and protein intake per capita per day
---- -----------------------------------------------------

Commodity Change in Calorie change in gms. of
Intake protein intake

---------------------------------------------------------

RIC -29.90 -0.600
FAT 0.04 --
BEF -2.38 -0.183
FMK -0.56 -0.032
PTN -3.32 -0.030
RBN 2.73 0.177
BEV -.01 -0.000
POL -0.32 -0.034
FDW -- --
FVG -0.47 0.005
POR -3.75 -0.268
SUG -22.43 -0.012
OCE 3.06 0.088
FIS -0.05 -0.011
BRE -0.76 -0.025
CAS -5.64 -0.042
OFT -1.57 -0.020
OMT -0.83 -0.073
EGG 5.48 0.419
ALC -0.33 --
ORT -4.32 -0.065
OMK -0.02 0.001
BAN -0.71 -0.008
CHE 0.07 0.007
OLG -0.34 -0.023
CVG 0.09 0.003
POT 4.95 0.164
NUT -7.95 -0.221
OBK -0.12 -0.002
ALL -70.25 -0.802


Source: Calculated from the information given in Tables 4.2
and 5.1.




85




Also, estimates based on finite changes would be most

accurate when the changes are near the average prices

observed in the data. Finally, the estimates assume ceteris

paribus conditions, i.e. only the price of rice changes.

The major reduction in calories was due to the drop in the

consumption of rice (RIC) wich accounted for a reduction of

29.9 calories. This was followed by a reduction of 22.4

calories due to the decrease in the consumption of sugar

(SUG). However, there were some gains in calorie intake as

the price of rice was increased. For instance, eggs (EGG)

contributed with 5.5 calories, potatoes. (POT) with 4.9 and

other cereals (OCE) with 3.1 calories.

The drop in rice (RIC) consumption as a result of the

price increase was mainly responsible for the reduction in

protein intake--accounting for 0.60 grams of it. Pork

(POR), beef (BEF), and nuts and oily seeds (NUT) caused

reductions of 0.27, 0.18, and 0.22 grams of protein

respectively. Major gains in protein intake came from an

increase in the consumption of eggs (EGG), potatoes (POT)

and red beans (RBN), which accounted for 0.42, 0.16, and

0.18 respectively. However, the increase in the consumption

of red beans was not expected since rice and red beans are

consumed together and considered to be complements in the

Dominican diet. The results suggest that the proportions of





86






rice and red beans used in the diet are responsive to

relative price changes.

These results provide an example of how the estimated

elasticities (Table 4.2) may be used to measure the effect

of an increase in the price of a commodity on the intake of

calories and protein in the Dominican Republic. Such

measures can provide useful insight into how a change in

price of a commodity can affect nutritional goals for

persons concerned with food price and nutritional policies.

The same analysis could be done for each commodity or group

of commodities included in the elasticity matrix.



lTh Effect of Changes in Income the Calorie and Protein
Intake


The effect of an increase of 10 percent in real income

was evaluated with the income elasticities given in Table

4.2 to see how such a change affected the overall intake of

calories and protein. As can be observed in Table 5.3, the

consumption of calories increased by 166 per capita per day.

The major increase was due to plantain consumption (PTN)

which accounted for 16.9 calories, followed closely by rice

consumption (RIC) which contributed 15.7 and by fats and

oils (FAT) which contributed 15.2 calories. Conditions on

these estimates similar to those given on pages 82 and 85





87






Table 5.3. Effect of a 10 percent increase in income on
calorie and protein intake per capita per day
---------------------------------------------------------

Commodity Change in Calorie change in gms. of
Intake protein intake



RIC 15.72 0.315
FAT 15.51 0.004
BEF 9.75 0.746
FMK 0.95 0.054
PTN 16.92 0.153
RBN 3.80 0.246
BEV 4.84 0.103
POL 4.96 0.540
FDW -- --
FVG 1.72 0.019
SPC 4.27 0.039
POR 4.66 0.333
SUG 9.60 0.005
OCE 9.14 0.261
FIS 0.88 0.183
BRE 7.86 0.258
CAS 4.64 0.035
OFT 4.52 0.057
OMT 2.99 0.264
EGG 15.19. 1.162
ALC 6.29 --
ORT 3.56 0.053
OMK 4.11 0.112
BAN 3.53 0.038
CHE 2.97 0.322
OLG 3.91 0.262
CVG 0.28 0.011
POT 0.49 0.016
NUT 2.13 0.059
OBK 1.17 0.022
ALL 166.38 5.675
------------------------------------------------

Source: Calculated from the information given in Tables 4.2
and 5.1.





88




are needed. Furthermore here any problem on the food supply

side are ignored. With a fixed supply of food an increase

in incomes would simple lead to faster inflation. It is

assumed here that the change is in real and not in nominal

income.

The cross-sectional survey data used in the analysis

had greater variation among households than did the price

data. As a consequence, the estimates of the income

elasticities were generally estimated with the smallest

relative standard errors (see t statistics in Appendix 2).

The overall increase in protein was 5.7 grams per

capita per day. The main contributions were from egg

consumption (EGG, 1.16 grams), beef consumption (BEF, 0.75

grams), and poultry cosumption (POL, 0.54 grams). Rice

consumption (RIC) contributed with 0.32 grams. It can be

seen that rice (RIC) and eggs (EGG) are of great importance

in the supply of both calories and protein to the Dominican

population and must be given special consideration in any

food policy. The contribution of cheese (CHE) and canned

vegetables (CVG) must be taken with caution due to their

small share in the consumer budget.

To analyze the impact of income changes on the

nutritional deficit at the national level the increase in

calories and protein has to be compared with the nutritional

needs of the population. ONAPLAN (1980b, p. 20) calculated





89




the recommended consumption of calories and protein per

capita per day as 2356 calories and 49.8 grams of protein.

The increase of the intake of these nutrients as a result of

a 10 percent increase in income would contribute importantly

to correcting the calorie and protein deficiencies of the

population. These per capita estimates at the national

level provide only broad guidelines and say nothing about

the distributional effects and the real nutritional status

of low income groups of the population. Measures of demand

and income elasticities are needed for different income

groups by geographic areas to understand the distributional

effects of food policies and nutrition. It would also be

interesting and informative to obtain biochemical parameters

data on the nutritional status of low income groups.

Fairness and equity aspects of food policies cannot be

ignored if food policies are to encourage political

stability. Therefore, distributional aspects are of major

concern and deserve serious analysis.



The Effect of a Change in Income and Price on the
Consumption and Imports of Selected Food Commodities


This section is concerned with analyzing the impact on

consumption and food imports of (1) a 10 percent increase in

real income per capita and (2) a 10 percent increase in real

income coupled with a 10 percent increase in price. The




90





second scenario provides some insight into a situation of

rising incomes and rising inflation. The change in

consumption was compared with the production, consumption,

and import goals of the Secretariat of State for Agriculture

(SEA) operative plan. The commodities included in the

analysis were those which have been imported or exported in

recent years and which were included as single commodities

in the elasticity matrix. These were rice (RIC), red beans

(RBN), and cassava (CAS). Plantain was excluded because it

was measured in units and not in weight in SEA's operative

plan.

The analysis was done separately for each commodity.

The consumption estimates used as a basis to calculate the

effect of changes in income and prices were those from SEA's

operative plan (TabLes 5.4, 5.5, and 5.6). These

consumption estimates were lower than those from the "First

National Survey on Households Income and Expenditure (SHIE)

for 1976-77. These SEA estimates were used in order to

unambiguously compare the results of the estimates based on

elasticities from Table 4.2 with 1982 goals of the SEA plan.

The SEA rice program proposed a consumption goal of

314 thousand metric tons for 1982 (Table 5.4). This implies

a per capita consumption of 0.32 pounds per day. This

consumption goal was based on a projected production of 285

thousand metric tons and a stock at the begining of the year





91










Table 5.4. Rice program from SEA's operative
plan

--------------------------------------

1980 1981 1982 goal
---------------------------------------
in thousand metric tons

Production 254.5 260.2 285.2
Initial Stock 30.8 34.1 65.8
Exports -- -- --
Imports 40.6 62.6 --
Supply 325.9 356.9 350.0
Consumption 291.8 292.1 314.3
Final Stock 34.1 64.8 35.7

in pounds per day
Consumption
per capita 0.316 0.308 0.322

in $ per pounds

Price 0.33 0.36 --

in thousand

population 5578.2 5731.6 5889.3


Source: Secretaria de Estado de Agricultura
(SEA). Plan Operativo 1982. Santo Domingo,
Dominican Republic: SEA, 1981; and ONAPLAN.
Plan Trienal de Inversiones Publicas,
1980,82, Santo Domingo, Dominican Republic:
PLANDES 42, ONAPLAN, 198Ua.




92









Table 5.5. Red beans program from SEA's
operative plan

-------------------------------------

Commodity 198u 1981 1982 goal
-----------------------------------
in thousand metric tons

Production 36.6 37.8 45.2
Seeds and
Crop Loses 4.4 4.2 5.0
Initial Stock 0.4 2.8 0.3
Exports -- -- --
Imports 6.6 -- --
Supply 39.2 36.4 40.5
Consumption 36.4 36.1 38.4
Final Stock 2.8 0.3 2.1

in pounds per day
Consumption
per capita 0.039 0.038 0.040

in $ per day
Price 0.63 0.64 --

Source: Secretaria de Estado de Agricultura
(SEA). Plan Operativo 1982. Santo Domingo,
Dominican Republic: SEA, 1981; and ONAPLAN.
Plan Trienal gde Inversiones Publicas,
1980,b2. Santo Domingo, Dominican Republic:
PLANDES 42, ONAPLAN, 198ua.





93










Table 5.6. Cassava program from SEA's
operative plan

--------------------------------------

1980 1981 1982 goal
-------------------------em----------------
in thousand metric tons

Production 80.9 119.3 134.4
Exports 1.2 3.8 4.4
Imports -- -- --
Supply 79.7 115.5 130.0
Consumption 79.7 115.5 130.0

in pounds per day
Consumption
per capita 0.086 0.122 0.133

in $ per day
Price 0.28 0.19 --

Source: Secretaria de Estado de Agricultura
(SEA). Plan Operativo 1982. Santo Domingo,
Dominican Republic: SEA, 1981; and ONAPLAN.
Plan Trienal de Inversiones Publicas,
1980,82. Santo Domingo, Dominican Republic:
PLANDES 42, ONAPLAN, 198Ua.




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EKGGL3FUF_O56ER0 INGEST_TIME 2015-01-16T18:12:01Z PACKAGE AA00026597_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

DEMAND FOR FOOD COMMODITIES IN THE DOMINICAN REPUBLIC AND SOME IMPLICATIONS FOR NUTRniONAL POLICIES BY MARINO CHANLATTE A' DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY University of Florida 1983

PAGE 2

ACKNOWLEDGMENTS The author wishes to express appreciation and gratitude to Dr. Max R. Langhara, chairman of his supervisory committee, for the inspiration, encouragement, and contributions throughout work on this dissertation. Special thanks are due to Dr. W. W. McPherson, a member of the supervisory committee, for his guidance during the earlier period of the author's Ph. D. program. The author also wishes to thank Dr. Carlton G. Davis and Dr. Gustavo Antonini, members of his supervisory committee, for their help and contributions. Special thanks are due to Mr. Rom Alderman for his assistance in computer programming, and to Mr. Laurent Ross for cleaning the data and preparing the computer tape used in this study. Dr. Arturo Martinez Moya and Gumersindo del Rosario of the Central Bank of the Dominican Republic, and David Jessee of USAID were very helpful in providing data. Dr. Mervin Yetley of USAID, Dr. Per Pinstrup-Andersen, Dr. Philip Musgrove and Mr. Ivan Castelar provided fruitful comments. ii

PAGE 3

Mrs. Pat Smart and Dr. Rekha Mehra edited portions of this work. Special gratitude to Dr. Ruben Nunez, of the Secretariat of State for Agriculture of the Dominican Republic and to Felipe Manteiga, of USAID, for their encouragement, moral support, and contributions. The author is deeply indebted to the Secretariat of State for Agriculture of the Dominican Republic for providing financial support without which this endeavor would not have been possible. Finally, the author wishes to extend appreciation and gratitude to his wife Nellie for her encouragement, sacrifice, and labor and to his daughters Michelle and Lena and his son Marino for their love and understanding. i i i

PAGE 4

1 TABLE OF CONTENTS PAGE ACKNOWLEDGMENTS ii ABSTRACT vi CHAPTER I INTRODUCTION 1 Problem Setting and Objectives 1 Characteristics of the Dominican Economy 5 Plan of Dissertation 19 II THEORETICAL FRAMEWORK AND REVIEW OF LITERATURE 21 Household Budget Surveys 22 Engel Curves 23 An Alternative Method of Estimating Engel Curves. ...29 Application of the Frisch Method 35 Price Elasticities from Cross-Sectional Data 40 Demand Estimates in the Dominican Republic 43 The Limited Dependent Variable Model 45 III METHODOLOGY 48 The Restrictions of Consumer Theory 48 The Demand Models 49 Use of the Tobit Model . .52 The Elasticities Matrix 53 The Effect of Price and Income Changes on Calorie and Protein Intake 59 The Data !61 Note ','.63 IV EMPIRICAL RESULTS 64 Tobit Model Estimates ; 64 The Full Matrix of Demand Elasticities .......67 The Family-Size Elasticities 78 The New Method of Estimating Engel Elasticities. !!! .'78 iv i

PAGE 5

1 V IMPLICATIONS FOR POLICY 81 The Effect of Changes in Prices on the Calorie and Protein Intake 82 The Effect of Changes in Income on the Calorie and Protein Intake 86 The Effect of a Change in Income and Prices on the Consumption and Imports of Selected Food Commodities 89 VI SUMMARY, CONCLUSION, LIMITATIONS AND RECOMMENDATIONS 99 Suranary and Conclusions 99 Limitations of the Study 101 Suggestions for Further Research 103 APPENDIX 1. COMPOSITION OF FOOD GROUPS 106 APPENDIX 2. REGRESSION COEFFICIENTS FROM TOBIT MODEL Ill APPENDIX 3. ROTATION OF AXES 142 APPENDIX 4. RESULTS FROM THE NEW METHOD OF ESTIMATING ENGEL ELASTICITIES 145 REFERENCES 146 BIOGRAPHICAL SKETCH 150 V

PAGE 6

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEMAND FOR FOOD COMMODITIES IN THE DOMINICAN REPUBLIC AND SOME IMPLICATIONS FOR NUTRITIONAL POLICIES By Marino Chanlatte April 1983 Chairmam: Max R. Langham Major Department: Food and Resource Economics The nutritional problem has been identified as one of the most important development issues in the Dominican Republic. The nutritional status and other development problems associated with food consumption projections and the effect of food demand on trade are of major concern for the Dominican policymakers. This study developed a full matrix of demand elasticities for food commodities at a disaggregated level. The primary concern of this study was to develop these estimates and to demonstrate their usefulness in evaluating nutrition and food consumption policies. vi

PAGE 7

1 Coefficients estimated from 30 demand equations were used to estimate the elasticities matrix. These demand equations were estimated using the tobit model and data from a national household consumption survey. This matrix was comprised of 996 price and income elasticities for 30 food commodities and groups of commodities, the all-food aggregate, and one nonfood group. Family size elasticities were also estimated. In general, the estimated cross and direct price elasticities for most of the commodities were within a logical range. The effects of changes in prices and income on calorie and protein intake were estimated using the elasticities. An increase of 10 percent in the price of rice, and also in income, were postulated, and the change in calorie and protein intake were estimated. An analysis of the effect of a change in income and prices on the consumption and import goals of the Secretariat of State for Agriculture (SEA) crop programs was undertaken for rice, red beans, and cassava. The conclusions were that, in the three cases, the projections of the SEA'S operative plan could be improved by making use of price and income elasticities. vii

PAGE 8

CHAPTER I INTRODUCTION Problem Setting and Objectives One of the main objectives of the governmental development plans in the Dominican Republic is to improve the nutritional status of the lower income groups of the population (Secretaria de Estado de Agricultura, SEA, 1976; SEA, 1979; Oficina Nacional de Planif icacion, ONAPLAN, 198ua). To make explicit the importance that the government has placed on this objective, and the means to achieve it, ONAPLAN published a detailed study on the nutritional situation of the country (ONAPLAN, 1978) and later presented a paper giving basic suggestions for a national policy on food and nutrition (ONAPLAN, 1980b). This paper led to discussions and to the current formulation of food and nutrition plans. Some of the proposed instruments to cope with the nutrition problem are increased income per capita, the redistribution of income, and increased food production. Along with these instruments, the plans being formulated also call for greater self sufficiency (fewer imports) of basic food crops. 1

PAGE 9

2 Some measures have been taken to increase the income of the low-income groups. The minimum wage for agricultural workers was increased from $2.50 for 8 hours of work to $3.50 (Law 25 of 1979, Quezada 1981, p. 39). In other sectors of the economy, minimum wages were raised to $125.00 per month (Law 45 of May 25, 1979). Also, an increase of 10 percent was established in the wages and salaries of all the workers who were earning $300.00 or less. Income taxes were also modified to favor the low-income groups (Law 39, May, 1979, ONAPLAN 198ua, p. 32). Regarding production goals, the SEA' s Operative Plan for 1982 contains specific production and consumption goals for 22 products along with a package of projects and programs to reach these production goals. ONAPLAN has pointed out the need to estimate the effect of changes in income among income groups on food consumption and the impact of increased income on the nutritional situation of each income group. This agency also called for the study of the effect of food policy on production and trade (ONAPLAN 1980b, p. 13). Some work has been done to estimate income elasticities (ONAPLAN 198ub). The range of income was divided into four income strata for 22 food commodities. Using data from the 1969-70 Central Bank survey on household income and expenditure, the change in expenditures on food

PAGE 10

3 between two income groups was used to calculate a gross income elasticity. With this arc elasticity and three postulated income distributions, ONAPLAN projected the consumption of food products and compared consumption with the production goals of food crops from SEA' s Operative Plan to analize the need for food imports (ONAPLAN ig^Ob, p. 38-39). A more sophisticated measure of the effect of income on food expenditure is needed to improve projections for policy purposes. The effects of prices on the food consumption and family budgets were not measured. Estimates of the direct and cross price elasticities are needed to study the effect of the change in the price of one product on the consumption of that product and on the other products being considered. Thus, estimates of income and price elasticities are necessary to analyze the effect of income and price changes on the consumption of food commodities and, through the nutrient characteristics of each commodity, to calculate the impact on the nutrition levels of different income groups. It is also possible to use such estimates to analyze the effect of government policies aimed at redistributing or increasing incomes on the food price level and food imports.

PAGE 11

4 The general objective of this study was to analyze the effectiveness and consistency of government food policies concerning income and price changes, and basic food imports. The study was partial in the sense that it focused mainly on the demand side. The impact on food imports (production deficits) was based on various assumptions about the effect of domestic policies on the supply side. Specific objectives of the study were as follows: 1 — to estimate a full matrix of demand elasticities of disaggregated food commodities from cross sectional data, 2 — to estimate expenditure elasticities of selected food commodities using a new method that provides more flexible coei f icients 3 — to estimate the effect of household size on consumption of food in the Dominican Republic, 4— -to estimate the potential effects of changes in prices and income on the nutrition level of the population of the Dominican Republic, and 5 — to analyze food and nutritional policies with the information generated and to determine the consistency or inconsistency of the food policy instruments proposed.

PAGE 12

5 Characteristics of the Dominican Economy The Dominican Republic occupies the eastern two-thirds of the Island of Santo Domingo (Figure 1.1). It encompasses an area of lb, 900 square miles and had an estimated population of 5.4 millions in 1979, a density of about 2b6 people per square mile (ONAPLAN, 198Ua, p. 1 ) It has a 1,000-mile coast line, and a 193-mile common border with Haiti. Its relatively high population density (compared with other Latin American countries) is of concern since only 44 percent of the total area is agricultural land (22.4 percent crop land and 21.6 percent pastures). Forest land and reserves comprise 54.4 percent and the rest consists of lakes and non-usable land (ONAPLAN, 19B0a, p. 1). Even though productivity is low, the agricultural sector has played, and must continue to play, an important role in the economic development of the country. In 1970, 55.5 percent of the labor force was engaged in agriculture; its average productivity (value added per person employed) was only $506. In comparison, 7.6 percent of the labor force was engaged in manufacturing and had an average productivity of $2,646 (Table 1.1). The shares for agriculture, manufacturing, and commerce, which account for more than half of the GDP, declined during the period 1976-79. Government,

PAGE 13

6

PAGE 14

7 -o 0) 1— I > 4-> C o 1CD Cm O -69n OS c Q o H c\j 1—1 vO •rH r1 LTi O vo ro o in o — (M O fC\J ^ ^ O OO vo 0^ ^ t— ro — T— T— r— MD ro ^ on c\j T— LTi I o o o 4-J CO c x: > o I— I Q. s t — — c\j o cx! oo CM CO c M \ O o I I 1 I I I I rI I 1 I I ^ I oo I I 1 CM I c~I o I ~ I c •H 0 t, •rH -P D 3 4^ E 4-> •J-) CJ CO rH 0 3 s tu 3 bO CO C 0 O C Cm -P 1. r-t rH H -H 3 (U to > to U C C 3 C > 1bO-H ca 0 0 0 0 ; 2: 2: a. 0 CO E-H

PAGE 15

8 o 0) Q. a s a w c 0) o D CO .-I i> cu •69Q 0^ o Lo vo ir\ -rCD CM =T — OO CO LTi LTi CM T— 0^ t— cr\ c\i oo cvj TTCO c I on I o 1 c\j I I t0) in on CO 1 Q.-1 0) £> 1 O 1 c c 3 c > S1 o 3 > CO 1 0) 1 H CD o o o O 1 \ c O H 1 00 1 a; aC_) o 00 H 1 00 Q o c CT3 4-) C o bO S o ICw T3 a; •o 3 >H O X
PAGE 16

construction, and housing showed the fastest growth rate per year (7.5, 6.3, and 5.9 percent, respectively) since 1976 (Table 1.2). Food production has lagged behind the population growth rate. Since most of the cultivable land is already being used, increased agricultural output will have to come from improved yields (World Bank, 1978, p. 14). Real income per capita in the Dominican Republic decreased from $489 in 1976 to $481 in 1979 (Banco Central, Dec. 1981, p. 199), and income distribution in 1976-77 showed that 50 percent of families with the lowest income receive only 18.5 percent of the total income. Income distributions for the country and for the urban and rural areas are depicted by the Lorenz curves presented in Figure 1.2. In 1976-77, 50 percent of the population had a per capita income below the poverty level, and the percentage was higher in the rural areas than in the cities (Quezada, 1981, p. 3). The decrease in income per capita along with the uneven distribution of income has worsened the nutritional conaition of the population. Consumption of calories per person decreased 9.2 percent from 1973 to 1977, and overall about 75 percent of the population consumed less food than is needed to fulfill the minimum nutritional requirements (ONAPLAN, 198ub, p. 2).

PAGE 17

10 Table 1.2. Gross domestic product in the Dominican Republic in 1976 and 1979, the change between these two years, and the equivalent compound rate of growth by sectors Value Shares Sector 1976 1979 1975 1979 millions of 1970 DR $ Percent Agriculture 429 .2 461 .7 17. 6 16. 8 Mining 146 .7 146. .5 6. 0 5. 3 Manufacturing 457 .4 504, .8 18. 6 ia. 4 Construction 153 .2 184, .0 6. 3 6. 7 Commerce 41 4 .0 455. .5 16. 9 16. 6 Transport 166 .7 195. ,8 6. 8 7. 1 Housing 156 .8 186. .0 6. 4 6. 8 Government Other ^' 189 .9 236. 1 7. 8 8. 6 329 .0 375, .7 13. 5 13. 7 Total 2,442.9 2,746.1 100.0 100.0

PAGE 18

n Table 1.2. Continued Growth 1976-1979 Sector RD$ millions Percent of Rate of Compound Rate of 1970 total growth growth of growth per annum (percent) Agriculture 32.5 10.7 7.6 2.5 Mining -0.2 Manufacturing 47.4 15.6 10.4 3.3 Construction 30.8 10.2 20.1 6.3 Commerce 41.5 13.7 10.0 3.3 Transport 29.1 9.1 17.5 5.5 Housing 29.2 9.6 18.6 5.9 Government 46.2 15.2 24.3 7.5 Other-S/ 46.7 15.4 14.2 4.5 Total 303.2 100.0 12.4 4.0 ^ Includes electricity, finance, communications and other services. Source: Banco Central de la Republica Domini cana. Boletin Mensual, Dec. 1981 Santo Domingo.

PAGE 19

12

PAGE 20

13 The pattern of development of the Dominican economy is characterized by the dominant role of the foreign sector. Agricultural products and some minerals are exported, and manufactures, oil, and capital goods are imported. In 1979, the value of exports represented 16 percent and imports 21 percent of the GDP (Banco Central, Dec. 19^1). Although the share declined from 91.2 percent in 1966 to 70.7 percent in 1976, agricultural exports still represent the bulk of Dominican exports. The share lost by agriculture has been gained by mineral products (Table 1.3). In 198u, for example, the four leading minerals (bauxite, ferronickel, gold, and silver) accounted for 39.4 percent of the total exports, and the four leading agricultural exports (sugar, coffee, cocoa, and tobacco and their by-products) accounted for more than 50 percent (Banco Central, Dec. 1981, p. 129). Intermediate goods have been dominant among imports (29.8 percent in 1974), followed by consumer goods (26.8 percent) and capital goods (23.9 percent). Fuels and lubricants showed the fastest growth rate among all imports — a 68.3 percent average annual growth from 1970 to 197U (Table 1.4) It is interesting to observe the increase of the imports of foodstuffs and crude oil from 1973 to 1977. The increase in crude oil imports was estimated at $145.68

PAGE 21

14 1 — (AJCOC— I^CMCM 1 CTi VO .1 C\J 1 0 1 fi • 1 a^ OO^OJOJ LTiLPiO c3^co f— vO o o vO 1 0 1 1 — CTi LTl • 1 0 V 1 — 0^ 1 o nLn=rc\ic\jc\joc\j 1 LTi — vO CM vO 1— 1 1 0 1 t— 1 • 1 • >> 1 (^^ CTv-3"X)C\lfY^oaNC3^0 o o CO 1 0 P 1 CO LTi =r • 1 0 •H C\J 1 "O o E S 1 in a\ 1 vO 1 0 O 1 1 • 1 • O 1 4-3 a^^o^-cM(^Joooo o C\J on CO o in 1 0 1 C CO LH ^ r— <— • 1 0 >. 1 00'— CMOOOOOO O vO o oo co o CM 1 0 3 CO in LO — • 1 0 Cl CM 1 t — (U o:: c 1 1 CM 1 1 0 CO vO 1 1 • o OOCMC\J(V^,— C-_t~_0 o \o o CO CO o CM 1 0 •r-{ c:7^ vO in • 1 0 c •H s o Q r— =T f— cTioo'— a^c^J 1 •=r CO 1 O LTi I in 1 0 1 • 1 • C3^ o m oo o OO 1 0 ON in • 1 0 o — E O X3 C+H CO sCO (U T3 s: "O O to c O CL CO CO i) cu O •o c O 0) CO CO o CL, CO OT o CO CO 4-3 4J 3 a; 4-3 CO 1 o o •a o 4J CJ 1 CO 3 3 o o o S•p4 1 4-3 X! "O s~ ex 0 4-3 1 o o o Q. CL, X CL ^ 1 3 SI, I — 03 1 X O 1 TD Q, Q, >> LU O 1 1 CO CQ to 0) CJ 1 L. ^ >> >i C 4-3 cc >> 1 x3 tn 1en CO (u a o 1-H 4J n 1 0 1—1 i-coco304->oox: 4-3 Cm CO cu 4-) CD l-i V X 0) O
PAGE 22

15 vo 1 c^-LTiLnooooooooonzri^oooi— r— LncMincoooio CI • CTi I ooLrif\J'— Of— vooo^-LOOt— c\j Lncgxit^onio i>-zT m r— c\j TI o w^^-^ www I rI I ^ ^ ^ ^ ^ ^ ^ I LO I CMLnc~-o>oc\jc\joo'— rocoocvjino zrcri^— =rio I I • ON I T— o.'njcmt— orocMoo^^roocM vO'— '-oncxjio CO ^ -O TI o wwww www I rI I ^^^^ I 1-p oc~-aNrooLnLria>c\j^cMC\ic\j^aD vocomdcm^t io c I • CTi I 0) O^O'— C\JOf— v000>-v£)0'— C\J >-C\I3-OC\l IO o oo LTi un I o t. wwww www I , I LH^CXJ'— CMOLTi^OrOvOOCMrO C\JP000OC\JlO f^ ^ C\J T— I o wwww www 1 ,I I ^^^^ I c\J I T-c^criiriT-c\jrovOr-voMDonmcriO a^onL^'— oio I I • CTv I oDOLnCMCMOLn^oooDooO'— on c^rrooo^io fLTi •=r I o -o E CO E T3 0) C •H -!-> c o o on a; I— I H CO er Q. 0) OS > 1 UJ 1 c o 1 1 CO CQ CO 0) 03 i. >> c OS >> 1 CQO CU a CQ iCQ 00 1) a — 1 p 1 o i3 t, 1— 1 x: 03 O O x: 1) o C iO -H o s~ O 1 4-> 1—1 cc 03 O jj o O x: -p Cm 03 ra •p
PAGE 23

16 Table 1.4. Imports into the Dominican Republic by major categories, 1 967-74 1967 1968 1969 1970 1971 Total Import Goods^ millions of US $ 171.3 200.8 221 .9 295.0 340.1 Consumer Goods 48.6 62.8 59.8 76.9 88.7 Food 31.5 44.8 38.5 44.3 53.6 Other 17.1 18.0 21.3 32.6 35.1 Fuels and Lubricants 12.8 13.8 18.1 19.3 24.8 Intermediate Goods 59.8 71.3 77.0 104.6 113.0 Capital Goods 50.1 52.9 67.0 94.2 113.6 percent of total import goods Total Import Goods 100.0 100.0 100.0 100.0 100.0 Consumer Goods 28.4 31.3 27.0 26.1 26.1 Food 18.4 22.3 17.4 15.0 15.8 Other 10.0 9.0 9.6 11.1 10.3 Fuels and Lubricants 7.5 6.9 8.1 6.5 7.3 Intermediate Goods 34.9 35.5 34.7 35.5 33.2 Capital Goods 29.2 26.3 30.2 31.9 33.4

PAGE 24

17 Table 1.4. Continued 1972 1973 1974 Average Annual Growth 70-74 % mill ions of US $ Total Import Goods-^ 382.0 448.5 796.4 2a. 2 Consumer Goods 104.3 124.9 213.6 29.1 Food 59.2 74.2 133.7 31.8 Other 45.1 50.7 79.9 25.1 Fuels and Lubricants 46.7 48.5 155.0 68.3 Intermediate Goods 120.3 145.9 237.5 22.8 Capital Goods 110.7 129.2 190.3 19.2 percent of total : import goods Total Import Goods 100.0 100.0 100.0 Consumer Goods 27.3 27.9 26.8 Food 15.5 16.6 16.8 Other 11.8 11.3 10.0 Fuels and Lubricants 12.2 10.8 19.5 Intermediate Goods 31.5 32.5 29.8 Capital Goods 29.0 2.8 23.9 V Based on UN Trade Publications. Includes OECD countries and Venezuela. This source is adjusted for petroleum, transport vehicles and wood. These totals will not necessarily coincide with total imports used in balance of payments which are based on Direction of Trade imports fron trading partners. Source: World Bank. Dominican Republic—Its Main Economic Development Problems. Washington D. C. 1978.

PAGE 25

18 millions. For the same period, the increase in foodstuff imported was estimated at $40.43 millions (World Bank, 1978, p. 146). These two items added considerably to the trade deficit of $557.8 millions in 1 980 — 10 percent of the GDP (Banco Central, Dec. 1981, p. 129). Between 1978 and 1980 prices were inflationary. The consumer price index increased 7.1 percent in 1978, 9.2 percent in 1979, and 16.7 percent in 1980 (Banco Central, Dec. 1981, p. 155). After a jump in the rate of inflation between 1979-80, prices were still increasing in 1981 but at a more moderate rate. During the 12-month period from November I98O to November 1981 the consumer price index increased 6.3 percent and this slower rate continued through 1982. The impact of the increase in prices has been different in the rural and urban areas. In 1980, the urban price index increased by 18.4 percent, that is, 1.6 percent more than the national increase. The rural price index increased by 15.4 (Banco Central, Dec. 1981, p. 155). Looking at the increase in the consumer price index by income strata, it is observed that in 1980 the index for the $300 to $400 income stratum showed the smallest increment (13.8 percent). The index for all other strata (above and below) grew relatively equally.

PAGE 26

19 The change in the comsumer price index by groups of commodities was 15.4 percent for food, alcohol, and tobacco; 10.2 percent for housing; 20.4 percent for clothing; and 31.8 percent for others (Banco Central, Dec. 1981, p. 157). Plan Disgertatiop Chapter II is used to present the traditional uses of household budget surveys and their applications to estimating Engel curves. This chapter is also used to review the literature within the theoretical framework applied in this study. This review included consumer theory and demand analysis, and applications of cross-sectional data in demand estimations. The methodology applied in this study is described in Chapter III, including the use of the restrictions of consumer theory, the models applied, and the estimation procedure. Steps followed to calculate the elasticities matrix are also explained in this chapter. In the last part, the method used to calculate the effect of prices an income changes on calorie and protein intake is discussed. The coefficients estimated are discussed in Chapter IV along with the presentation and discussion of the full matrix of demand elasticities and family-size elasticities. A final comment is made about the results of the new method of estimating Engel elasticities.

PAGE 27

20 The implications for policy are discussed in Chapter V, showing the effects of changes in prices and income on the calorie and protein intake. The last part of this chapter presents the implications of changes in income and prices for consumption and import goals. Finally, Chapter VI gives the summary and conclusions. Limitations of this work are identified and some suggestions for further research are offered.

PAGE 28

CHAPTER II THEORETICAL FRAMEWORK AND REVIEW OF LITERATURE In applied demand analysis there has been concern about the theorical plausibility of empirical demand estimates. One way to proceed in order to ensure the theoretical validity of the estimated demand model is to specify a utility function. The demand functions to be estimated can then be derived from the first order conditions obtained when maximizing the utility function subject to the budget constraint of the consumer. A disadvantage of this approach is the requirement of specifying well-behaved utility functions and the derived demand equations which result may be too complicated to be estimated due to the number of parameters or to nonlinearities. The other approach is more pragmatic and leads to the direct specification of demand functions with causal variables that are believed to capture the behavior of the consumer. In reviewing the literature, an attempt was made to fill the gap between the theory of consumer behavior and empirical demand analysis. Emphasis was upon those studies 21

PAGE 29

22 using cross sectional data to estimate Engel curves and income and price elasticities of demand. Household Budget Surveys Household budget surveys have been widely used to estimate income-consumption relationships and some attempts have been made to estimate price elasticities of demand. This kind of survey provides data indicating all expenditures on consumer goods and services made by individual families. A basic objective of the budget survey is to obtain the percentage budget share of all consumption items to be used as weights in the official cost of living index (Phlips 1974, p. 101-102). These cross-sectional data can be used to estimate Engel curves, which express the expenditure on a good as a function of income, i.e., E^= p^-q^= f^(y) ; i=1 ,n where p represents price of good i; q represents quantity of good i; y represents income and n is number of goods. The fact that prices are (almost) constant is used to isolate the influence of income and the simultaneous variation of both prices and income encountered in time series analysis is avoided. However a cross section does not automatically satisfy the assumption of a single given utility function, because the preference ordering may change from family to family (Phlips 1974, p. 103). To overcome

PAGE 30

23 this problem it is necessary to take a collection of families that have different incomes which are as homogeneus as possible with respect to such things as social class and profession. Phlips adds that to ensure the theoretical plausibility of estimated Engel curves the general restrictions of demand theory must be satisfied. All of the price derivative restrictions of demand theory (i.e., homogeneity, symmetry, negativity of the own substitution effect) disappear because prices are constant (Phlips 1974, p. 105). The only restriction that remains is the adding-up condition, i.e., that 5:f^-(y) = y or similarly E (dE ./dy) = 1 the sum of the marginal propensities to consume (or the marginal budget shares) has to be equal to one at all income levels, Engel Curves The first attempt to analyze cross-sectional data based on a theoretical model was the work by Allen and Bowley (1935). They estimated linear equations which automatically satisfy the adding-up restriction but were

PAGE 31

24 very restrictive in term of the underlying utility function that was implied. In addition, the implications of linearity in terras of the income elasticities were unsatisfactory from an economic point of view (luxury goods resulted with negative dE^. /dy), and the statistical fit was very poor (Phlips, 1974, p. 109). Pollak (1971) thoroughly analyzed the implications of linear Engel curves. He defined a demand system by {h^ ( p, u) h'^ ( p, u) } where p represents the price vector (p-j, p^) and n u = E E = y 1 = 1 He characterized the demand functions in the following way: if the demand functions are of the form h''(p,u)rx^. (p) + z^.(p)u in some region of the price-income space, the demand functions are locally linear in income; if h\p,u) =z-(p)u

PAGE 32

25 the demand functions exhibit expenditure proportionality, this is a especial case of linearity. The demand functions are linear in income if and only if the marginal budget shares of the ith good (p,u) = (3p^-h^. (p,u))/ 3u are functions of prices alone, independent of income. The income-consumption curves are straight lines in a region of the commodity space if and only if the demand functions are linear; they are straight lines radiating from the origin if and only if the demand functions exhibit expenditure proportional ity Pollak proved a theorem stating that additive utility functions yield demand functions which are locally linear in income if and only if they are of the form ) > 0, Z a^ =1 ; U r^Z^a^ log( Qk ), a >0, (q. ^ -^k n ^ : or n E k = 1 U = -Ea^e"^k^k ,a, ,B >0.

PAGE 33

26 He pointed out that in demand analysis the assumption that all income elasticities are unity has little merit and that the additivity assumption is defensible if the arguments of the utility function are taken to be broad aggregates of goods rather than individual commodities. Prais and Houthakker (1955) analyzed non-linear functions to estimate Engel curves more adapted to income consumption patterns. In fitting income-consumption equations they took into account a priori economic as well as statistical considerations. They incorporated two properties in the algebraic formulations: a) that there is an initial income below which a commodity is not purchased, and b) that there is a satiety level, that is, a maximum to the quantity of the commodity consumed which is not exceeded however high income may rise. Based on these considerations they chose the following models: log E-j = a^ + b-jlog y; double-logarithmic, log E-j r a-j (b-j /y); log reciprocal E-j = a-j + b-jlog y; semi-logarithmic E-j = a-j (b-j/y); hyperbolic.

PAGE 34

27 Prais and Houthakker concluded that none of the functions were uniformly satisfactory. The doublelogarithmic form gave fairly satisfactory results for most commodities except for the difficulty of treating zero expenditures. The semi-logarithmic form gave a satisfactory description of expenditures on most foodstuffs, and all non-linear forms proved to give better estimates than that given by the linear form. With respect to the work of Prais and Houthakker, Phlips pointed out that, while much has been gained with this pragmatic approach in terms of descriptive power, much has been lost in terms of theoretical plausibility. Non-linear specifications are not compatible with utility maximization, as they do not satisfy the adding-up criterion. He makes the same criticism of the work of Aitchison and Brown (1957), Wold and Jureen (195i) and Champernowne (1969). Laser (1963) investigated the properties of various forms of Engel functions satisfying the additivity criterion using nine commodity groups. He took into consideration several factors in choosing among various functional forms. These factors included the close connection with a direct or indirect utility function, the validity of the function for all positive values of total outlay within a wide range, and the plausibility of the income elasticities besides the

PAGE 35

28 statistical goodness of fit. He concluded that the equations used by Working (1943) of the form w-j = a^ + b-j log y + e^where w-j = E-j/y describe a behavior which would seem acceptable, and that the goodness of fit criterion tends to reinforce the theoretical considerations in favor of this model The above model was rejected by Prais (1953). In that work he comments that Stuvel and James in their study of households expenditure on food in Holland showed that neither the linear nor the exponential forms were appropriate for the whole range of budgets in that collection. In his conclusion, Prais suggests that a typical Engel curve may be described by the equation ( E-j/N )' = a + b log ( y/N ) where N is household size in equivalent adults. The above equation means that the income elasticity is inversely proportional to the expenditure on the commodity. To specify the behavior of the Engel curve for quantities he takes into account the change in quality measured by the average price per physical unit of the commodity and describes it by the equation p = h + k log ( y/N )

PAGE 36

29 where p is average price paid per unit of commodity. Dividing expenditures by price he obtained the quantity consumed q a + b log (y/N) I.e. N h + k log (y/N) q B + log (y/N) n C + log (y/N) where A, B, and C are constants respectively equal to b/k, a/b, and h/k (Prais, 195i, p. 101-102). Prais pointed out that this equation for the Engel curve for quantities seems very satisfactory, giving an asymptote equal to A, and consumption falling to zero at a positive income. M Alternative Method M Estimating Eneel Curves A new method of estimating Engel elasticities was proposed by Kakwani (1978). The new equation of the Engel function was based on a coordinate system for the Lorenz curve developed by Kakwani and Podder (1976). The equation of the Lorenz curve was fitted to the expenditure data and then the expenditure elasticities were derived in terms of

PAGE 37

30 the estimated parameters of the Lorenz curve for total expenditure and expenditure on each commodity. The income elasticities derived from the Engel functions investigated above were either constant or monotonically increasing or decreasing. The actual data indicated that the elasticities may not be either constant or raonotonic (Kakwani, 1978, p. 103). To estimate the Lorenz curve, Kakwani and Podder (1976) defined x as per capita total expenditure or income of a family; x is a random variable with mean M and the probability distribution function F(x). If F(x) is the proportion of families having income less than or equal to x, then Fn (x) = ( 1/M) ;^x g(x) dx 0 is the first moment distribution function, where g(x) is the density function of x. The Lorenz curve is the relationship between F(x) and F (x). This relationship is shown in Figure 2.1. Let P be any point on the curve with coordinates F and F Then, by introducing a new coordinate system (see Appendix 3), TT = (F + F )/^ and n = (F F)/f2~,

PAGE 38

31 Figure 2.1. Kakwani and Podder coordinates for estimating a Lorenz curve

PAGE 39

32 the equation of the Lorenz curve in terms of n and tt can be written as n= f( ). The derivatives of n with respect to are f ( TT ) = (M x)/(M + x) (1) and f"( ^ ) = -(2 V~rM2)/g(x)(M+x)2 < 0 (2) The model fitted was n= au^ (V2-. )^ a > 0, > o, and B > 0. Kakwani (1978, p. 104) denoted v.(x) as the Engel function of the ith commodity which has the mean per capita expenditure M ; then the function Fl[v.(x)] = (1/M.) dx is interpreted as the proportion of per capita expenditure on the ith commodity of the families having a total per capita expenditure less than or equal to x. The concentration curve of the ith commodity is given by the relationship F(x) and F^[v.(x)] (Kakwani 1978, p. 104; 1977 p. 720). In terms of n. and ^ the result is ^= (1/2)[F(x) + F^(v.(x))] n. = (1/2)[F(x) F^(v.(x))J

PAGE 40

33 The concentration curve can be written as n. =f .( tt ), and 111 the derivatives of n^with respect to tt^are M.v,-(x) (3) M • + V ( X ) 1 1 2 2V2M.v'.(x) -g(x)[M.+ v.(x)] ^ (4) Combining (1), (2), (3), and (4) the Engel elasticity of the ith commodity is obtained as Sv^Cx) X v;. (x)x f^-'(^.)[1 + f'(^)]2 [1 f'(^)J 9x v^(x) v^(x) f" ( w )[1 + fl (tt.)]2 [1 f. (^.)J (5) Kakwani (1978, p. 105) pointed out that if the data on different commodities are available in grouped form arranged according to total per capita expenditure or income, this equation can be used to compute expenditure elasticities at the end point of each expenditure class. To compute elasticities at any income level the following relationship can be used: 1T + fCir) = TT. + f (tx .) 1 1 1 The function fitted was the same assumed by Kakwani and Podder (1976)

PAGE 41

34 n. = a .TT^i (V2 ^. and was estimated using ordinary least squares method. Blaylock and Smallwood (19^2) applied the method discussed above to estimate expenditure elasticities among three urbanizations for several food groups. They introduced a generalized functional form to describe the Lorenz and concentration curves. The model developed was n^^)= A*+ a^^''^ + B*(V2 T,)(5)+u where a v 5 A* and B* are parameters and u is the error term. This equation approaches the function used by Kakwani and Podder as all transformation parameters approach zero. The equation was estimated using maximum likelihood technique since it is nonlinear in the parameters. Blaylock and Smallwood concluded that, as income increases, elasticity patterns can be nonmonotonic and the pattern found in their analysis could not be obtained using traditional methods of analysis. This added flexibility for estimating Engel parameters is important for policy analysis involving the impact of income transfer programs on population subgroups located at extremes of the income distribution (Blaylock and Smallwood, 1982, p. 139).

PAGE 42

35 Application of the Frish Method In an effort to compute all direct and cross price elasticities, Frisch (1959) made the assumption of want independence and suggested a method of obtaining the price elasticities with knowledge of the budget shares, the income elasticities, and the flexibility of the marginal utility of money. In his work, he defined marginal utilities as a function of the quantities u-j = u-j(q],...,qn) ;i = 1,2, ...,n and the inverse function as q^ = qi (u-| u^ ) ; i = 1 2 n ; he also defined i = 1 2 n i = 1 2 n as the utility acceleration which is the elasticity of the marginal utility with respect to q |^ And he defined 3u i ( q 1 q ^ ) 3q u.(q, ...,q ) ; 1 1 n q = — — — — = Q ""^ 3U, q (u . u) ^ 1^ k 11 n as the want elasticity. This elasticity tells the proportionate change in the want for a good with respect to the change in quantity of other goods. Frisch defined good

PAGE 43

36 i as want independent of all other goods if q ^ |^ = 0 for all k^i, or equivalently q^_^ = 0 for all k ^ i. Another equivalent definition he gave was that good i is want independent of all other goods if u^.|^ = 0 or Uj^^ = 0 for all k ^i; u^-|^ = 0 means that the marginal utility of good i depends only on the quantity of good i and not on any other quantity. These concepts can be applied to broad aggregates; want independence willnot hold with single commodities. The marginal utility of money is 8U/3y. From the first order conditions of constrained utility maximization X = (u. )/(p. ) and the elasticity of the marginal utility of money with respect to income is ( 3A / 3y) (y/ a) = and is called the flexibility of the marginal utility of money or money flexibility. Using these results Frish developed the relationship for the price elasticity as where e-jis the cross price elasticity of good i with respect to the price of good j; w. is the budget share of good j; e^-^ is the income elasticity of good i; and ^ and a..

PAGE 44

37 are as defined above. This relationship is equivalent to the classical Slutsky relation e and =e..(w./w.)+w.(e. -e. ) i j = 1,2, ,n which can be compared with the classical relation that states the restriction that Z e • • = -e (homogeneity restriction) Now if a good i is want independent of all other goods, a0 for all i j and the cross price and income elasticities can be simplified to the relation 1 J elasticity ) e so thus ly 1 1 '1 1 1 i Wje^.^ (1/ *)Wje.ye.^ = 1 1 11 e^-yWj[1 + (ej^/(t))] (cross price = '^^^^ > (income elasticity) (ei/ ) ii"i^iy ^T/* ^"i^-y^-y (eiy/ (}) ) w^-e^-y (1/ )^.e.^e^y w^-e^-y + e^-y[ ( 1-w^-e^-y)/ $ ] Siytw^( 1-w^-e^-^)/ ^ )] from which <^ can be obtained, i. e. ^e^..+

PAGE 45

38 (e.. +e. w.) = e. w.e. e. e w e e ly 1 iy iy ^ = (George and King, 1971, p. 22-23). The disadvantages of using this approach are that want independence imposes complete additivity in the utility function (which implies that the marginal utilities of i and j are unaffected), and that the a-, are not invariant for 3 any monotonic transformation of the utility function. Frisch pointed out that his development can be extended to the case where there are groups of commodities in which there may be dependence within each group, but want independence between a good in a given group and all the goods in the other groups. In this case, in addition to the budget share, w, the income elasticity, e-^, and the money flexibility, <}) the cross price elasticities for the commodities within the groups are needed. George and King (1971) made a full application of this method, estimating a demand system for food commodities in the United States. They used most of the theoretical restrictions on demand systems and made full application of the Frish relations developed above. In order to apply the Frish method, they imposed the assumption of want independence for all commodities outside

PAGE 46

39 a given group to obtain all the cross price elasticities of commodities between different groups. The direct and cross price elasticities of commodities within each group were estimated directly using time-series data. They estimated a demand system for 49 commodities grouped in 15 separable groups. The demand equations were specified with quantities as dependent variables. The independent variables included income, the price of all commodities belonging to the same group, and the price indices of other groups. To obtain the income elasticities from cross-sectional data, they used both ordinary and weighted regressions. The regression equation obtained was log q = a + b log y where q is per capita .quantity consumed and y is per capita income. They also evaluated the effect of household size on consumption in the cross-sectional analysis. Expenditure and quantity were used as dependent variables. The regression equations were of the form log q = a + b log y + c log N and log E = a' + b' log y + c' log N where E represents expenditures, and N household size. The quality elasticity was also analyzed, with and without household size. This measure was obtained from the

PAGE 47

40 decomposition of the expenditure elasticity into quantity and quality elasticities. Combining the coefficients calculated in the ways mentioned above with the restrictions of the theory of consumer behavior, George and King calculated all the remaining coefficients. Price Elasticities From Cross-Sectional Data Several attempts have been made to estimate price elasticities from household budgetary data. The approaches to this problem can be classified as three kinds. First, estimates of direct and cross price elasticities have been obtained making strong assumptions about the underlying utility function. The second approach has been based on pooling cross-sectional and time series data to overcome the restriction on the number of observations that usually exists when using only time series. The third kind is aimed at capturing the price variations during the time period covered by the household budget surveys, which in most cases requires enough time to observe some price variation. Friedman (1935) discussed a method suggested by Pigou, as early as 1910, of using budgetary data to obtain the ratio of the elasticity of demand for one commodity to that for another. Pigou' s method is based upon the assumptions that the marginal utility of a commodity to an individual is a function solely of the quantity of that commodity

PAGE 48

41 possessed, and that the marginal utility curve for a commodity is the same for each individual. He then makes the utility elasticity equal to the price elasticity. Friedman shows that that assumption would hold only if there is a very large number of commodities purchased or if at least one commodity has a very large utility elasticity. Leser (1941-42) reviewed the assumptions that have to be made in order to estimate price elasticities from budget data. He pointed out that the assumption made by Marshack — that the difference between the income and price elasticities of demand for a commodity is negligible if the amount spent on it is small compared with total expenditure --was rejected by Frisch who showed that the observed income elasticities can not be interpreted as price elasticities. In his work to obtain price elasticities, Leser assumed that the cross elasticities of demand depended on the nature of the good whose price changes, and not on the nature of the good for which the effect was studied. This assumption excluded a high degree of substitution or complementarity, and thus would be valid only for broad aggregates and not for single commodities. Hassan and Johnson (1976) estimated consumption functions and Engel curves based on cross-sectional data from a family budget in Canada. They also examined hypotheses of possible structural change related to six

PAGE 49

42 socioeconomic characteristics. These included city or urban community in which the family resided, class of tenure, age of family head, family life cycle, family income, and education of family head. In this study the Engel curve was approximated with a linear function. To justify this functional form they pointed out that the aggregate level of the commodity groups should not involve a substantive misspecif ication error; that is, the commodities as they were grouped were neither luxuries nor inferior goods. In a new study using cross-sectional data from an urban family food expenditure survey in Canada, Hassan and Johnson (1977) estimated demand parameters for 122 food items and computed, from these estimates, elasticities for expenditure, income, quality, family size, and price. The price elasticities were computed by using implicit prices derived from the expenditure and quantity data. This time they fitted a semi-logarithmic function to approximate the Engel functions, which implied that elasticities vary inversely with the level of consumption. Among the reasons they gave for choosing this functional form were that the serai-logarithmic function gives a satisfactory description of expenditures for most food commodities; that this form allowed observations of zero for the dependent variable (which occurs frequently when raicrodata are used); and that even though this function does not satisfy the adding-up

PAGE 50

43 criterion, it provides an appropriate representation of Engel curves for the range in which the income elasticity is less than one. To estimate the price elasticities, Hassan and Johnson (1y77) used the price variations within the period over which the survey data were collected. In the survey which provided the source of their data the period was one year. They pointed out that during such a period, price variations are likely to be small and statistical measurraent cannot be overly ambitious. So the estimated models must include a large amount of prior information. They included only own price elasticities in the demand functions. They concluded that the price variations that occurred within the sample period were sufficient for estimating the coefficient necessary to compute the elasticities, and that the estimated parameters were consistent with theory and with available elasticity estimates from time series data. Dgmapd Estimates In the Dominican Republic A few attempts have been made in the Dominican Republic to estimate income and price elasticities. All of them faced the problem of inadequate data; time series data did not exist at all for some crops or food commodities and when they existed they were very inconsistent (USDA, 1977, p. 103). Cross-sectional data first became available in

PAGE 51

44 1971, from a household income and expenditure survey (Banco Central, 1971), but only for the city of Santo Domingo. Later, in 1976-77 the Central Bank conducted a new household income and expenditure survey (this time at a national level), whose final results were available in May 1982. Erichson, House, and Nunez estimated income and price elasticities for several crops (USDA, 1971). In their work, they rejected the use of time series because of the lack of meaningful data which led to very poor results. They used the Frish method (discussed earlier) to estimate price elasticities. Estimates of expenditure elasticities and budget shares were derived from the 1969 household income and expenditure survey. The money flexibility coefficient was taken from cross-country studies done by de Janvry, Bieri and Nunez (1972), Lluch and Williams (1975), and Le-Si and Pomareda (1976). In this way they calculated direct price elasticities for nine crops. Another study using data from the 1969 household survey was done by ONAPLAN to estimate gross income elasticities between two income groups (see discussion on pages 2 and 3) In a preliminary study using data from the 1976-77 household survey, Del Rosario and Musgrove (1980) estimated consumption functions for 13 groups comprised of 67 single commodities. They worked with data grouped by 10 income

PAGE 52

45 classes. The expenditure on each commodity, E ^, and the quantity of each commodity, q^, were regressed separately against total expenditure, y. The models fitted were as follows : log q^= a^-+ b^/y + c^log y and log E-j = a^ + b^/y + c^-log y where a^ = a-j + log p^ The authors suggested extension of their work by including more explanatory variables such as family size and geography location. They also suggested that this kind of estimation be done with single household data (Del Rosario and Musgrove, 1980, p. 28). Quezada (1981, p. 162, 219) estimated demand equations for 6 commodities using data from a 14-years time series on prices and expenditure. He fitted a linear model deflating all prices and consumption expenditures by the consumer price index. In the model he included per capita consumption of the product as a function of its own price and of prices of some substitutes or complements, and per capita consumption expenditures. T he Limited Dependent Variable Model When working with cross-sectional data from a household survey many of the observations on the' consumption of specific goods will be zero, since at the time of the

PAGE 53

46 survey the consumer did not buy that good, so the consumption variable is not observed over its entire range (Tobin, 195a p. 24). As described by Maddala (1977, p. 162), if q = bx + e and q is observed only if it is greater than zero, then q = bx + e if bx + e > 0 or e > -bx or q = 0, i.e., e = -bx otherwise. Ordinary least squares cannot be used to estimate the regression equations using only the observations for which q> 0 because the residuals do not satisfy the assumption that E(e) = 0, if only those residuals for which e > -bx were used. Making some specific assumption about the distribution of the error term, e, the maximum likelihood method can be used to estimate the parameters. If, for instance, it was assumed that e has a normal distribution, with mean zero and variance a the joint distribution of the observations is L = ^1 f ( Cq bx )/ a ) tt F(-bx ./a) I 2 where f((q.j bx^)/a ) is the standard normal density, and represents the observations for which q. was observed; and

PAGE 54

47 F(-bx-/o) is the curamulative normal density, and represents the observations at the limit, i.e. where e = -bx when the limit is 0. The ML estimates are obtained by maximizing L with respect to b and a. (Maddala, 1 977, p. 136). Tobin (1958) pointed out that if only the probability of limit and nonlirait responses were to be estimated, the probit analysis would be appropriate, but that it would be inefficient to throw away information on the value of the dependent variable when it is available. So he developed what he called a hybrid of probit analysis and multiple regression, that was later to be known as the tobit model.

PAGE 55

CHAPTER III METHODOLOGY Thg Rggt-r ict iong Consumer Tbgpr y In this study, a demand system for food commodities was estimated using the general restrictions of consumer theory. The restrictions also serve to reduce the number parameters to be estimated. Some price elasticities were estimated directly. In such cases the variation in prices over the period in which the data were collected was used. The general restrictions on demand coefficients comprise the following: le = -e (Homogenity Condition! i, j = 1 n Zw. e_.^^ = 1 (Engel Aggregation) i= 1 n 1 e .. = (w-./w.)e^.. w. (e.^ e^.^ ) (Symmetry Relation ) i j = 1 n ^w. e. = -w. (Cournot Aggregation) 48

PAGE 56

49 i j = 1 n where e. represents the elasticity of good i with respect to the price of good j, e,^ represents the income elasticity of good i, w. = p. q./zp. q. is the budget share of good i, and n is the number of goods. The Demand Models A semi-logarithmic model was hypothesized for the demand functions. The equations were of the form Q^. = a*^^ +a^^ log p^ + + a^.^ log p^+b^ log y +c^. log m + u_. where q^ represents quantity consumed by the household of good i, for i=1,...,n; p^ represents prices of good j, for j=n-k and k=0,...,n-1; n is the number of commodities (some are single commodities and some represent groups of commodities); y represents the household total expenditure (which is also referred to as income in this study); m is the number of members in the family, adjusted for adult equivalents; u. is the error term; and a'-^ = log a^-^. There exists an extensive literature on the measurement of household size. Brown and Deaton (1972, p. 1178-1186) reviewed many of the proposed scales to adjust the size of the family for adult equivalent; they pointed out that such scales, including the nutritional scales, are

PAGE 57

50 based on normative judgments rather than on market behavior. Nevertheless, for this study it was considered that from a practical point of view it would be better to make some adjustment in the number of members, since all of them do not consume the same quantities of the different food commodities. The scale used was to consider any family member 14 years old or less as 0.5 of an adult equivalent. Using the proposed semilog models, the estimation procedure was as follows: the first equation had n prices as independent variables plus total expenditure and number of adult equivalent household members, the second equation had n-1 prices as independent variables plus total expenditure and number of adult equivalent household members, and so on up to the 30th equation which had only its own price along with total expenditure and number of members. The estimation procedure is shown in Table 3.1. A few exceptions were made to include the prices of some products that were known, a priori, to have an important effect on the respective demand equations, and that would not be included if the procedure had been followed. These exceptions were made for cassava, where the price of plantain was included; for other-meats, where the prices of beef and pork were included; for other-roots, where the price of cassava was included; for other-milk, where the price of fresh-milk was included; and for banana,

PAGE 58

51 CM + + e e o o — 1 — 1 CM o o 1 T — w bO o o — 1 — 1 CM X) + + o o CO CO o. o. bO o o 1— 1 i-H o O CO CO CM CD + + CM Q. Q. bO bO o o — 1 — ( CTi CM CM *\ CM CD CD • • • • • • + + CM CNJ Q. a bO bjO O. o CTi CM + S o i-H o + >. bO O rH > bO O rH o CO + O CO Q. bO O o CO CI o 00 CD + CD + CM CM CD + bO O (0 + o o o o o CM CM CO CD CD CD CD II II II II o CM CM CO cr • • • cr cr

PAGE 59

52 where the price of other-fruits was included. However, the coeificients of these prices were not considered in the calculations of the price elasticities; they were included just to lessen specification errors for the estimated parameters used in the calculations. (See estimated coefficients of regressions in Apendix 2.) M Jig To^i t Model To estimate the demand equations the tobit model was applied to the semilog forms for each single equation. The usefulness of this estimation method is twofold; it provides a way to use all observations including those which are concentrated at the zero limit value; it also allows the use of all the observations in the estimation of each equation; this was important, since otherwise each equation would have had a different number of observations and it would not have been possible to establish the relationships between the elasticities of different goods consistently with the same data set for each equation. To overcome the remaining problem of missing values in the prices of some goods in each equation, the average prices of the respective goods were used in place of the missing values (Maddala, 1977, p. 202). Since the purpose of the estimation of these equations was to establish a functional form for the demand functions

PAGE 60

53 and to obtain the respective regression coefficients, the observations were not weighted. A similar procedure was followed by Tobin (195b, p. 31) who stated that it may well be that the sample gives urabiased estimates of the parameters of the relationship, even though it gives biased estimates of the separate frequency distributions of the v ari abl es The estimations of the mean quantities consumed and of the mean prices, as well as that of the budget shares, were made using weighted observations. XtLg Elast ic it ie s Matrix The elasticities matrix was comprised of direct and cross price elasticities plus income elasticities for 30 food groups, the all food aggregate, and one nonfood group. The 30 food groups and the variable names used for each of them are listed in Figure 3.1. Since the estimated values of the elasticities would not be invariant to the ordering of the commodity groups within the matrix, the groups were arrayed in descendent order according to the budget shares. In this way the group, or commodity with the largest share of the consumer expenditure would be in the first row. As can be seen in Table 3.1, the item in the first row is the one with more coefficients estimated directly. The number of coefficients

PAGE 61

54 RIC Rice FAT Fat and oils BEF Beef FMK Fresh milk PTN Plantain RBN Red beans BEV Non-alcoholic beverages POL Poultry FDW Food away from home FVG Fresh vegetables and soups SPC Spices and miscelaneous POR Pork SUG Sugar OCE Other cereals FIS Fish BRE Bread CAS Cassav a OFT Other fruits OMT Other meats EGG Eggs ALC Alcoholic beverages ORT Other roots OMK Oter milk and milk products BAN Banana CHE Cheese OLG Other dried legumes CVG Canned vegetables and soups POT Potato NUT Nuts OBK Other bakery products ALL All food commodities NON Non-food commodities Figure 3.1. Definition of consumption groups

PAGE 62

55 estimated was decreased by one for each additional row down the matrix. The strategy was to attempt to minimize specification errors for the most important items in the budget. The elasticities were calculated using the coefficients estimated with the tobit model. Own price and cross price elasticities, e with ij respect to those prices which were included in the estimation equations (see Table 3.1) were calculated directly. The estimated regression coefficients were divided by the weighted mean of the dependent variable (the quantity consumed of each commodity or group) to obtain the estimated elasticities. To see the reason for using this procedure consider the semi-logarithmic model, k ^'io "^.^^ij y + c. log m where k = the number of prices included as independent variables in the ith equation. Here, aq. / 3 p = a. /p and e.. = (sq./sp ).(p /q. ) = (a., /p. ).(p./q.) = a. /q the income elasticity e. and the elasticity with respect to

PAGE 63

56 the number of members, e ^ were calculated following the same procedure, so e.y = b. /q. and e^^ = o^/q Nelson (1976j has derived elasticity estimates for the tobit model based on the expected quantities consumed given values of the independent variables (in this study price, income, and family size). His derivation leads to the formula ei= [BiPiFCBP/a)J/LB' PF(B' P/a) + af(B'P/a)] For the semi-logarithmic model specification used in this study the ?^ would cancel and the formula would reduce to e j_ = [B^F(B'P/a)]/[B'PF(B'P/a) + af(B'P/a)] ^ ^.i B' P + [af(B' P/a)/F(B' P/a)] This value for the elasticity is approximately equal to (ay/qi) for values of (B'P/a) over 2.5. There is some overestimation of the elasticities at the mean quantities versus expected quantities for purposes of national policy. However, the overestimation with the method used was minor. 1/

PAGE 64

57 The rest of the elasticities for the food commodities and groups were calculated using the symmetry restriction. For example, the element in the second row and first column, e2-] of the elasticity matrix in Table 4.2, was estimated as ^2,1 = (Wi/W2)ei^2 ^1 ^^2,y -l,y )• The rest of the elasticities in that row were obtained using the estimated coefficients as explained above. The elasticities of each food group with respect to the prices of the all-food aggregate, e-j^ were obtained as the sum of all the price elasticities in the ith row, or j The elasticity of the all-food aggregate with respect to the prices of each food group, e was calculated using the symmetry relation, from the price elasticities of the food groups with respect to all-food, e^.^ The own price elasticity of the all-food aggregate, e was obtained as the sum of all the price elasticities of the all-food aggregate with respect to the price of each food group, e^. or e^^ = ^fj' 3 The price elasticities of each food group with respect to the price of the non-food group, e. were calculated using the homogeneity restriction

PAGE 65

58 = -e,, but ? i j ,nf "^iy ze^j = e^.f then ],nf ~ "^iy "if The elasticities of the non-food group with respect to the price of each food group, e ,were calculated using the symmetry relation and the price elasticities of each food group with respect to the nonfood group, e^ The income elasticity of the all-food aggregate, e^^, was obtained as the weighted sum of the income elasticities of each food group, using the budget shares as weights, or e, = zw.e. /ew. = zw.e. /w^ i ^ ^ • 1 -ly f The income elasticity of the nonfood group was obtained using the Engel aggregation, or Ew.e., + w Je =1 i.e. .j 1 iy nf nf,y e„. = ( 1 E w e. )/w nf,y ^. 1 ly nf The cross price elasticity of the nonfood group with respect to the price of the all-food aggregate was calculated using the homogeneity restriction or

PAGE 66

59 E e^j = e^^ then J The cross price elasticity showing the effect of the all-food aggregate on the nonfood group, e^^ ^ was calculated from the symmetry relation based on e^ The own price elasticity for the nonfood group, nf,nf obtained from the homogeneity condition as Uig Effect ilf P rioe and income Changes on Calorie and Protein Intake The elasticity matrix was used to estimate the effect of changes in prices on the intake of calories and protein. The procedure, in general, was as follows: the contents of calories and protein in one pound of each commodity, cand p^, respectively, were estimated from tables on food composition from INCAP-ICNND (1961), ONAPLAN (igaob), and Whitney and Hamilton (1977); and the percentage change in the average consumption of each food commodity or food group, per capita, per day, q. when the price pchanged 1

PAGE 67

60 percent was estimated from the price elasticities, e. The 3 equation for estimating the change in consumption was AQ. = (e.. /100)q. 1J TJ 1 Thus, the estimated change in calories consumed from the good i when p. changed 1 percent was 3 AC.j = Aq.. .c. and the estimated total change in calories consumed from all food commodities, when p • changed 1 percent was J AC = Z Ac • • = A Cn • + A c o ij ; ^ij ""Ij ""2j ... + -^30,3 where i, j = 1 ,. ,30. The change in protein intake as a result of a change in price and changes in calories and protein intake as a result of a change in incomes were determined in an analogous way. In the. policy analysis some assumptions on price and income changes will be evaluated and compared with the goals of nutrition in regard with income changes and food imports.

PAGE 68

61 UOM Data Data from the "First National Survey on Households Income and Expenditure" (SHIE), collected by the Central Bank of the Dominican Republic and the National Office of Statistics, were used. The first objective of the survey was to obtain the information necessary to calculate the official cost of living index. The data were collected during the period from May 1, 1976, to April 30, 1977. They were based on a probabilistic sample by areas in multiple stages. The probability of one household being selected in the sample was 0.005 or one in each 200 households. The final sample used in this study was comprised of 4,027 households. The urban and rural distribution and a comparison with the 1970 census are shown in Table 3.2. In the data there were 477 consumption items, of which 222 were food. The 477 items were grouped into 30 food groups (some groups were single commodities) and one nonfood group. The 30 food groups are shown in Figure 3.1. (Consumption items in the food groups are given in Appendix 1.) In the equations estimated (Appendix 2) the logarithms of the prices used as independent variables were designated with the three letters given in figure 3.1 preceded by the letter L. For example, the logarithm of the price of fats and oils is designated LFAT in Appendix 2.

PAGE 69

62 4-5 C O s_ a. o O Cw To o t-on cvj =r cr> ^ ^ ON CTl o to t^on 4-3 C 0) o Sa> cu o on O X) O ^ LTi to -a o sz rH 0) CL CO E 3 CD O CO X Cm o tvO c\j CO o vo on CO (0 CD -P O H CD D

PAGE 70

63 1. For example, estimating the elasticities for rice at the expected quantities consumed led to the following estimates -.46, .01, -.12, -.22, -.05, .08, .01, -.00, -.07, .05, -.05, -.25, -.22, .05, -.00, -.01, -.07, -.07, -.07, .06, .02, -.06, .02, -.01, .03, -.00, .02, .15, -.04, -.00, .24, .56. When respectively compared to the first row of Table 4.2 it can be seen that the overestimation was slight.

PAGE 71

CHAPTER IV EMPIRICAL RESULTS Tobib Model Estimates The tobit model was used to estimate the demand equations following the procedure described in the methodological section. These coefficients and the t values associated with their implicit standard errors are shown in Appendix 2. The coefficient for the own price in each demand equation, except in the demand for canned-vegetables -and-soups (CVG), had the expected negative sign. This sign implies an inverse relationship between the price of a good and the quantity demanded of it. Also the absolute value of the coefficient of the own price in each demand equation, with the exception of CVG, was greater than the coefficients of all other prices. This result shows that, as expected, the effect of the own price is more important than the effect of other prices on the demand for a good. The coerficient of total expenditures for all the commodities had a positive sign implying that there were no inferior goods among the food groups. 64

PAGE 72

65 The t statistics reported in Appendix 2 clearly indicate that the elasticities of demand with respect to own price and the elasticities of demand with respect to income were estimated most accurately relative to their standard errors. With cross-sectional survey data the variation in income is greater than in prices and it would be expected a greater accuracy in measuring income effects. The high t ratios for own price effects suggest that there was enough price variation to consistently measure price effects. The t ratios for the number of household members was also in general indicative of an effect of household size on the purchases of food items. The coefficients showing the effect of the size of the family on the quantities consumed of food commodities, in general, had positive signs. The exceptions were a few commodities (BEF, FDW, OMT, ALC, OFT, OMK, CHE, and OBK) which might be considered luxury goods within the food groups due to their high prices (see Table 4.1). This conclusion is reinforced by the estimated income elasticities for those commodities, which were all greater than one. The point is that when the family size increases with a given income, cheaper foods are substituted for more expensive foods. All the coefficients from the regression in one demand equation were divided by the mean quantities consumed of the

PAGE 73

66 Table 4.1. Average expenditure proportions, quantities consumed, and weighted average prices paid by households in the "First National Survey on Household Income and Expenditure", by food group Food Group Expenditure Proportions Average Quantity Consumed Weightei Average pounds/month $/pounds RIC 0.0728 63.9910 0.25 FAT 0.0490 12.0711 0.97 BEF 0.0303 8.3404 0.78 FMK 0.0257 4.541 1 1 .25 PTN 0.0256 47.241 9 0.14 RBN 0.0233 12.4066 0.43 BEV 0.0228 13.2299 0.79 POL 0.0224 7.9530 0.64 FDW 0.0209 1 .9729 2.89 FVG 0.0200 20.7630 0.30 SPG 0.0176 19.6044 0.24 PGR 0.0168 5.0203 0.77 SUG 0.0160 20.3194 0.18 OCE 0.01 45 14.2361 0.32 FIS 0.0137 3.7067 1 .02 BRE 0.0137 9.5707 0.40 CAS 0 01^5 J J J J J J n DQ OFT 0.0119 16.5314 0.25 GMT 0.0107 2.2902 1 .04 EGG 0.0092 30.6472 0.07 ALC 0.0089 1 .3867 7.00 ORT 0.0081 24.3326 0.14 OMK 0.0075 1 .9638 1 .29 BAN 0.0067 15.6197 0.11 CHE 0.0052 0.8674 1.31 OLG 0.0044 3.5431 0.34 CVG 0.0025 1 .0745 1 .01 POT 0 .0023 3.5716 0.15 NUT 0.001 9 3.8800 0.20 OBK 0.001 9 0.2641 5.09

PAGE 74

67 respective commodity (see Table 4.1) to obtain the price, income, and family size elasticities in the diagonal and the upper off diagonal part of the price elasticity matrix. TJi£ Full Ma t ri x D emand E last i c i t i es The calculated elasticity matrix (Table 4.2) had 900 price elasticities plus 30 income elasticities for the 30 food groups, 32 price elasticities plus one income elasticity for the all-food aggregate, and 32 price elasticities plus one income elasticity for the nonfood group, for a total of 996 prices and income elasticities. The 32 groups and their variable names were given in Figure 3.1. The 30 family-size elasticities will be discussed in a separate section. Applying the general restrictions of the consumer theory, only 495 elasticities or (n ^ + n) ( (n ^ + n + 2)/2) where n is equal to 31 (30 food groups plus one nonfood group) were calculated directly making use of the estimated regression coeif icients. The rest were derived as explained in Chapter III. All the coefficients presented in the elasticity matrix in Table 4.2 were the result of the applied

PAGE 75

68 Table 4.2. Elasticity matrix for food and nonfood commodities RIC FAT — BEF — FMK — PTN RBN BEV— RIC -.^68678 .0150396 -. 121 812 -.224391 -.048324 .076431 1 .0090591 FAT .0013273 -1 .37831 .0399466 -.021 1 91 .0139341 .1068585 -.010728 BEF -.391032 .0125062 -1 .70453 .3880998 .2422791 .4621612 .0178289 FMK -.704585 -.072877 .4695052 -1 .73966 .0900230 .1765227 -.041158 Pra -.1 90008 .0053786 .3057456 .0962717 -1 .62926 .4988961 -.035854 RBN .2329542 .2348904 .6392313 .2171300 .5644442 -1 .28724 .01101 83 BEV -.001672 -.029456 .051 9074 -.032684 -.032460 .0032815 -.7741 1 3 POL -.079073 .0521671 -.210296 -.017451 -.260845 .1398433 -. 1 41376 FDW -.654877 -.48941 1 -.397863 -.1 1 9285 -.193705 -.1 97665 -.133329 FVG .1568708 .0777268 .1764044 -.038167 -.054107 .0351349 -.130922 SPC -.249605 .0030094 .29621 88 .1424885 -.022344 -.132987 .0255896 POR -1 .20349 -.345123 .5686421 .3979459 -.513988 -.424806 .0943125 SUG -1 .02641 -.443872 .7838838 .2092158 .1700612 -.476323 .2366396 OCE .2106366 .01 98445 -.021 178 .2769948 -.345445 -.003008 .0291 1 95 FIS -.050510 .0570831 .3696716 .3022375 -.136728 .0887535 .0381 152 3RE -.089414 -.022468 -. 10691 9 -.131639 .00526 85 .0421 949 -.003413 CAS -.408111 -.41 9758 .8980362 .5091213 .527383i< -.239844 -.159968 OFT -.504434 -.196100 .0771819 .0442710 -.377728 -.06 9468 -.059093 OMT -.621277 .1314742 .6459895 -.025551 -.520553 -.516326 -.165390 EGG .4077358 .iJ468458 1 .595364 -.401605 -.099060 .ii 926756 .0705563 ALC -.521719 -.680581 .0814897 -.096 825 -.21 1606 -.1 14294 -.585757 ORT -.574183 -.073812 .2093247 .5150462 -.324768 -.195058 .12391 40 OMK .0178859 -.239134 -.415854 .5440894 -.378302 .0174882 -.206608 BAN -.147904 -.402379 1 .614299 .4270427 -.263817 -.107713 -.159442 CHE .1107597 -1 .39507 -2.17848 -1 .23576 -1.83773 -.924305 1 .762128 OLG -.086048 1 .646997 4.586 966 3.544175 -.790643 .0617924 -.067987 CVG .5123 973 1.360553 -.648218 .0240597 1.325407 .8198891 -.039534 POT 4.905634 4.761 158 -2.64277 -.341124 5.547962 1 .557051 -.425352 NUT -1 .73880 2.475468 1 .651586 .6731463 -2.51 944 -1 .45126 -.245286 OBK -.406634 .021J7874 1.607111 -.062502 -.639433 .1291453 -.068498 ALL -.263497 -.154201 .1014618 -.00 9737 -.125162 -.026405 -.048393 NCN -.008912 .0328441 -.038649 .3928913 .0272895 1391381 .0378603

PAGE 76

69 Table 4.2. Continued POL-— r Dvi FVG SPG POR — SUG — OCE DTP -.074039 .049^003 A C )( ^ 0 ^ .05^001 -.257391 .222875 .0474739 TT AT no Qn Q • yo oy o 1 Ul 0 DO A A C A A 0323 500 A A -1 C A Q A .001 59o9 1 A ll A 1 0£:425 -. 1 4671 4 .0072156 -. 103077 1 o9d 92 A AC £ T A •! .0950791 153o7^1 A i 1 1 A 0 ^ A .3143062 .3959892 024 1 96 f nJS. 0 i h33d -.003545 A 1 1 A ^ A A -.0421 93 A 0 ^ A T A h 0 06 27 94 .2657309 1 1 82323 1 478950 r ill -.222641 -.050000 -.050174 .022455 -.3291 82 .0979258 -.200498 1 55 1 0 f 4 uo 5o 1 0 .0349733 -.095038 2 87 1 1 3 -.325544 .0025309 BaV -. 126207 -.017574 -. 1 16448 0 1 7 92 1 5 .0835985 .1625031 .0179366 rUL -1 .72d'i0 -. o5949 A A ^ ^ 1 1 0 -.0911 48 1 170500 .3168737 .04381 99 -.00731 8 TNT T 245-^1 4 -6,o491 9 -.584827 -.1 96007 2.61 8950 1 .893060 -.129860 FVG -.087474 -.506591 -1 .30310 -.02791 5 1 904397 .1 900881 -.049925 o n SPG 16 46086 -.126602 -.03161 1 -.567362 -.121687 -.036109 .0121044 POR 4 165246 3.342894 .2071881 -.143399 -3.16503 -.0731 82 -.368103 QTTP 079331 5 2 5 o2 0l 5 A A A It A A ^ .23 9403 1 A A ^ 'l A ^ -.037401 A r~ A A 0 ^ .059087 -.457660 -.05/468 OCE .0023722 -.032038 -.06 9930 .0135673 -.413074 -.077615 -2.40712 FIS .0025164 .7293255 -.038098 -.040031 -.206233 -.004602 .0039534 BRE .1070434 .0679063 .0638405 -.013524 -.125610 -.063524 .0366223 CAS .631^724 1 .591905 -.258442 -.066166 -1 .00354 -.198893 -.426279 OFT -.001356 .7512477 -.043646 -.007968 -.481341 -.104555 -.408852 OMT -.821357 3.554726 .2976448 -.127213 .4339475 -.087508 -.166115 EGG .6032040 -.030757 -.061632 .2897114 1 .274996 .001 9316 -.338014 ALC -.486680 .6610031 -.293090 -.263621 -.461080 -.190489 -.134547 CRT -.40806 2.314882 -.135975 .0159748 .0363135 -.025087 .0914517 OMK -.364127 -1 .10712 -.0251 13 -.212570 -.008941 -.142061 -.191409 BAN .5122970 .6602959 .0325937 .231 1630 -.208516 -.098515 -.316049 CHE -.414248 7.714869 .3238960 -.285560 -1 .74674 .0702598 -.290836 CLG .7914259 -1 .02578 .8370777 -.523126 1 .809225 -.200454 .2520830 CVG .6067594 1 .51 9959 -.812643 -.229105 1 .140329 .1418944 .41 1 1551 POT -.683898 -20.2892 2.747935 .790 9645 6.325170 .1229049 1 .214449 NUT -1.34475 5.21 8851 -.188226 -.436674 1 .106062 -.51 9088 .5771732 OBK -.400466 -6.51 834 -.006278 -.05981 1 -.203133 -.245502 -.360782 ALL -.090387 .0160301 -.052093 -.040872 -.023864 .0271916 -.117495 NGN .0387742 .2351353 .0200807 .0210751 .0734059 -.016122 .0031538

PAGE 77

70 Table 4.2. Continued FIS • — BRE — CAS — CFT — Cf-IT .__EGG — ALC— RIC -.001616 -.007526 -.074492 -.06 8217 -.070040 .0593708 .0234377 FAT .01 98739 _9.i}HE-4 -.11 8365 -.036749 .0471208 .0389397 -.038762 BEF 1563^78 -.057491 .3831360 .0286797 .2354929 .4777240 .0988804 FMK 1 557346 -.073793 .2557332 .0236068 5 .946E-4 -. 1 43623 .0451657 PTN -.075092 .002201 4 .2695404 -. 1701 16 -.204217 -.033944 .0072965 RBN .0589122 .0329744 -.138741 -.022037 -.216772 .2008286 .0428401 BEV .0249389 .001 4891 -.099094 -.021459 -.060726 .0325475 -.144362 POL -.004162 .O6O9455 .3676344 .001 8735 -.381390 .2450016 -.114095 FDW .4104132 -.020528 .9546401 .3776693 1 .783323 -.054995 .3189737 FVG -.022656 .0486153 -.177951 -.015499 .1778163 -.023128 -.046082 SPG -.027642 -.005601 -.05391 1 .0052233 -.059533 .156061 9 -.048841 POR -.176821 -.109873 -.820205 -.341312 .2842243 .6396559 -.166921 SUG .0013140 -.047590 -.168976 -.065631 -.039150 .0074313 -.020138 OCE .0063922 .0386272 -.401030 -.326501 -.105435 -.203884 .0012925 FIS -2.26643 .0865737 .0767535 .0938758 .0220144 .3681 472 -.003480 BRE .0851 989 -1 .65064 .7229140 .1 992 851 -.148714 .1 104829 -.022331 CAS .0844492 .7405521 -2.46890 -.167235 -.15821 8 -.817234 -.052326 OFT .1047978 .2213606 -.204471 -1 .91782 -.003799 -.533802 -.304935 OMT .0085992 -.207905 -.225103 -.013667 -5.54957 1 .01 1021 -.070650 EGG .5461401 .1623003 -1 .21607 -.691 475 1 .195802 -.896023 -.034385 ALC -.131739 -.159447 -.210890 -.512510 -.16 9077 -.117199 -9.00939 ORT -.309221 .3689959 .1517886 -.003153 .2702793 -.058410 1 .631070 OMK -.099549 -.014837 -.066880 -.527333 -.129997 -.106430 -1 .11082 3A^^I -.678602 .3803565 .3286 830 .3250292 -.i] 88552 -.166960 -.061887 CHE -.115647 -.455627 -.531897 -.166492 .4436546 -.404729 2.O87747 OLG 2.553636 2.041350 -3.02329 .3955413 3.287324 .2433079 3.3I6II7 CVG .0083157 .3797364 .2558534 -.706085 -.172108 1 1 93439 -6.11 157 POT 1 .435144 -1 .53408 2.076496 -2.96326 .691 9679 .4707097 -2.87398 NUT -5.60772 1 .099851 2.512835 -.326205 -1 .17740 -.14196 4 -3.61483 03 K -.21 1 946 .1436658 -.387002 -1 .46915 .4510727 -.150557 -4.13735 ALL -.028171 -.00531 -.06 8436 -.116379 -.02553 .0550306 -.137937 NGN .02 8706 4 .86101 9 -.001 164 .0154369 .0446424 -.003223 .0972331

PAGE 78

71 Table 4.2. Continued ORT CiMK BAN CHE CLG CVG POT RIC -.062313 .02i<5456 -.010298 .0323853 -.001 903 .02100^45 .1560048 FAT -.012758 -.016055 -.053591 -.124810 .1508645 .0716505 .2240309 BEF .0470362 -.090332 .3507754 -.355727 .6675116 -.052791 -.203732 FMK .1570338 .1742324 .1080805 -.229946 .6093731 .0035234 -.032239 PTN -.107195 -.093487 -.070486 -.351804 -.136563 .1300202 .4986379 RBN -.066835 .0277433 -.028138 -.181702 .0147180 .0905002 .1544178 BEV .0425854 -.048315 -.046259 .4230644 -.011716 -.001769 -.043394 POL -.154080 -.106350 .1493774 -.076449 .1550118 .0680749 -.071998 FDW .8570169 -.414417 .1780637 1.909078 -.237570 .1700044 -2.24716 FVG -.056182 .0108510 .0120647 .1070317 .1875307 -.098098 .3173145 SPC .0066465 -.070352 .0891739 -.061491 -.129828 -.029519 .1036145 POR .0092026 .0093420 -.087923 -.520344 .4733354 .1682160 .8638874 SUG -.012456 -.045208 -.039130 .0462464 -.052900 .0250253 .0178056 OCE .0500558 -.079165 -.145166 -.081645 .0786311 .0728992 .1929321 FIS -.186610 -.036151 -.332374 -.022365 .8268338 .0036421 .2411059 BRE .2156893 .0094873 .1848034 -.151901 .6608601 .070653 -.259636 CAS .0925466 -.015106 .1655990 -.180459 -.988304 .0502906 .3550703 CFT -.010144 -.317904 .1775954 -.054260 .3304197 -.146134 -.575366 OMT .1906417 -.083226 -.315087 .2289360 1.352089 -.041176 .1448370 EGG -.057239 -.071129 -.124661 -.209585 .1170743 .0336735 .1171167 ALC 1.417590 -.989746 -.109034 1.195267 4.110202 -1.72782 -.769369 ORT -2.33373 -.251440 -.331321 -.250602 .0938124 -.118438 .3969489 OMK -.295547 -5.20549 1.012308 .0646695 -.309854 -.194263 -.689520 BAN -.404852 1.153188 -3.70049 .3306530 -1.70372 -.181111 .2261567 CHE -.429190 .0807426 .3980064 -5.76217 2.681870 -.332943 -.783130 OLG .1680809 -.507711 -2.57713 3.165181 -7.25628 .1804332 -.488211 CVG -.398642 -.575928 -.495484 -.680496 .3193484 .7426848 -1.57164 POT 1.399601 -2.22046 .6578235 -1.73918 -.934774 -1.68530 -.475924 NUT -.643160 1.281608 .1693147 1.379173 -2.84034 -6.35253 -1.73916 OBK -.481366 -4.75277 -.749203 -.180294 -.524247 -.254458 -.07^981 ALL .0029225 -.153847 -.043972 -.OI383 .1440205 -.037763 -.027783 NON -.01.5717 .0233134 .0032229 .03806 .0005787 .0033905 -.0001 03

PAGE 79

72 Table 4.2. Concluded MTTT UOrt— — nljc; 1 1 Q FAT n Qfi 1 fin 1 REF 1 01 7Ufin FMK 048^1 ?Q \J ~ \J ^ \ J 6 166E-4 1 88'i2'^ 041770 HBN 118'?8Q n 1 7 2j 1 81 BEV • V t. \J J R lj2iPF-H ^ • T T ^ i-t" 1 POL ^^^f^f^7 • 1 1 ^ xJ U ( — U tf. U f J 1 FDW "=504711 FVG 01 81 7? 00^8^2^ SPC ^•jj ij— 3 POR 1 P'^PI QO 0l82iJfi snn OCE FIS -.781730 -.023471 BRE .1523711 .0257034 CAS .3550119 -.047577 OFT -.133957 -.229291 OMT -.212780 .0831820 EGG -.030828 -.025878 ALC -.79511M -.897658 ORT -.150757 -.105821 OMK .320i<962 -1 .20296 BAN .0477090 -.206579 CHE .4979151 -.067903 OLG -1 .22408 -.21 9353 CVG -4.88981 -.190790 POT -1 .47873 -.055998 NUT -1 .35568 -.047835 OBK -.054649 -6.99516 ALL -.057541 -.097627 NCN -.000414 .0066494 ALLF OOD-N CN -F GOD-IN COiME— -1.30598 .1886992 .2464378 -1.21402 .4408520 .5352039 1 .501769 -1 .06439 1 .598464 -.161558 7.418284 1 .198487 -2.30136 .4213434 .9696625 -.101468 3.196350 .3243274 -.766214 .8697111 .6663845 -2.00511 .6177352 1.238625 -1.82417 3.107933 5.669399 -.962248 .5916515 .5693156 -.817862 .6905149 .5647457 -.830175 1.804281 1.497870 1.255585 -.349708 .4390878 -3.73929 .1679450 .6279950 -.813257 1.011188 .8223554 -.029459 31.39321 .9248112 -2.07493 .1619863 .3361097 -4.97955 .2924275 1.456156 -1.68737 1.332207 2.247922 3.066789 -.368595 1.130403 -12.1737 .8261623 10.05747 .5699425 -.830109 .4731143 -11,2678 .2906487 3.274978 -3.02762 .2442520 .7395020 -3.18660 1.557967 4.964625 16.39067 -.060932 .9987187 -7.83437 .2150543 1.684815 -5.63910 .0367209 .4890273 -14.6953 .0316775 .4655107 -27.0397 .1462476 3.954494 -1.41957 .1655520 1.254023 .4192827 -2.30366 .7461540

PAGE 80

73 methodology. Some of these coefficients are out of a logical range; however they were not intuitively adjusted as in George and King (1971, p. 39) or prespecified as in Hassan and Johnson (1976, p. 41). This was done in order not to obscure the results obtained with the data and method used. As indicated earlier, the results of the estimation procedure were affected by the ordering of the commodity groups. The groups were arrayed in a descending order by the budget shares. As a consequence, any error build-up associated with the procedure will fall on those food groups that are least important in household budgets. Elasticity coefficients associated with these minor items in the budget are the ones closer to the bottom of the matrix. Problems with errors build-up were by and large associated with estimated coeificients corresponding to commodity groups with an expenditure proportion equal to or less than .006 (see Table 4.1). These are cheese, other dried legumes, canned-vegetables-and-soups potatoes, nuts-and-otheroily-seeds, and other bakery products. All of them together represent 1.8 percent of the average total expenditures of households in the sample. Looking at the matrix in general, it can be observed that most of the coefficients have a negative sign. For the own price elasticity this result was expected. For the

PAGE 81

74 cross price elasticities a negative sign means that the commodities are complements. Many of these commodities with negative cross price elasticities were expected to be substitute or independent groups. In many cases the unexpected negative sign can be explained because of the general substitution effect. Even though two goods i and j were independent, a change in the price of j does not leave the demand for i unaffected because there is an overall effect due to the fact that all goods compete for the consumers' income (See Phlips, 1974, p. 63). In a complete demand system this fact becomes obvious. Observing closer the results for own price elasticities, the inelastic commodities, those with own price elasticities greater than -1, were rice (-0.4?) non-alcoholic beverages (-0.77), spices and miscellaneous (-0.57), sugar (-0.46), eggs (-0.90), and potatoes (-0.48). The expenditure on those commodities will increase if their prices increase. Ten commodities were in the elastic range of -1 to -2, and include fat-and-oils (-1.38), beef (-1.70), fresh milk (-I.74), plantain (-1.63), red beans (-1.29), poultry (-1.73), fresh vegetables (-1.30), bread (-1.65), other fruits (-1.92) and nuts-andoily-seeds (-1.36). Seven commodities had own price elasticities less than -2. These coefficients seem too high in absolute value and included food away from home, alcoholic beverages, other-milk-and-

PAGE 82

75 milk-products, cheese, other dried legumes, and other bakery products. Price elasticities for beef and pork were estimated by the USDA (1978) for Mexico and Central America. The own price elasticity of the demand for beef was estimated to be -0.4; elasticity of the demand with respect to the price of pork, 0.1; income elasticity of beef, 0.7; own price elasticity of the demand for pork, -0.3; elasticity of the demand for pork with respect to the price of beef, 0.1; and income elasticity of pork, 0.6. The signs were consistent, but in each case the absolute values of the elasticities in this study were larger (by factors of 2.3 to 10.7) than the USDA estimates for Mexico and Central America. Also, in this study, some of the cross price elasticities were excessively high--for example, potatoes, other dried legumes, canned vegetables, nuts-and-oily-seeds and other bakery products. Also the elasticities showing the effect of the price of food away from home on the consumption of several commodities were believed to be too high. In general, the estimated cross and direct price elasticites for most of the commodities were within a logical range. For instance, on consulting the elasticities table it follows that for a 10 percent increase in the price of rice the consumption of rice will decrease by 4.7

PAGE 83

76 percent. The cross price elasticity of beef with respect to the price of rice (row 3, colunm 1) shows that for the same increase of 10 percent in the price of rice the consumption of beef was reduced by 3.9 percent. Also a 10 percent increase in the price of nonfood decreased the consumption of rice by 1.9 percent. An increase of 10 percent in income increased the consumption of rice by 2.5 percent, and so on. All-food has an estimated income elasticity of 1.3 which is reasonable for the country. The average total expenditure level was $220 per month, of which 50 percent was spent on food. For comparison Del Rosario and Musgrove (1980) estimated expenditure elasticities for the food, beverages, and tobacco group of 1.6, 1.0, 0.6, and 0.5 for expenditure levels of $100, $200, $400, and $800 respectively.. The estimated income elasticity for food was greater than the nonfood income elasticity. This result is reasonable in a country with a low total expenditure per household and implies that with a given increase in income, the consumption of food will increase more than the consumption of nonfood items. This result has implications for the nutritional policy discussed in the next chapter. The income elasticities for alcoholic beverages and food away from home are probably too high. The problem might be related to their high price/weight ratio since these elasticities are the result of dividing the

PAGE 84

77 coefficient for total expenditures from the regression equation, b j_, by the mean quantity consumed measured in pounds per month. Most of the income elasticities are within a logical range. For comparison FAO (1971) estimated income elasticities for the Dominican Republic for rice (0.60), starchy roots (0.29), sugar and sugar products (0.60), oilseeds (0.50), vegetables (0.40), fruits (0.2), beef and veal (0.80), poultry (1.0), pig meat (0.70), eggs (1.0), fish (0.60), milk (0.70), cheese (1.0), and fat and oils (0.69). The own price elasticity of all-food was -1.4. As indicated before, this estimated elasticity was obtained by summing the price elasticities of all-food with respect to each food group, e fj. One would expect the absolute value of this elasticity to be lower. However, in comparison with the own price elasticity for nonfood (-2.8), all-food was more inelastic than nonfood, as was expected. Further interpretation of this table may be interesting for specific commodities and in general, cross price elasticities in a demand system are important. The usefulness of this matrix and some implications for policy analysis will be discussed in the next chapter.

PAGE 85

78 The Familv-Size El9?ticitig? A family-size elasticity may be interpreted as the percentage change in the consumption of a good as the result of a 1 percent change in the size of the family. Most of the elasticities were positive and within a range of 0 to 1 (Table 4.3). This implies that the demand for those goods will increase when the family size increases, but in a smaller percentage. The commodities with a negative elasticity, as indicated earlier, were beef, food away from home, other fruits, other meats, alcohol, other milks and milk products, cheese, and other bakery products. Except for other fruits, all were most expensive products (see Table 4.1) and can be considered as "luxury" foods. Thg New Method ^ Estimating Eneel E la gt icite? The results of the new method of estimating Engel elasticities were applied to the consumption of rice. The estimated values of the elasticities were disappointing and excessively high. The results are presented in Appendix 4. Estimated expenditure elasticities used in the policy analysis were the ones obtained from the elasticity matrix. These coefficients are consistent with the price elasticities since they belong to the same demand system. The elasticities obtained using the new method which was described in the theoretical framework section though

PAGE 86

79 Table 4.3. Estimated family-size elasticities r Ai-1. SlZii nlC r AT 1 il ll c J 3Er — 1 yc; 0 r riK 0 y J 0 y £5 0 • 1 J yy^uy h 0 n 1 "7 c 1 H by 1 ^51 BEV 1 <10 44 Oh POL UOoo 1 <; yo UHii BRE .30^40631 CAS .7693678 CTT -.176609 ot-rr -.39^077 EGG .0015564 ALC -4.05590 ORT .8337375 -1 .38754 BA1-! .681 4934 CHE -1 .87372 OLG .4978635 CVG 2.606377 POT .7955863 :^UT .7251 987 CBK -.100358

PAGE 87

80 seemingly overestimated were logical in relationship to total expenditures. The elasticities for rice increased as total expenditures increased up to $200 per month where they reached the maximum. At total expenditure level over $200 the expenditure elasticities for rice began to decrease.

PAGE 88

CHAPTER V IMPLICATIONS FOR POLICY The data used in this research recorded the amount and the value of food items purchased by sampled households. In order to derive nutritional implications, it was assumed that the food purchased was consumed by the household. In the Dominican Republic this assumption is close to reality, especially for low-income households. Since the data do not provide knowledge of the distribution of food among family members, it was further assumed that food was distributed in proportion to the needs of each household member. Using the results of the elasticity matrix, the nutritional implications of the demand system were analyzed from the point of view of the potential impact of income changes on calories and protein intake. Income and price elasticities were used to analyze how a change in income affected nutrient intake. Before going further it is necessary to clarify the meaning of nutrition as this concept is used in this study. To analyze nutritional policies, some indicator about nutrition has to be used. But to choose an indicator is a controversial matter. In this study the indicators were 81

PAGE 89

82 determined by the kinds of data available to be calories and protein intake per person per day. These are flow measurements, but are also commonly taken as shorthand indicators of current or likely future nutritional status (Dowler et al., 1982, p. 103). Indicators of state are also commonly used. These indicators are based on anthropometry or on biochemical nutrition parameters. For a discussion of state indicators see Moussie (19^1, p. 10); and Davis (1982) The Effect iLf Changes jn Prices on the Calorie and Protein Intake Following the method described in chapter IV, the change in calorie and protein intake as the result of a change in the price of a commodity was estimated. The estimates were based on consumption per capita per day of each commodity, the price elasticities (Table 4.2), and the calorie and protein contents of each commodity (Table 5.1). The estimated effect of a 10 percent increase in the price of rice on calorie and protein intake was a reduction of 70 calories and 0.8 grams of protein in the overall per capita intake per day (Table 5.2). It should be pointed out that such estimates based on discrete changes can only be approximated by the point elasticities given in Table 4.2 since point elasticities are based on infinitesimal changes.

PAGE 90

83 Table 5.1. Calorie and protein in one pound of edible portion by food commodity group Commodity Calories grams of protein R T r 1 11 c; c a T C R 1 id 1 0 u y 1 1 nv 1 1 u / 9.11 ft CM V r n IS. 1 c n PT M c; Q7 D n n M ^ CO f\ 1 DiiO yy u D Ci U 0 *+ J 1 M n 1 0 u PDF 77 1 0 4 u r V u 5 O 5 d D o r ^ D y u e; 11 PnR run you 7 n n SUG 1678 n Q u y OCE 1674 47.9 FIS 453 94.0 BRE 1424 46.7 CAS 598 4.5 OFT 295 3.7 OMT 907 80 .0 EGG 671 51 .3 ALC 616 CRT 526 7.9 OMK 1002 27.2 BAN 498 5.4 CHE 1007 108.9 OLG 1526 102. 1 CVG 216 8.5 POT 412 13.7 NUT 1890 52.8 OBK 1769 33.0 Source: Calculated by the author based on average for the groups from information about the food composition for single commodities from: INCAP-ICNND, 1961; Whitney and Hamilton, 1977; and ONAPLAN, 198ub.

PAGE 91

84 Table 5.2. Effect of a 10 percent increase in the price of rice on calorie and protein intake per capita per day Commodity Change in Calorie change in gms. of Intake protein intake RIG -29.90 -0.600 FAT 0.04 — BEF -2.38 -0.183 FMK -0.56 -0.032 PTN -3.32 -0.030 RBN 2.73 0.177 BEV -.01 -0.000 POL -0.32 -0.034 FDW — FVG -0.47 0.005 PGR -3.75 -0.268 sue -22.43 -0.012 OCE 3.06 0.088 -0 oil BRE -0.76 -0.025 CAS -5.64 -0.042 OFT -1 .57 -0.020 GMT -0.83 -0.073 EGG 5.48 0.419 ALC -0.33 ORT -4.32 -0.065 OMK -0.02 0.001 BAN -0.71 -o.ooa CHE 0.07 0.007 OLG -0.34 -0.023 CVG 0.09 0.003 POT 4.95 0.164 NUT -7.95 -0.221 OBK -0.12 -0.002 ALL -70.25 -0.802 Source: Calculated from the information given in Tables 4.2 and 5.1.

PAGE 92

85 Also, estimates based on finite changes would be most accurate when the changes are near the average prices observed in the data. Finally, the estimates assume ceteris paribus conditions, i.e. only the price of rice changes. The major reduction in calories was due to the drop in the consumption of rice (RIC) wich accounted for a reduction of 29.9 calories. This was followed by a reduction of 22.4 calories due to the decrease in the consumption of sugar (SUG). However, there were some gains in calorie intake as the price of rice was increased. For instance, eggs (EGG) contributed with 5.5 calories, potatoes. (POT) with 4.9 and other cereals (OCE) with 3.1 calories. The drop in rice (RIC) consumption as a result of the price increase was mainly responsible for the reduction in protein intake--accounting for 0.60 grams of it. Pork (FOR), beef (BEF), and nuts and oily seeds (NUT) caused reductions of 0.27, 0.18, and 0.22 grams of protein respectively. Major gains in protein intake came from an increase in the consumption of eggs (EGG), potatoes (POT) and red beans (RBN), which accounted for 0.42, 0.16, and 0.18 respectively. However, the increase in the consumption of red beans was not expected since rice and red beans are consumed together and considered to be complements in the Dominican diet. The results suggest that the proportions of

PAGE 93

86 rice and red beans used in the diet are responsive to relative price changes. These results provide an example of how the estimated elasticities (Table 4.2) may be used to measure the effect of an increase in the price of a commodity on the intake of calories and protein in the Dominican Republic. Such measures can provide useful insight into how a change in price of a commodity can affect nutritional goals for persons concerned with food price and nutritional policies. The same analysis could be done for each commodity or group of commodities included in the elasticity matrix. lilS Eff get jQl Change? In Income Sm Uie Calorie and Protein Intake The effect of an increase of 10 percent in real income was evaluated with the income elasticities given in Table 4.2 to see how such a change affected the overall intake of calories and protein. As can be observed in Table 5.3, the consumption of calories increased by 166 per capita per day. The major increase was due to plantain consumption (PTN) which accounted for 16.9 calories, followed closely by rice consumption (RIC) which contributed 15.7 and by fats and oils (FAT) which contributed 15.2 calories. Conditions on these estimates similar to those given on pages 82 and 85

PAGE 94

87 Table 5.3. Effect of a 10 percent increase in income on calorie and protein intake per capita per day Commodity Change in Calorie change in gms. of Intake protein intake RIC 15.72 0.315 FAT 15.51 0.004 BEF 9.75 0.746 FMK 0.95 0.054 PTN 16.92 0.153 RBN 3.80 0.246 BEV 4.84 0.103 POL 4.96 0.540 FDW FVG 1.72 0.019 SPC 4.27 0.039 PGR 4.66 0.333 SUG 9.60 0.005 OCE 9.14 0.261 FIS 0.88 0.183 BRE 7.86 0.258 CAS 4.64 0.035 OFT 4.52 0.057 GMT 2.99 0.264 EGG 15.19. 1.162 ALC 6.29 GRT 3.56 0.053 OMK 4.11 0.112 BAN 3.53 0.038 CHE 2.97 0.322 OLG 3.91 0.262 CVG 0.28 0.011 POT 0.49 0.016 NUT 2.13 0.059 OBK 1.17 0.022 ALL 166.38 5.675 Source: Calculated from the information given in Tables 4.2 and 5.1.

PAGE 95

88 are needed. Furthermore here any problem on the food supply side are ignored. With a fixed supply of food an increase in incomes would simple lead to faster inflation. It is assumed here that the change is in real and not in nominal income. The cross-sectional survey data used in the analysis had greater variation among households than did the price data. As a consequence, the estimates of the income elasticities were generally estimated with the smallest relative standard errors (see t statistics in Appendix 2). The overall increase in protein was 5.7 grams per capita per day. The main contributions were from egg consumption (EGG, 1.16 grams), beef consumption (BEF, 0.75 grams), and poultry cosumption (POL, 0.54 grams). Rice consumption (RIC) contributed with 0.32 grams. It can be seen that rice (RIC) and eggs (EGG) are of great importance in the supply of both calories and protein to the Dominican population and must be given special consideration in any food policy. The contribution of cheese (CHE) and canned vegetables (CVG) must be taken with caution due to their small share in the consumer budget. To analyze the impact of income changes on the nutritional deficit at the national level the increase in calories and protein has to be compared with the nutritional needs of the population. ONAPLAN (1980b, p. 20) calculated

PAGE 96

89 the recommended consumption of calories and protein per capita per day as 2356 calories and 49.8 grams of protein. The increase of the intake of these nutrients as a result of a 10 percent increase in income would contribute importantly to correcting the calorie and protein deficiencies of the population. These per capita estimates at the national level provide only broad guidelines and say nothing about the distributional effects and the real nutritional status of low income groups of the population. Measures of demand and income elasticities are needed for different income groups by geographic areas to understand the distributional effects of food policies and nutrition. It would also be interesting and informative to obtain biochemical parameters data on the nutritional status of low income groups. Fairness and equity aspects of food policies cannot be ignored if food policies are to encourage political stability. Therefore, distributional aspects are of major concern and deserve serious analysis. The Effect 2l 3, C han ge n ipoom e and Pri ce sm the Consumption .aM imports 2lL Selected Food Commodi ties This section is concerned with analyzing the impact on consumption and food imports of (1) a 10 percent increase in real income per capita and (2) a 10 percent increase in real income coupled with a 10 percent increase in price. The

PAGE 97

90 second scenario provides some insight into a situation of rising incomes and rising inflation. The change in consumption was compared with the production, consumption, and import goals of the Secretariat of State for Agriculture (SEA) operative plan. The commodities included in the analysis were those which have been imported or exported in recent years and which were included as single commodities in the elasticity matrix. These were rice (RIC), red beans (RBN), and cassava (CAS). Plantain was excluded because it was measured in units and not in weight in SEA' s operative plan. The analysis was done separately for each commodity. The consumption estimates used as a basis to calculate the effect of changes in income and prices were those from SEA' s operative plan (Tabl.es 5.4, 5.5, and 5.6). These consumption estimates were lower than those from the "First National Survey on Households Income and Expenditure (SHIE) for 1976-77. These SEA estimates were used in order to unambiguously compare the results of the estimates based on elasticities from Table 4.2 with 19^2 goals of the SEA plan. The SEA rice program proposed a consumption goal of 314 thousand metric tons for 1982 (Table 5.4). This implies a per capita consumption of 0.32 pounds per day. This consumption goal was based on a projected production of 2b5 thousand metric tons and a stock at the begining of the year

PAGE 98

91 Table 5.4. Rice program from SEA' s operative plan 1980 1981 1982 goal in thousand metric tons Production 254.5 260.2 285 • 1 Initial Stock 30.8 34. 1 65. 8 Exports Imports 40.6 62.6 Supply 325 .9 356.9 350 0 Consumption 291 .8 292. 1 314. 3 Final Stock 34.1 64.8 35. 7 in pounds per day Consumption per capita 0.316 0.308 0. 3 in $ per pounds Price 0.33 0.36 in thousand population 5578.2 5731 .6 5889 3 Source: Secretaria de Estado de Agricultura (SEA). Plan Gperativo 1982. Santo Domingo, Dominican Republic: SEA, 198l; and ONAPLAN. P l an Trlenal ^ Inversiones Publicas, 1980,82. Santo Domingo, Dominican Republic: PLANDES 42, ONAPLAN, 198Ua.

PAGE 99

92 Table 5.5. Red beans program from SEA' s operative plan Commodity 198u 198 i 192 goal in thousand metric tons Production 36.6 37.8 45.2 Seeds and Crop Loses 4.4 4.2 5.0 Initial Stock 0.4 2.8 0.3 Exports Imports 6.6 Supply 39.2 36.4 40 .5 Consumption 36.4 36.1 38.4 Final Stock 2.8 0.3 2.1 in pounds per day Consumption per capita 0.039 0.038 0.040 in $ per day Price 0.63 0.64 Source: Secretaria de Estado de Agricultura (SEA). Plan Operative 1982. Santo Domingo, Dominican Republic: SEA, 1 98 l ; and ONAPLAN. Plan Trienal ^ Inversiones Publicas^ 1 9 8 0 H 2 Santo Domingo, Dominican Republic: PLANDES 42, ONAPLAN, 198ua.

PAGE 100

93 Table 5.6. Cassava program from SEA' s operative plan 1980 1981 1982 goal in thousand metric tons Production 80 .9 119.3 134. 4 Exports 1 .2 3.8 4. 4 Imports Supply 79 .7 115.5 130. 0 Consumption 79 .7 115.5 130. 0 in pounds per day Consumption per capita 0 .086 0.122 0. i: in $ per day Price 0 .28 0.19 Source: Secretaria de Estado de Agricultura (SEA). Plan Operativo 1982. Santo Domingo Dominican Republic: SEA, 1 98 i ; and ONAPLAN. Plan Tr i ^nal Inversiones Publicas, 1980,82. Santo Domingo, Dominican Republic PLANDES 42, ONAPLAN, 1980a.

PAGE 101

94 of 64.8 thousand metric tons. No imports were projected for 198^. Since 1982 was an election year it was more favorable to the government not to promise to import rice. In order to avoid importing in 1982, the commodity was imported in excess in I981 and stored to be used in 1982, As a consequence, rice imports in 198l were one and a half times those for 198u, though production was greater and per capita consumption decreased in 198l. At the end of 1981, stocks were almost double both those for 1980 and those projected for 1982. Rice consumption per capita was projected to increase in 1982 after falling from 198u to 1981. The price increased by more than 9 percent from 1980 to 1981. No hypothesis was given in the plan to support 1982 consumption goal. In this study, an alternative projection was proposed and was based on a method using the information provided in the cited operative plan and the price and income elasticities from table 4.2 for the respective commodities. The results are presented in table 5.7. The elasticities allow to make specific hypotheses about the effect of income and price changes. Population growth for 1982 was used and the consumption per capita was the average for 1980 and 1981. In the projection of rice consumption for 1982, a 10 percent increase in total income was first assumed. The

PAGE 102

95 Table 5.7. Alternative consumption estimates for 19^2 applying price and income elasticities for rice, red beans, and cassava to 198l figures from Secretariat of State for Agriculture (SEA) operative plan Commodity 1982 consumption, assumming a 10 % increase in income 1982 cosumption, assumming a 10 % increase in income & 10 % increase in price in thousand metric tons Rice Red Beans Cassav a Rice Red Beans Cassav a 311.4 297.7 38.9 34.2 105.3 100.6 pounds per capita per day 0.319 0.305 0.040 0.035 0.108 0.103 Source: Estimated by the author based on data from SEA, 1980; ONAPLAN, 198ua.

PAGE 103

96 result was a consumption of 311 thousand metric tons. This is less than SEA' s projected 19^2 goal. Assuming a 10 percent increase in income along with 10 percent increase in price, projected rice consumption was 298 thousand metric tons. SEA'S red beans program proposed a consumption goal of 38 thousand metric tons for 1982 (Table 5.5). In 1981 there were no imports and none were projected for 1982. Consumption remained almost the same during 1980 and 1981. To avoid imports in 1982, production was projected to be 20 percent higher than in 1981 during which year it grew by only 3 percent. With the assumption of a 10 percent increase in total income red beans consumption was projected in this study to be 39 thousand metric tons (see Table 5.7) which is almost the same as the goal from the plan. When a 10 percent increase in red bean price was assumed along with a 10 percent increase in income, projected consumption dropped to 34 thousand metric tons. SEA'S cassava program consumption goal was 130 thousand metric tons, or about 13 percent greater than in 1981, and 63 percent greater than in 1980 (Table 5.6). Exports of this commodity were projected to be greater in 1982 than in 198u and 198l. To fulfill these goals, cassava production was assumed in the SEA plan to increase 13 percent in 1982 with respect to 1981. Between 19au and 1981

PAGE 104

97 actual growth was estimated to be 47 percent. The projected consumptions in this study were 105 thousand metric tons, assuming a 10 percent increase in income, and 101 thousand metric tons, where a 10 percent increase in price was also assumed (Table 5.7). Both estimates are lower than the 19^2 consumption goal from SEA's operative plan. Finally, if the consumption data from SEA's operative plan (Tables 5.4, 5.5, and 5.6) are compared with the consumption data from SHIE for 1976-77 (Table 5.8) it is observed that the consumption level from the plan is lower in absolute terms for red beans and cassava. In per capita values, consumption data from the SEA plan are lower for all three commodities analyzed. The alternative projected consumption levels based on elasticities of this study (Table 5.7) are lower than the goals from the SEA plan. This result does not mean, however, that consumption from the SEA plan is overly optimistic. It is possible that data for the base year 1981 as well as that for 1980 was underestimated. The SEA goals were not subjected to technical critique but were developed and justified from a political point of view.

PAGE 105

98 Table 5.8. Consumption from the "First National Survey on Household Income and Expenditure (SHIE) and the operative plan of the Secretaria de Estado de Agricultura (SEA) for rice, red beans, and cassava Commodity 1976-77 SEA's from SHIE 1982 goal thousand metric tons Rice Red beans Cassav a 309.8 60. 1 190.5 314.3 38.4 130.0 pounds per capita per day Rice Red beans 0.361 0.070 0.322 0.040 Cassava 0.222 0.133 Source: from Tables 5.4, 5.5, and 5.6.

PAGE 106

CHAPTER VI SUMMARY, CONCLUSION, LIMITATIONS, AND RECOMMENDATIONS Summary and Conclusions The concern of development policy in the Dominican Republic provided the impetus for this study. Nutritional and other development issues associated with food such as food consumption projections and the effect of food demand on trade were of major concern. The analysis developed estimates of elasticities of demand for food commodities at a disaggregated level. The primary concern of this study was to develop these estimates and to demonstrate their usefulness in evaluating nutritional and food consumption policies. Estimates of direct and cross price elasticities were used to study the effect of a change in the price of one commodity on the consumption of that commodity and other goods. Such estimates were also used to analyze the effect of a change in income on the consumption of each commodity or group of commodities included in the study. And through the nutrient characteristics of each commodity, the impact on the nutritional intake of the population of price and income changes were estimated in a partial analysis. 99

PAGE 107

100 The main feature of this study was the estimation of 30 demand equations whose coet"f icients were used to estimate a full matrix of demand elasticities. The estimation of the demand equations and the calculation of the demand elasticities matrix were from cross-sectional data from a national household consumption survey. To estimate the demand equations, the tobit estimation model was applied to the semilog forms for each single equation. This estimation model was the most appropriate since it allowed the use of all observations available in the saraple--including those which were concentrated at zero. The estimated elasticities matrix was comprised of 996 price and income elasticities for 30 food commodities and groups of commodities, the all-food aggregate, and one nonfood group. Family size elasticities were also estimated. In general, the estimated cross and direct price elasticities and the family size elasticities for most of the commodities were within a logical range. The effects of changes in prices and income on calorie and protein intake were estimated using the elasticities. An increase of 10 percent in the price of rice, and also in income, were postulated, and the changes in calorie and protein intake were estimated. The results reinforce

PAGE 108

101 that rice and eggs are of great importance in supplying calories and protein to the Dominican population. For Dominican policymakers, this kind of analysis can provide useful insights into evaluating the effects of changes in prices and income on the nutritional intake of the population. However, broad per capita measures at the national level do not reveal the distributional aspects of nutrition problems and particularly the nutritional status of low-income groups. Finally, an analysis of the effect of a change in income and prices on the consumption and import goals of the Secretariat of State for Agriculture (SEA) crop programs was undertaken for rice, red beans, and cassava. The conclusions were that in the three cases, the projections of the SEA'S operative plan could be improved making use of price and income elasticities. It was also concluded that the established goals were motivated mostly by political criteria rather than by technical consistency. Limitations ^ iJig Study As pointed out in the section on problem setting and objectives, the study was partial in the sense that it focused mainly on the demand side. The impact on food consumption and imports was based on estimates from SEA's operative plan.

PAGE 109

1 102 Elasticities were estimated only at the national level. The effects of prices and income on the demand for food for specific income groups of the population were not estimated. Nutritional policies affect specific target groups and estimates are needed for each such group if enlightened policies are to be developed. There were limitations of the data used. Data from the First National Survey on Household Income and Expenditure (SHIE) was not primarily intended to be used for the evaluation of nutritional policies. Nonetheless the detail on disaggregated commodities consumption, the scope of the sample, and the accuracy of the survey provided a vast amount of data. However information on the actual consumption of food purchased, the distribution of food among family members and other nutritional aspects such as food preparation were not collected. Use of the data in this study was based on the assumptions that foods purchased were actually consumed by the household, and that food was distributed proportionally to each household member. It is doubtful if these assumptions are at great variance with reality in the Dominican Republic. To estimate the effect of a change in price on calorie and protein intake ceteris paribus conditions were assumed,

PAGE 110

103 i.e. only the price of rice was assumed to change. The effect of the change in price on the supply side was not considered. Suggestions for Further Research The present study can be extended, and experiences from this effort will be useful to further improve demand research in the Dominican Republic. The estimation procedure described in the methodology, and sketched in Table 3.1, can be done in an alternative way including the prices of all the commodities or group of commodities in each equation. This procedure might diminish the specification errors in those demand equations corresponding to commodities with the lower budget shares. Work can also be extended to analyze the effect of income and price changes on the demand for food commodities by income group and geographical area. Such an effort would require the creation of new data sets for the selected food groups, and the estimation of all demand equations and an elasticities matrix for each income group and area. The number of nutrients included in the analysis could be extended to be more complete. Also, the nutrient content of the food groups could be estimated as a weighted averages to gain accuracy.

PAGE 111

104 Finally, what happens on the supply side is, of course, quite important to the policy making process and information on the supply response of farmers to price changes is important in assessing the total impact of price policies on the economic system.

PAGE 112

APPENDIX 1 COMPOSITION OF FOOD GROUPS

PAGE 113

COMPOSITION OF FOOD GROUPS Rice Other cereals oats corn-flakes noodles, macarronies, and other pastes corn flour wheat flour corn grain Bread hard crust soft sandwich and others Other bakery products cakes biscuits cookies crackers (salty) crackers (soda) crackers (regulars) Cassava Potatoes Other root crops sweet potatoes cassava bread mapuey yam malanga Sugar and sugar products brown sugar white sugar conf ectios candies syrups honey Dried red beans Other dried legumes peas 106

PAGE 114

107 chick peas pigeon peas white beans broad beans lima beans and others Nuts, coconut, and oily seeds dried cocunut nuts and seeds (mixed) Fresh vegetables peppers garl ic pumpkin eggplant onion chive pigeon peas jack beans red beans 1 ettuce okra cucumbers radishes beets cabbage tomatoes carrots other fresh vegetables Canned vegetables and soups pigeon peas sweet peas chick peas other canned vegetables condensed soup vegetable soup beef noodle soup other soups Bananas Plantain Other Fruits avocado sugar cane coconut (fresh) guava granadillo

PAGE 115

108 papaya 1 emons sweet limes mangoes apple sour oranges oranges pineapples grapefruit grapes other fresh fruits dried fruits dried canned natives dried canned imported canned fruit juice (imported) canned fruit juice (native) Beef Pork Poultry Other meats goat meat rabbit pig feet other fresh meats haras saussages other preserved and canned meats Fish and shellfish Eggs Fats and oils copra oil sun flower oil peanut oil soybean oil olive oil butter margarine lard other oils Non-alcoholic beverages cocoa chocol ate

PAGE 116

coffee soft drinks mabi mal ta other beverages Fresh milk Other milk and milk products ev apor ated condensed cream other milk products Cheese Alcoholic beverages beer rum others Food away from home Spices and miscellaneous pickles blak pepper .oregano coriander salt tomato sauce ketchup mayonnaise other sauces flavors and extracts vinegar gelatin baby food olives ginger other miscellaneous

PAGE 117

APPENDIX 2 REGRESSION COEFFICIENTS FROM TOBIT MODEL

PAGE 118

RSGS^^SSICN CGlFFlCIcMTS, ASSOCIATED T-3TATISTICS 0EPEND6NT VAPI4BLE = RIC VAr! I AdLE COEFF I CI ENT T-5TATI ST IC LP IC -29.5512 -7. 126 LF AT 0 .9624 0.594 LBEF -7.7945 -3.C53 LF MK -14.3550 -6.416 LP TN -3. 3923 -2.233 L P SN 4.8909 2.432 LB E V 0.5 757 1 2 77 L P CL —0.1 354 —0.051 LF 0 W — 4 • 7 3 '-J 0 —3.36 8 LF VG 3.1 654 3 94 8 LS PC —3.4991 —4.75 0 LPC P -16.4? 07 —5.925 LSOG -14.2620 —6 .430 LCCE 3.0 379 2.577 LF I S -0.1 034 -0 092 LBRE -0.4 316 -0.54 1 LC AS -4.7668 -3 26 3 LOFT -4.3653 -6.361 LO"T -4.4319 -2.0 18 LEGG 3 .7992 1.713 LALC 1 .4598 2.254 LOPT -3.9375 -5.339 LOMK 1 .5707 2.167 LOAN -C.6590 -0.435 I f* w c L. 1n t c • J r \j • J 1 LCLG -0.1218 -0.052 LC VG 1.3441 1 .459 LPCT 9.9829 4 .4 74 LNL'T -2.6872 -2.874 LCBK -0.2284 -0.364 LCCI NC 1 = .7698 20. £17 LCGMEM 36 .6581 33.524 CONSTANT 142. 9078 -11.184 THE MEAN SQUAKEC ERRCR. USING E(Y) FCP PREDICTED VAL'JES. IS = 664.34843 THE MEAN APSOLLTE ERRCR IS = C. 30004 THE "1-SDICTF:0 PRCbAEILITY OF Y > LIMIT GIVTN AVERAGE X(I) = 0 .9529 THE 08SERVEC FREQUENCY OF Y > LIMIT IS = 0.9366 in

PAGE 119

112 RHGPESSIGN CC EFF I C I E'l T3 A3SCCIATED T-^TATISTICS ePE^OENT VASIASUE = FAT VAKI ASLE CDEFF I C I EMT T-STAT I ST IC LF A T -6.cj77 -33.6 40 1_ a EF 0 •4822 0.302 LF — 0 • 2 553 —0.405 L P T N 0 • 1 6 32 0.440 L P 6 N 1 399 2 • 27 1 LS E V — 0 • 1 29 5 -1.010 LPC L 0 • 4 79 1 0.640 L F *' — 1 • Z 27 I — 3 .5 25 LF VG 0 • 3 905 1.724 LS PC 0 • ^ 19 3 0 • 093 LP C P — i •2 2 64 1 577 LS L G — 1 • 7 7 1 0 — 2 93 1 LCCE 0 • G 37 1 0.302 LF I S 0.5399 0 75S LB '^E 0 J 1 1 4 -0 045 LC AS -1.4 253 -3.465 LCFT —0.4436 -2.291 LCVT 0 • bo83 0.906 LEGG 1 0 73c 1.717 L ALC -0 .4 o79 -2.524 LOFT -0 : 54 0 -0.731 LOMK -0 1 93a -0.946 LEAN -0^c469 1 526 LC t-E 1 • 5 06b — 1 .821 LCLG 1 • a 2 1 1 2. 727 LC V3 0.3649 3.220 LPCT 2 7 04 2 4.289 LNUT 1.1610 4.035 LOGK 0.0900 0 507 L Jul NC 6 .4605 29. osn LCGW^M 2 2 'j Jo 7.3?" Ct.NST ANT I 7. 6957 -5.54 TH2 "^HAN SGUAREC SRRPP. USING E(Y) FOR PPEOICTED VALUES. IS = THE Mr. AN ABSOLUTE =f^^0^ IS = 0.477ea THE PREDICTED PKrGAEILITY OF Y > LIMIT C-IVEN AVERAGE X(I) 0.9500 THh' QDSERVEC PSECUENTY OF Y > LIMIT 13 = 0.9912 52.22757

PAGE 120

113 SEGSESSICN COEFFICIENTS. ASSCCIATED T-STATISTICS DEPENDENT V^RIAFLE = BEF VA^ I A FLE CCEFF IC I ENT T-ST AT I ST IC LBEF -13.394 LFWK Z,22t9 2 727 LPTN 2.0207 2.804 LREN 3 .3 546 3.3ec L3EV 0. 1 467 0 6 C5 LPCL — 1 .3 tjc 3 —0.99? L F D —1.5321 —".442 LP V 'o C 7 9 d 0 1.775 L ^ P C 1 — C J f "3 ^ n c L P C t "TP! 1 / C 1 L S C G ~ C 2 7 LCCE — 0.201*^ J w — 0 LF I S 1 J 04 0 LO P c — 0 -1 7 95 — 1 .004 LC AS 3.1 955 4.047 LCFT C 2 39 2 0.^44 LCy T 1 564 1 1.611 LE GG 3 V £4 4 443 LALC 0.8247 2.505 LO P T C 3 923 0.962 LOKK -c .7534 —1.980 Lti AN 2.9 2 56 3.446 L C H E —2.9669 2 C2 S LCLG 5.5673 4.271 LCVG C .4 403 -0.9 33 LPCT -1.6992 1 .437 LNLT C; 94?6 1.557 LCEK C. 3777 2.645 LQGI NC 13.3313 3 0.443 LGGMtw -1.2533 -2.296 :CNSTAN7 -2C.6722 -3 .44 : THE VEAS EC'JASSr fSRCP, JSir.G E(^) FCS PREDICTED VALUES, IS = THE V1EAN AtSCLUTE EPPCfi IS = C.173£9 THE Pi^EOICTEO PPCSAEILITY Y > LIfIT GIVEN AVERAGE X(I) = 0.6C99 THE OPSERVEC FRECUENCY CF Y > L I I T IS = C.cCCA

PAGE 121

114 S£G,7ES31C.\ CC5FF I C IcNTS ASSCCIATEO T-STATISTICS DEPENDENT VA'7IA';l5 = FMK VA^I AEL€ CCEFFICIcNT T-STATIST IC LF •< -7.39^0 -15. 047 LPTN O.*038 1.242 L P E N C .3 0 1 6 1.591 LB ;v -C. 1 36 9 -i.e?2 LPCL 0 0 6 E 1 -0.102 LF3* -0.0 l-l -C.C54 LFVG -0 : <; 1 e -0 =57 LS PC c .3<; 1 a 2 155 LPC t-" 1.2 067 1 .a 10 LSUG 0.5369 1.010 LCCE 0.6716 2 .656 LF IS 0.7 072 2.571 LSSE -C.3251 -1 .542 LC A S 1.16 13 3.247 LCFT 0.1 072 0.635 LC WT 0.0 027 0.005 LEGG -0.6522 1 20P LA LC C.2 0S1 1 .325 LOFT 0.7 151 3.955 LCVK C .7<;i2 4.474 LO AK C.4';03 1.336 LCI-5 1 .0442 1 .525 LOLG 2.7672 4.64? LC VG 0. D 1 cO C,. :7 3 LPCT C 1 464 -0.275 LM. T C .2 203 0. e?4 LC EK CO C2H 0.019 LCG I KC 5.44 24 23. 135 LCC^C" 0.4264 1.719 CCNST ANT 13 2544 -4.82: ThE VEAN 3QUAMEC EKPCS, USING E(Y) FCfi PRECICTEO VALUES, IS = THE MEAN AESOLUTE £?RCR IS = 0.ie3S4 THE -^REDICTEO OPCEAEILITY Y > L I NI T GIVEN AVr=AG£ X(IJ = 0.6942 THE cesefve: Fi;EcuE^cy of y > livit is = o.6e9c

PAGE 122

115 "^EGRCSS ICV COEFFICIENTS, A33CCIATE[3 T-STATISTICS CEPENCENT VARIA5LE = PJH V AK I A H LC T — bTATI ST tC LPTiN *• 2 2 • 0 A S LR 3 N 7 T ^ A a c J • O O c 9 4.471 L3E V "—1 •63'''^ — 1 • 4 4 c LPCL — lO**" 180 — 1 • S £ 3 L F C 'A ~ 2 • 8 ^ 4 S 0 ? 9 5 LF N/G ~ 2 • 3 7 C ~ — 1 • 1 ? 3 L S PC ~ 1 • 0 6 G 9 -0 • 5 5 e LPCF -2 • 2 1 4 L3 LG ft •6262 0 € 2 2 LCCE — Q 4 7 i <3 LF I 5 —3,5475 ~ I •233 L3 =E 0 • 1 0 ft 0 0 • 0 4 5 LC AS I 2 • 73 36 3 • 4 1 fc LOFT -S .0 366 H • 3 ** f LC/T — 9 .6 4 76 LEGG ~* 1 • 6 0 3 6 — r ? o c U • LIVJT GIVEN AVERAGE X(!) = 0.717C The 33SERVED FSECUENCY GF Y > LIVIT IS = C. 799ft 3 1 ;

PAGE 123

115 !:?G?-3SiC^' C::FFFIcrFMT3. 1SSCCIATE0 T-STATISTIC Orc5h3FM MiPHBL=RBN C C ~ F F I C I f N T T-S T AT I STIC LP '^N 1 5 -! 7 0 3 -23 74 LS C V :.: 2^7 0.799 LP CI. 1 2= 1 1.941 LFD^-r 0 ^ 1 i 0 1 7 1 r LF \/r. C ? ~ 9 1 4 ] L3 PC 1 •19 15 — c ,275 L = CJ 5 6 £ 1 —3.440 LS UC — • 11 3 C — 4 c 1 A L C C F 0 0 ~ 1 4 ^ Q o LF 1 S C • 7 ~; 3 -1 1 / LP r C • 4 0 91 1.221 LC A S ~ 1 • 7 2 1 — w i c 1 LCFT — C 7 ~ 1 C 7 LC T — 2 • ri 39 4 it i LT GG 2 c J ^ LA LC C • 5 3 lb c 1 ] LCFT —3.2292 2 '-^ 2 LON K 0 • 3 -* 5 1 .3 7 LE AN — C • J A 9 1 — C • ^ 1 *^ L C K ^ 2 • ? 5 4 3 -? 04£ lcl;; 0 1 sa'^ C 2 C2 LCVC 1 • 1 22 3.223 LPCT 1.9155 2.225 LNLT 1 .4* = S -3.670 LC 0 2 1 f. ; 0 9 0 6 L3G r ^C 4 0 2 14.27! LLGM i'-{ .0 690 1 5 £ 4 2 Cr, \ 5T -3a. 253'. -9.124 Tl-:-JiFAN 3C'Ji-eC rhPCP. USING £(V) -CP FPEDICTEn VALUFS, IS = T."f '-lEAN A.33:JLUT^ -f;9C9 IS = C.3152C Ti-3 ^s^DiCTjo P"'C3aeil:ty :,r y > limit civen avepage x(i) = o.sp^^ THE GHSE^VEC F9EGU-3NCY CF Y > LI"IT IS = C £ 9 £ 0

PAGE 124

117 ^£G-ES3:C^: CQE^P I C I E T 3 • ASSOCIATED T-3TATISTICS OEPEN'CFNT VAPIAFLE = BEV V AR I A CLE C CEFF I C Ewr TSTATISTIC LP E V I C • i •* 1 o 3f .2 1 2 CL — 1 .6 f 57 1 .104 LF 0 >> — 0 • i 225 C 3 3 0 LF VG 1 .5 A06 -3,317 L5 PC 0.2371 0.5 = 5 1 C ft 0 0 7 03 L S 2.1499 1 f 99 LC CE 0 373 0.405 LF I S 0.3306 0.513 1 u z c L T t 0 • C I 9 7 0.339 L C A S ~ 1 .C 1 I 0 — 1 .5 62 1 t~. ^ T ~* 0 83 a -0.732 LO T *" C • ^ 0 ^ 4 — 0 12 2 LE G2 0.^306 0 — 4 0 LALC —1.9099 LC PT 0 63^ 1 "J 1 Q i 1 ^3 L C' V K — G • 6 ~ ^ ? — 1 .^47 L5 AN — 0 1 1 ^ 0 — C 7 I 1 LC t-E S fi 97 1 O -T O LCLG — 0 i 5 5 0 J i 1 LC V G —0.0234 -0 045 LPCT C • -J 7 a I -0.45 2 LNUT -c.r76: -0.483 LC OK CO O"?? 0 .020 LCGI NC H .8 I e2 20.067 LCGMF" •_ 7 2g 2 eft 1 CCNST ArJT -4 1 62 3 ; -e 550 E MEAN SOUAREn EtlRCB, US ING E ( Y ) FCf. =::ec;icted MEAN AESCLUTE E K C K IS = 1 .94964 THE PREDICTED D^^C2AcILITY Y > LIVIT GIVEN AVERAGE X(I) = 0.79C5 THE aaSERVEO FPECUESCY ZF Y > L I IT IS = R 9 0

PAGE 125

118 F LI~1T GIVEN AVERAGE X(I) = j.SvOP THE 3=1SEKVED r^PGUENCY C! F Y > LIVIT IS = C.3S01

PAGE 126

119 REC^ SSmfl COEFFICIENTS. ASSOCIATED T-STATI3TICS OePENOcNT VARIABLE = FDW V AK I A BL fc: COEFF IC I ENT T-3TAT I S TI C LFOW -13.3127 -15.904 LF V& 1 1 533 — 1 .422 LS PC -0.3867 —0.528 5 • ; 6 6 Q 2 • 085 L S UG 3.7 348 2 Q I 3 LCCE — 0 .2 5 2 — 0 265 LF I S 0.3097 0 754 LB RE -0.0405 — 0 04 6 LC AS 1 3 834 1 .34 6 LOF T 0.7451 1.157 LOMT 3 .5 1 83 1 589 LE GG -0 : 085 -0 053 LALC 0.6 29 3 1 1 64 LQRT 1 .6 90S 2.427 LO "K -C a 1 76 -1.230 L9 AN 0.3513 0 233 LC HE 3.7664 1 535 LGLG -0.4687 -3.2 10 LC VG 0.3354 0.410 LPOT -4.4334 -2 267 LNUT O.T 1 79 0 '574 L08K -1.1 732 -2. 120 LQGl NC 1 1 : as 1 15.706 LOGMEM -3.6738 -4 028 CONSTANT -4.b. '4 3 1 -4.630 THE MEAN SQUA.^EO ERRGrT. USING E(Y) FOR PCEDICTEO VALUES. IS = 53.7 THE MEAN ABSOLUTE SnROR IS = 0.52375 THE PREDICTED ^'RCaAEILITY OF Y > L I •' I T GIVEN AVERAGE X(I) = 0.1773 THE OeSERVED FREGUESCY OF Y > LIMIT IS = 0.2411

PAGE 127

120 ^;~G?;SSICN CC?FF IC If NTS. ASSOClaTED T-3TA^ISTICS OEPCN.-JNT VAPIABLH = FVG VAP I ACLH CC.= FF I C 1 rNT T-ST ATI STIC LFVG -27 :5r 3 -35.642 LSPC j .5 796 -1.030 LPCr 2 5^1 1 .ago LSLG 3 "-^6 3 2 334 L0C5 1 0 3o6 1 .331 LF I S -C .-704 -0.549 LD=F 1 0 0 '-i 4 1 .434 LCA3 -3.6 941 -3.368 LOFT c ; 2 1 -3 -0.626 LC'T 3.6-i2 J 2.170 LE^Z : i c 0 2 0 ??4 LALC -0. ^356 -1.917 LCCT 1 I 6 f 5 2 0 4 t LC.v< 3 .£253 0.408 LH AN D ^ 505 0.219 LCI-5 2 i 12 3 LCLG 2.a-?37 2.145 LC VC -? .03i3 -2.39! LPcr f -384 3 SS 1 LNL'T 773 0 4 r; 6 LCHK 0.1211 0.253 LCGI NC 1 : .3237 20.428 LCGVSM 2.3-70 3.651 CONSTANT -6"^. 25.35 -3.335 THi VEAN SQUA£3 £R = cr-, 'JSINo ( Y ) '^UQ PaEOICTHC VALU = £. IS = THii .'/EAN AESCLJTr £ C IS = 3 0231 4 THz '^.-ir-r; ICTEO P^C3AEIl:TY of Y > LI'-IT GIVES AVFSAGc XCI) = 0.3446 THE CdSEPvEC FREQUENCY -j F Y > LIMIT 1 3 = 0 974 2

PAGE 128

121 PfGP'-SSrrN COEFF ICIFMT3, ASSCCIiTED T-3TATISTIC3 OEPFNOENT =SPC VAP I AELH CG5:F= I C I ENT T-S TA T I ST IC LSPC -11.; 223 -27 795, LPCfi -2 J as6 -1.676 LSLG -0 .61 7 LOCE 0.2 ,'73 0.451 LF I S -0.5^19 C 934 LacE C 1 098 -0 23S LC AS 1 j 560 1 .42 3 LCFT 0. ; 02* 0.294 LC VT 1 I c72 -1 .013 LECG 3.0 5^.5 2 t70 LALC -O.v 575 2 ci 2 7 LCP T 0.1303 0.339 LCVK -1.3 742 -3.ce4 LB AN 1 .74 = 2 2.24a LC hE — 1 .2 055 LGLG -2.5452 -2 078 LC VG -0 .5767 -1 .21 C LPCT £.0313 1 .769 LNLT -0.9321 -1.804 LCSK -C .0005 -0 .002 LCGI NC 1 1 .C 715 27. 754 LCGWEM 1 .'-i^lQ 3. 763 CCNST ANT -4e 0975 -a 53( THE Vt-AN SGUAKFD =i-SfiCR, USING E(Y) FCS PREDICTED VALUES. IS = 180.0 = 72? TmE mean AeSCLUTE EPCCP IS = 0.a528e THE Pf^EDICTEC PR3BAEILITY JF Y > LIMIT GIVEN AVERAGE xtl) = 0.C269 ThS COScSVED FfiEGUENCY OF Y > LINIT IS = 0.9903

PAGE 129

122 ^EGPtSSICN COSFF ICIEMTS. ASSCCIATE") T-STATISTICS OEPENCENT VARIABLE =POR VAk I ABLE CL-:FF IC I ENT T-5 T IT I ST I C LPCR -1 5.ie--;5 -12.211 LSUG -0.-674 -0 2 ?<5 LCCE 1 r! 4 = 0 -2 .5e 1 LF 13 -0./3S77 -1.511 LBiiF. -0.5516 -1.195 LC AS -4 : 77 -5.655 LC^T -1.^135 -4. 993 LOWT I -> 269 1.291 L£ GG ; • 4c 23 3. CR3 LALC -C .8380 -2.614 LC CT 0 C 462 0.123 LC 0.0 46 9 0.129 LEAN -0.4414 -0.582 LChc -Z.C123 1 .797 LCLG 2. r 763 1.^36 LC VG 0.3 44 5 1 .7ac LPCT 4.2 370 3.652 LMJT 0 t, 1 ? 5 1.231 LCEK -0.09 16 -0.292 LQOINC 7 .5 198 19.675 LCGHEW 0.2016 0 .578 CONSTANT -77.2230 -6 727 THE "EAN SGUAt^EC EtiRC. USING E(Y) •^OS PREDICTED VALUES. 15 = THE MEAN ASSCL'JTE 7CRCR IS = 0.10472 THE PQEDICTEO Pf^CeA£ILITY OF Y > LIVIT IIVEN AVERAGE X(I) = 0.4333 THE QBSEf-VEu FSECUE^CY c: F Y > LtviT IS = 0.4954

PAGE 130

123 = -C;P-3SlCN CC5FF I Clr.vTS 4S3CCIAT=0 T-3TATI3TtCS QrOENT^NT VARIi"LE = SUG CCEFF ICI ENT T-STATISTIC LSUG -<;,? 9Tb -9.7=1 LOCE -1 ..: 7 09 -3.135 UP IS C. 3^67 3 056 -0 670 -2.526 LC AS -3.U.335 -5.r 1 : LCFT 1 J46 -4.65= LCVT -0 7c;55 0 a 3 3 LE C-G 3.1=10 0.159 LALC -C 4 092 -1 .45f LOCT -0.2531 -0 .794 LGK 0 > 1 2 5 -2. 959 LB AN C 7 g 5 1 -1.234 LC t= -0.9 397 -0 746 LQLG 1 .0749 1 .06 3 LCVG 0.5 03 5 1.251 LPCT 0 : e 1 3 0 .37 V LNLT • .2=37 -2.9 1 LGEK -C .i=57 -1 cP7 LCGI NC e 2 2 0 2 .5 3 2 5 LOGVS.'^ t .7550 1 0 94 0 CONSTANT -e LIWIT GIVEN AVEtAGE X(I) = C.9c29

PAGE 131

124 CHC-t^TSSI CaSFF I CI I^'T 5 ASSCCMTCT T-STaTISTICS r-F'^CN^tlN T VA^IA3LJ =OCE V A!-, I A L E c Cr. =F ICI ENT '-STATISTIC LCCE 2 a ? i -3 0 3 4 q 2 1 LF IS 0 0 I 0 0 C? 1 LORE : 6 ? LC AS -5.7 0<)1 -4 40 1 LCFT a f 4 ^ 1 -7.77£ LCVT -1.5317 -0 ""A-; LE GG -2. 7-M 1 : ? LALC 0.0 l-?4 0 0? I LGRT 0.7126 1 07S LCK -1.1 270 -1 .75 1 La AN — 2 h66 1 s 3 e LCI-S 1 1 2 -0 .44LCLG 1 I l'~-4 0 .532 LC VG 1.0373 1 ; 5 <; LPCT 2.74£6 1 394 LNLT 1 ; 7t2 1.207 LLaK C e 2 2 1 :)4 i LCGI NC 'J "J 4 0 2 14. 547 LCGy C 4 1 2f D t c 0 4 CONSTANT 102. c787 1 1.564 £ iVEAN SGUAPEC ISPCF: U3 ING ( Y ) F2?; PKECICTFD MSAK ABSOLUTE EKRCr IS = 2.31117 THE aKEClCTEO PKCHACIlITY QF Y > LINIT GIVEN AVERAGE X(I) = 0.6=25 THE OSSEf-VED r-.ECUENCY CF Y > L I w I T 13 = 0.8450

PAGE 132

125 REGRESSION COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = FIS VARIABLE COEFFICIENT T-STATIST IC LF IS -8.4009 -24 586 LBRE 0.3209 1 1 52 LCAS 0.284S 0.654 LOFT Q.3 66S 1 786 LO*T 0.0816 0.117 LEGS 1.3646 2 034 LALC -0.0129 -0 064 LORT -0.691 7 -2 .967 LOMK -0.1 340 — 0 604 LBAN —1.2320 —2 638 LCHE -0.0829 -0.092 LOLG 3.0648 4.200 LCV6 0.0 1 35 0. 047 LPOT 0.8937 1 .306 LNUT -2.8978 -9.505 L08K -0.0870 -0. 449 LOGINC 3.0482 14. 1 17 LOGMEM 0.4304 1 .548 CONSTANT -14. 1405 -4.61 THE MEAN SQUARED ERROR. USING E LIMIT GIVEN AVERAGE X(I) = 0.5703 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.6387

PAGE 133

126 KEGfiCSoICN CCHFF ICIENTS. ASSCCiaTED T-57ATISTICS CEP£^OE^^ v*pi4fle =BRE V SR I ASLZ C3EFF I C I ENT T-STAT I ST IC lepe 1 S 7?7e -26.372 LC,4o c.9i!ja 7.933 LCFT I SC73 4.724 LCVT I .4233 1 0£ C LEGG I .0574 C. 793 LLC -0.2142 -o.sse LCPT 2 J e 4 3 4. se 1 LCXK o.osoe 0 .2 1 1 LE,4N 1.7c87 1.929 LCI-E 1 .4536 -0.646 L CLG e .324? 4.363 LCVG 0 c 7c 2 -I .240 LFCT -2.4349 -1.39 4 LM.T 1 .45d3 2 .401 LCEK u .2400 O.e S7 LCG INC a .651 I 20.9 S 1 LCGOEV 2.9101 4.793 CCNSTVINT -22 339 1 -3.797 ThE ME*N SOUABED ERRCR. USING E(Y) FCP PPECICIEC V4LL5S. IS = 149.4C7e4 THE VEAN ABSCLUTE EPfiCh IS = 1.35114 ThE PRECICTEO PRCBAcILITY Zr r > LIMIT GIVEN AVERAGE XtlJ = 0.6646 THE CBSEfiVED FPEQUENCY CF Y > LIWIT IS = 0.7331

PAGE 134

127 REGRESSION COEFFICIENTS. ASSOCIATED T-STATI5TICS OEPENOENT VARIABLE = CAS VAK l*DL.tl — G 7 1 l_ P T N 1 U • 1 I r U 5 A f9 A LOFT ~6 • 5 8 1 6 ^ 3 2 9 A LO w t ~6 •2 267 *" 0 .9^9 LE GG — 321 625 —4.375 LALC -2.0 593 — 1 .073 LORT 3*6 422 1.648 LOMK — 0.5 9-^5 -0 .2 75 LB AN 6*51 72 1.419 LCHE -7. 1 020 -0.818 LOLG -38 .8950 -5.596 LCVG 1 .9792 0.722 LPQT \ 3.9 739 2.115 LNUT 13.9716 4.639 LOBK -1 .8724 -0. 997 LOG I NC 13.2277 6.506 LQGMEM 30.2 787 1 0.022 CONSTANT -361 .6895 -1 1 .970 THE MEAN SQUARED ERROR, USING E(Y) FOR PREDICTED VALUES, IS = 1661.3970? THE MEAN ABSOLUTE ERUOR IS = 7.44377 THE PREDICTED PRCflABILITV OF Y > LIMIT GIVEN AVERAGE XII) = 0.6235 THE OBSERVED FWEQUENCY OF Y > LIMIT IS = 0.7261

PAGE 135

128 ReCRESSION COEFFICIENTS, ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = OFT V AW ( A t3L_Jl 1 — 5TATI STIC — 3 1 #73 5 LO M T — 0 0 628 0 0 1 8 LEGG — 6 •B 24 5 — 2 69 7 LA LC —5 •0410 5 32 C LORT — 0 I 677 — 0 • 1 5 1 LO — 5 • 2 554 —4 • 980 LB AN 2 • 9 359 1.2 79 LCHE -0a970 -0.211 LOLG 5,4623 1 .=37 LC VG -2,415a -1 .790 LPOT -9.5 116 -2 .899 LNUT -2.2 145 -1 495 LOBK -3, 7905 -4. 1 45 LG6INC 24.0723 22.776 LOGMEN -2.91 96 -I .923 CONSTANT -131. 9559 -12.841 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = 377.201S9 THE MEAN ABSOLUTE ERROR IS = 4.20680 THE PREDICTED PROBABILITY OF Y > LIMIT GIVEN AVERAGE X LIMIT IS = 0.6794

PAGE 136

129 REGRESSION COEFFICIENTS, ASSOCIATEO T-STATISTICS DEPENDENT VARIABLE = QMT VARI ABLE COEFP tC I ENT T-CTATIST tC LOMT -21 .232 L8EF 0 1 561 0 263 LPOR 0.9 055 1.213 LEGG 2.3 1 54. 3. 775 LALC -0.1 618 -0 .924 LORT 0. 4366 2.069 LOMK -0 1 906 -0.968 LB AN -0. 7216 1 .697 LCHE 0.5243 0 674 LOLG 3.0 965 4. 544 LC VG -0.0943 -0.373 LPOT 0.3317 0. 554 LNUT -0.4873 1 .774 LQBK 0. 1 905 1 .099 LOG INC 5.143 1 25. 286 LOGMEM -0. 9025 -3. 1 85 CONSTANT -15.791 1 -5.977 THE MEAN SQUARED EKRQCi. USING E(Y) FOR PREDICTED VALUES. IS = 23. THE MEAN ABSOLUTE EBRCR IS = 0.68973 THE PREDICTED PPOBABILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.4860 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.5947

PAGE 137

130 REGRESSION COEFF tC irNTS, ASSOCIATED T: T A T I S T I C S DEPENDENT VARIAeUE =EGG VAl) I ABLE COEFF I C r ENT 1 — S T AT I ST IC LEGG 2 7 .A 606 — 8 • 357 LALC — 1 • 0 538 1 069 LORT -1.7 SA2 — I • 53 1 LOMK ~" 2 • 1 799 1 .985 LB AN — 3 • 8 2 05 -1 .670 UC HE — 6 • 4 232 -1 .461 • LOLG 3.5880 0,996 LC VG 1 .0 320 0.740 LPOT 3.5893 1 .069 LNUT -0.9 A48 -0.621 UOBK -0.7931 -0.830 LOGI NC 34.6 437 31 .697 LOGMEM 0.0477 0.031 CONSTANT -226. 753 1 -15.717 THE MEAN SQUARED ERROR, USING E LIMIT GIVEN AVERAGE X(I) = 0.7190 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.7298

PAGE 138

131 REGP5SSI0N COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = ALC VA/^ I ABLE CCE FF I C I ENT T-STAT I STIC LALC 1 2 .A 935 -21. 996 LORT 1 € 9658 1.995 LOMK 1 J725 1 .560 LB AN -0.1512 -0 C 70 LCHE I .6575 0.5*1 LOLG 5.6997 1 .768 LCVG -2 .3 960 -2 .282 LPCT 1 .0669 -0. *05 LNUT -1.1 026 -0.892 LOaK -1 .2**8 -1 .698 LOG I NC 1 3 .9*69 15.59* LOGMEM -5.6 2** -*.203 CONSTANT -61 .8951 -6.298 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES, IS = 47.03200 THE MEAN ABSOLUTE ERROR IS = 0.*2*10 THE PREDICTED PRCeASILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.0766 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.1329

PAGE 139

132 REGPeSStGN COEFFICIENTS. ASSOCIATED T= T A T I ST I C S DEPENDENT VARIASLE =ORT VAR I ABLE COEFF ICI ENT T-STAT I STIC LORT -56.7857 -28.f567 LCAS -7.3 136 -2.037 LOMK -6. 1 82 -3.120 UBAN -e.oeio -1.959 LCHE -6 .0978 -0.762 LOLG 2.2827 0. 349 LC VG -2.831 9 1 138 LPOT 9.6588 1 .ei 0 LNUT -3.6683 1 344 LOBK -2.5749 -1.502 UOGINC 11.5121 6.24 1 LQGMEM 20.2 870 7. 254 CONSTANT -233.3558 10.302 THE MEAN SQUARED ERROR. USING e LIMIT GIVEN AVERAGE X{I) = 0.4825 THE OBSERVED FPEQUENCY OF Y > LIMIT IS = 0.5585

PAGE 140

133 PEGPfSSION COEFFICIENTS. ASSOCIATED T-STATISTICS OEPSNOENT VARIABLE =OMK VARIABLE CGEFFTCIENT T-STAT 1 STIC LOMK 1 0.a227 -25. 684 LFMK S.l 845 3 .359 LB AN I • 9 380 1.804 LCHE 0.1 270 0.068 LOLG -0.6085 -0.366 LCVG -0.3815 -0.616 Lf QT 1 3541 -0. 922 LNUT 0.6294 0.920 LOBK -2.3624 -6 ; 44 LOGINC 6.4315 1 3. 554 LOGMEM -2.7249 -3.843 CONSTANT -34.8294 -6.812 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = 28.3S115 THE MEAN ABSOLUTE ERRCR IS = 0.14549 THE PREDICTED PROBABILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.1923 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.2339

PAGE 141

134 REGOESSION COEFFICIENTS, ASSOCIATED T-STATISTICS DEPENOENT VAPIABLH = BAN VAR t ABLE CCE FF I C [ ENT T — CT \T J CTT r~ LS A N — Q 7 "~ 1 6 • I 6 1 t~lCT l_Ur 1 o • ^: y 1 LCHE s. 1 647 0.664 LCLG -26. 6 116 -4 330 LC VG -2. 8239 -1 165 LPOT 3. 5325 0. Sq3 LNUT 0. 7 4 5?. 0.232 LCBK -3. 2267 -1 .960 LOG INC 1 1. 5508 6.589 LOGMEN 10. 64.48 3 .943 CONSTANT -262 .4132 1 3 .684 THE MEAN SQUARED ERROR. USING E(Y) FCP PREDICTED VALUES. IS = 1 220.6c5<; THE MEAN AaSOLUTE ERROR IS = 2.71464 THE PREDICTED PROMAHILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.3S31 THE OaSERVeO FREQUENCY OF Y > LIMIT IS = 0.4271

PAGE 142

135 REGRESSION COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = CHE VARIABLE CQEFFICI ENT T-STATISTIC ™ 1 i • VJ 1 O LOLG 2.3263 4.891 LCVG -0.2888 -1.729 LPOT -0.6793 -1.698 LNUT 0.4319 2.181 LOBK -0.0589 -0.516 LOG INC 4.3 064 30.922 LOGMEM -1.62S3 -8.218 CONSTANT -19.0405 -15.793 THE MEAN SQUARED ERROR, USING E LIMIT GIVEN AVERAGE X(I) = 0.2429 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.3005

PAGE 143

136 REGRESSION COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE =OLG COEFF IC I ENT T-STATISTIC LOLG -25.7 100 -14.450 LCVG 0.6393 0.757 LPOT -I .7298 -0.900 LNUT -4.3371 -5. 037 L08K -0.7772 -1 .389 LOGI NC 3.5386 6. 254 LOGMEM 1,7640 1 .991 CONSTANT -69.825A -13.052 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = THE MEAN ABSOLUTE ERROR IS = 0.48944 THE PREDICTED PROBABILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.2779 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.3022

PAGE 144

137 REGRESSION COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = Q V G VAR I AEUE CCEFFICI ENT T-STATISTIC LCVG 0.7989 0.871 LPOT 1 .6887 -0809 LNUT -5.25A0 -5.649 L06K -0.2 050 -0.337 LOGINC 1 .a I 03 3.038 LOGMEN 2.8005 2.931 CONSTANT -37. 6233 -7. 288 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = THE MEAN ABSOLUTE ERROR IS = 0.52868 THE PREDICTED PROBAEILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.2753 THE OBSERVED FREQUENCY OF Y > LIMIT IS = C.3022

PAGE 145

138 REGRESSiaN COEFFICIENTS. ASSOCIATED T-STATISTICS OePENOENT VARIABLE = POT VARIABLE COEFFICIENT T-STATISTIC LPOT -1.6998 -0.815 LNUT -5.2814 -5.682 L08K -0.2000 -0.328 LOGINC 1.7466 2.954 LOGWEM 2.8415 2.977 CONSTANT -37.4833 -7.265 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = 85.25366 THE MEAN ABSOLUTE EBRCR IS = 0.52871 THE PREDICTED PROBABILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.2753 THE OBSERVED FREQUENCY OF Y > L I K I T IS = 0.3022

PAGE 146

139 REGRFSSIQN COEFFICIENTS. ASSCCIATEO T-STATISTICS QEPENDENT VARIASUE = NUT VARIAeLE COEFFICIENT T-STATISTIC LNUT -5.2601 -5.662 LOBK -0.1856 -0.305 LOGINC l.a062 3.078 LCGMEM 2.8 138 2.950 CONSTANT -34.5180 -<9.*34 THE MEAN SQUARED ERROR. USING E(Y) FOR PREDICTED VALUES. IS = THE MEAN ABSOLUTE ERROR IS = 0.52870 THE PREDICTED PROBAEILITY OF Y > LIMIT GIVEN AVERAGE X(I) = 0.2754 THE OBSERVED FREGUENCY OF Y > LIMIT IS = 0.3022 85.24761

PAGE 147

140 REGRESSION COEFFICIENTS. ASSOCIATED T-STATISTICS DEPENDENT VARIABLE = QBK VARIABLE COEFFICIENT T-STATISTIC LOBK -1.8471 -27.096 LOGINC 1.0442 12.344 LOGMEM -0.0265 -0.196 CONSTANT -5.0284 -11.263 THE MEAN SQUARED ERROR. USING E LIMIT GIVEN AVERAGE X(I) = 0.1753 THE OBSERVED FREQUENCY OF Y > LIMIT IS = 0.2357 1 .37228

PAGE 148

APPENDIX 3 ROTATION OF AXE

PAGE 149

Rotation. of Axes: Relationship Between the X,Y and the x,y Coordinate Systems. 142

PAGE 150

143 Rotation of Axes: Relationship Between the X, Y 9P<;i the X, Y Coordinate Systems Assume that the X,Y coordinate system is obtained by rotating the x,y axes through an angle 9 about th.e origin. Then any point P in the plane has coordinates (x,y) and (X,Y) in the two systems. If P is not the origin, let $ be the angle from the positive X axis to the line segment joining P to the origin, and let r be the length of that line segment, then X = r coS(t) and Y = r sintj) using the trigonometric identities for sin ( e+<() ) and cos ( Q+(p ) thus X = r cos (9+41) = r(cos6cos4)sine sincj) ) = (cose)(r costj)) (sin9)(r sin(j)), and X = (cos 6 )X (sin 6 )Y. Similarly y = r sin i Q + ^) r(sine cos^) + cosesincp ) = (sine)(r coS(})) + (cose)(r sin (j)), thus y = (sin 9)X + (cosa )Y. (Robert and Gulick, 1978, p. 640. ) Kakwani and Podder (1976), rotating the coordinate system obtained TT = (1/V2)(F + FJ and n= (1/V2)(F F,).

PAGE 151

APPENDIX 4 RESULTS FROM THE NEW METHOD OF ESTIMATING ENGEL ELASTICITIES

PAGE 152

RESULTS FROM THE NEW METHOD OF ESTIMATING ENGEL ELASTICITIES, FOR RICE Income Class Elasticity $ per month per household 50.00 8.19 100.00 15.12 1 50.00 23.60 200.00 25.08 300.00 21 .51 400.00 5.56 500.00 1 .74 600.00 0.78 800.00 0.87 1 ,000.00 .25 1 ,500.00 -0.47 2,000 .00 -0.01 145

PAGE 153

I REFERENCES Aitchison, J., and J. A. C. Brown. The Loenormal Distrilpution with SpgciaJ. Reference Jlq its Uses JLd Economics. Cambridge, England: Cambridge University Press, 1 957. Allen, R. G. D. and A. L. Bowley. Family Expenditure. London: Staples Press, 1935. Banco Central de la Republica Dominicana. Estudio Sobre PregypuestQS Familiares. Santo Domingo, Dominican Republic: Banco Central, 1971 Banco Central de la Republica Dominicana. Boletin Mensual Santo Domingo, Dominican Republic: Banco Central, Dec. I98I. Blaylock, J. R. and D. M. Smallwood. "Engel Analysis with Lorenz and Concentration Curves. American Journal of Agricultural Economics. 64 (1 982 ): 1 34-39 Brown, Alan and Angus Deaton. "Surveys in Applied Economics: Models of Consumer Behavior." The EconopijQ Journal. 82( 1972) :1 1 45-1236. Chapernowne, D. G. Uncertainty and Estimation jji EconQmjQs, vol. 2. Edinburgh: Oliver and Boyd, I969. Davis, Carlton G. "linkages Between Socioeconomic Characteristics, Food Expenditure Patterns and Nutritional Status of Low-Income Households: A Critical Review." Affieriqan JQUrnal ^ Agricultural Economics. 64(1982) :1017-1025. de Janvry, A., J. Bieri and A. Nunez. "Estimation of Demand Parameters under Consumer Budgeting: An Application to Argentina." Apierican Journal ^ Agricultural Economics. 54( 1972) :422. Del Rosario, G. and P. Musgrove. Funciones ^ Consumo para ^J ^nes Aliment iQjosj. Un P rime r Anaiisis sl^ 1^ Datos ^ ^^J^F. Santo Domingo, Dominican Republic: Banco Central 1 980 Dowler, Elizabeth A., Philip R. Payne, Young Ok Seo, Anne M. Thomson and Erica F. Wheeler. "Nutritional Status Indicators." Food Policy. 7(1982): 99-112. 146

PAGE 154

147 Ellis, Robert and Denny Gulick. Calculus with Analitic Geometry New York: Harcourt Brace Javanovich, Inc., 1978. Friedman, M. "Professor Pigou's Method for Measuring Elasticities of Deraand'from Budgetary Data." Quarterly Journal of Economics. 50 ( 1 93 5 ) : 1 5 1 Frisch, R. "A Complete Scheme for Computing all Direct and Cross-demand Elasticities in a Model with Many Sectors. Econometrica. 27 ( 1 959 ) : 177 George, P. S. and G. A. King. Consumer Demand for Food In the United States with Pro lections Xar 1980. Berkely : Gianninj. Founaation Monograph No. 26, University of California, 1971. Hassan, Z. A. and S. R. Johnson. Family Expenditure Patterns in Canada A Statistical Analysis of Structural Homogeneity Ottawa, Canada: Economic Branch Publication No. 76/3, Agriculture Canada, 1976. Hassan, Z. A. and S. R. Johnson. Urban Food Consumpt ion Patterns in Canada An Empirical Analysis. Ottawa, Canada: Economics Branch Publication No. 77/1, Agricuture Canada, 1 977. INCAP-ICNND. Food Composition Table for Use In Latin America. Bethesda, Maryland : National Institutes of Health, 19b1 Kakwani, N. "A New Method for Estimating Engel Elasticities." Journal of Econometrics. 8 ( 1 978 ) : 1 03-1 1 0 Kakwani, N. and N. Podder. "Efficient Estimation of the Lorenz Curve and Associated Inequality Measures from Grouped Observations." Econometrica. 44 ( 1 976 ) : 1 37-1 48 Leser, C. E. V. "Family Budget Data and Price Elasticities of Demand." Review of Economics Studies. 9(19^1-42). Leser, C. E. V. "Forms of Engel Functions". Econometrica 31 ^ iyo3) :694-703. Le-Si, Vihn and Carlos Poraareda. Direct and Cross Price Elastigitie? from Expenditure Elasticities : Some Estimates for Food in Zambia. Draft. Washington, D. C. : Development Research Center, International Bank for Reconstruction and Development, June, 1976.

PAGE 155

148 Lluch, C. and R. Williams. International Patterns in the E l gst i g iti es SlL the Marginal Utility ^ Income and Expenditure. Discussion Paper 11. Washington D. C. : Development Research Center, International Bank for Reconstruction and Development, Aug., 1975. Maddala, G. S. Econometrics. New York: McGraw-Hill, Inc., 19Y7. Moussie, Menwouy ellet. "Impact of Socioeconomic Characteristics on the Nutritional Status of Adolescents from Low-Income Households in Florida." Ph. D. dissertation, University of Florida, 1 98 i Nelson, Forrest D. "On a General Computer Algorithm for the Analysis of Models with Limited Depenaent Variables." Anpgls SlL E con omic msl Social Measurement. 5 ( 1 976 ; 493-509 Oficina Nacional de PI anif icacion (ONAPLAN). Situacion Mimentgria X Nutricional ils 1^ Republica Dominicana. Santo Domingo, Dominican Republic: PLANDES 37, ONAPLAN, 1978. Oficina Nacional de Planif icacion (ONAPLAN). Plan Trienal Jie InversiOPes Pyb l ica s, 1980-82. Santo Domingo, Dominican Republic: PANDES 42, ONAPLAN, 198Ua. Oficina Nacional de Planif icacion (ONAPLAN). Sueerencias de Pg'^tas Bas i c as para Jms Pc litic a Nacional As Alimentacion x Nutriciont (Documento Preliminar para Discusion). Santo Domingo, Dominican Republic: ONAPLAN, 1980b. Phlips, L. Applied Consumption Analysis. New York: North Holland Publishing Company, 19-74. Pollak, R. A. "Additive Utility Functions and Linear Engel Curves." l^view of Economic Studies. 38 ( 1 97 1 ) : 40 1 -1 3 Prais, S. J. "Non-linear Estimates of Engel Curves." Review Sil Economic Studies. 20( 1 953):87. Prais, S. J. and H. S. Houthakker. Ihs Analysis ^ Family Bi^dgetg. Cambridge: Cambridge University Press, 1 955 Quezada, N. A. "Endogenous Agricultural Price and Trade Policy in the Dominican Republic." Ph. D. dissertation, Purdue University, 198i.

PAGE 156

149 Secretaria de Estado de Agricultura (SEA). Diagnostico j Estrateeia ^ DesarrQUo Agr opecua r io 1976-86. Santo Domingo, Dominican Republic: SEA, 1976. Secretaria de Estaao de Agricultura (SEA). Plan As. Desarrollo Agropecuario 1980-82. Santo Domingo, Domincan Republic: SEA, 1979. Secretaria de Estado de Agricultura (SEA). Plan Operativo 1982. Santo Domingo, Dominican Republic: SEA, I981. Tobin, James. "Estimation of Relationships for Limited Dependent Variables." Econometr ica 26 ( 1 958 ): 24-36 United Nations, Food and Agriculture Organization. Agricultural Commodity Projections, 1970-19gO, Rome: FAO, 1971 US Department of Agriculture, IDS/SAIG. "Interin Report on Sector Analysis Project Model of Agriculture in the Dominican Republic." (Preliminary draft), December, 1977. US Department of Agriculture. Alternative Futures for World Food in 1985. Washington, D.C. : ESCS For. Agr. Rep. 146, April 1978. Whitney, E. and E. M. N. Hamilton. Understanding Nutrition New York: West Publishing Company, 1977. Wold, H and L. Jureen. Demand Analysis. New York: Wiley, 1953. Working, H. "Statistical Laws of Family Expenditure." Journal of the American Statistical Association. 38(1943): 43-56. World Bank. Dominican Republic — Its Main Economic Development Problems. Washington D. C. : World Bank, 1978.

PAGE 157

BIOGRAPHICAL SKETCH The author was born on June 4, 19^9, in Santo Domingo, Domincan Republic. In 1972 he received the degree of Bachelor of Arts, with honors, in economics, and in 1974 received his Master of Economic Planning, both from the University of Puerto Rico. During the year 1973 he worked part time at the Puerto Rico Planning Board as an economic consultant. In 1974 he joined the Secretariat of State for Agriculture of the Dominican Republic where he served as an Assistant to the Under-Secretary of Agriculture for Sectoral Planning, as Coordinator of the Sector Analysis Project, as Director of the Division of Agricultural Planning, and as Associate Director of the Department of Statistics and Information. From 1975 to 1979 the author worked as part time professor in the universities: Universidad Nacional Pedro Henriques Urena, Universidad Central del Este, and Institute Tecnologico de Santo Domingo. In September 1979, the Secretariat of State for Agriculture of the Dominican Republic awarded him a 150

PAGE 158

151 scholarship to pursue a Ph. D. degree in food and resource economics at the University of Florida. He is married to Nellie and they have three children.

PAGE 159

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Max R. Langham, Chairmam Professor of Food and Resource Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Woodrow W. McPherson Graduate Research Professor of Food and Resource economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. / Carlton G. D^is Professor of Food and Resource Economics

PAGE 160

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. 'Gustavo A. Antonini Professor of Latin American Studies and Geography This dissertation was sutxnitted to the Graduate Faculty of the College of Agriculture and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. April, 1983 Agridtrlti Dean/ College of Agriculture Dean for Graduate Studies and Research