Export supply, import demand and trade policy


Material Information

Export supply, import demand and trade policy
Physical Description:
Scott, Samuel W., 1956-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 26485189
oclc - 36725195
System ID:

Full Text








I am indebted to the chairman of my supervisory committee, Dr. John J.

VanSickle. He is the individual after my parents and immediate family who literally

invested in my education. Since the summer of 1989, he accepted the risk in believing

in my abilities and has provided guidance, mentoring, encouragement, and direction in

all my research at this University, including the preparation of my dissertation. He was

always open to my maverick ideas and was willing to steer me to full potential. My

committee was selected based on the specific roles that they were able to play in the

preparation of this work. I am extremely grateful to Drs. Mark Brown, Robert Emerson,

Max Langham and Edward Zabel for their assistance in keeping me in line with standard

economic theory, applied principles and econometric computations. I would like to

acknowledge suggestions and comments from Dr. Ron Ward at various stages of the

research. Their contributions cannot be valued sufficiently; these individuals helped to

keep me focused and aided in clarifying and consolidating the research issues. I am very

thankful for financial support provided by the Food and Resource Economics

Department/IFAS, and directly for the assistantships provided by Dr. VanSickle; this

made it much easier to remain in the program.

Finally, I cannot repay my family for their support especially in moments when

I felt it would be better to have given up. To Marcia, Sadeeka and Samae sincere thanks

for all the support provided. To my friends, too numerous to enumerate here, thanks.



ACKNOWLEDGMENT....................................... i

ABSTRA CT ........................................... vi


1 INTRODUCTION ................................... 1

Overview of the World Coffee Market, Trade and Pricing Patterns ..... 2
Problem Statement ................................... 5
H ypotheses .. ............. ..... .... ....... ........ 9
O bjectives ......... .. .......... ... ... ..... ... ..... 10
Methodology, Scope and Organization of the Study ................ 14


Major Exporters and Market Shares in the U.S. Coffee Market .....
Price Trends and Pricing Patterns in the U.S. Coffee Market .......
Marketing Channels and Distribution Systems ................
Import Channels ..................................
Marketing Channels .................................
Market Structure, Concentration and Commercial Entities .........
Milestones in the Development of the World Coffee Market
and ICO/ICA .....................................

. . 17
... 18
. . 26
.... 28
.... 30

. .32

3 LITERATURE REVIEW ...........................

. ...... 35

Institutional Framework and International
Commodity Agreements ............................... 35

Theoretical Framework ................................ 41
Market Structure and Export Quotas ......................... 41
Market Demand and Structural Changes ....................... 43
Previous Studies ......................................... 47
Descriptive Analyses ..................................... 47
Empirical Analyses ...................................... 55
Summary of Previous Coffee Demand Studies in the U.S ............ .63


Conceptual Framework for Aggregate Supply and Import Demand ...... 65
Modeling Aggregate Export Supply and Import
Demand of Coffee Traded ............................ 70
General Export Supply and Import Demand Functions .............. 71
Aggregate Supply Equation ........................... 73
Aggregate Import Demand Equations for the U.S
and Rest-of-the-world ............................. 73
Modeling Conditional U.S. Import Demand for Coffee by Type and
Importers' Preferences ............................... 74
Estimated Equations and Proposed Models .................... 78
Estimation M ethods .................................. 79
D ata . . . . . . . . . . . . . . . . . . . . . . 83


Empirical Results .................................... 84
Aggregate Export Supply and Import Demand for Coffee ......... 84
Test for Evidence of Structural Changes in the Aggregate Model ..... 90
Conditional U.S. Import Demand by Coffee Types .............. 91
Test for Structural Changes in the Disaggregate Model ............ 95
Discussion and Implications ............................. 98

6 SUMMARY AND CONCLUSIONS ....................... 106

Sum m ary ....................................... 106
Conclusions ...................................... 109
Suggestions for Further Research ......................... 111

REFERENCES ........................................ 113

BIOGRAPHICAL SKETCH ................................. 119

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



Samuel W. Scott

December, 1996

Chairman: Dr. John VanSickle
Major Department: Food and Resource Economics

World demand for coffee has not grown at the same pace as production. The

concern for stabilizing prices and export revenues, and reducing over-production over the

long term led to the formation of the International Coffee Organization (ICO) and the

signing of the first International Coffee Agreement (ICA) in 1962. The purpose of the

ICA was to stabilize prices by controlling supply through the use of export quotas. The

absence of incentives for cooperating have resulted in ICO member countries not

deciding on a new agreement with economic provisions since the spring of 1993.

United States, being the major consuming market, presents an interesting case for

studying the effects of the export quota system, structural changes and demand in the

market given the fact that it withdrew its membership from the ICO in 1993.

The general objective of this study is to assess the impacts of the ICA and the

behavior of exporters and importers of coffee given the effects of the ICA export quota

system. At the aggregate level the study examines the export supply and U.S. and rest-

of-the-world import demand for all coffee traded. The study at the disaggregate level

examines the trading of the four types of coffee imported into the U.S. market. These

include Colombian Mild, Other Milds, Brazil and Other Arabica, and Robusta.

The results suggest that at the aggregate level, annual inventory levels and

domestic production along with spot market price influence export supply decisions. The

aggregate demand model was price inelastic and income inelastic. In the disaggregate

model, expenditure elasticities were statistically significant for all types of coffee;

however, they were elastic for Colombian Mild, Brazil and Other Arabica and for

Robusta, but were inelastic for Other Milds. Own price elasticities were statistically

significant for Other Milds and Robusta. Cross price elasticities for some (6)

combinations were statistically significant, where the premium Arabicas (Colombian Mild

and Other Milds) were complements with the lower grade coffee types (Brazil and Other

Arabicas, and Robusta). In addition, substitute relationships were seen among the

premium coffee types as a group and among lower grade coffee types as a group.

Importers preferences were seen for Colombian Mild and Robusta. The results suggest

that structural changes have occurred in both the exporting and importing countries.



Coffee is grown in the tropics and is considered a perennial shrub. Trees take

three to four years to bear and have a productive life of at least 20 to 30 years. Coffee

can be harvested twice per year (especially for those plantations located at the equator),

but is generally considered a single-harvest crop. When ripe, the berries are picked by

hand; however, in Brazil some harvesting is done by machine. The seeds of the berries

are removed by a process called "pulping" and are then dried and shipped as "green"

coffee beans. Value-added activities including roasting, grinding and blending are

performed in importing countries to yield the rich flavored coffee that consumers


Historically, coffee was a plantation crop. Currently, 98 percent of the world's

coffee is grown by small farmers and two percent by plantations (Bread for the World,

1990). It is estimated that worldwide coffee production employs some 20 million people.

The two main varieties grown and traded are Arabica and Robusta. Arabica is preferred

for its flavor and commands a higher market price. It is grown mainly in the highland

areas, with the major producing countries being Colombia, Costa Rica, Kenya, Tanzania,

Jamaica and Haiti. Large amounts of another type classified as unwashed Arabica are

grown in Brazil. Robusta coffee is native to sub-Saharan Africa, but is grown

commercially in Latin America and Asia. It grows better in higher temperatures and


lower altitudes. The increase in demand for instant coffee and blends has stimulated

rapid expansion in the production of Robustas. Coffee is further classified into four

types by the trade industry: 1) Colombian Milds; 2) Other Milds; 3) Brazilian and Other

Arabica and; 4) Robustas. Coffee is sold in the wholesale market as green beans, roasted

beans, soluble and extracts, and instant coffee.

Overview of the World Coffee Market. Trade and Pricing Patterns

Based on value, coffee is the leading primary agricultural commodity in world

trade, second in total trade only to petroleum oil. It is a major source of foreign

exchange for over 50 countries. For example, Colombian coffee accounts for 50 percent

of the legal foreign exchange of that country. It accounts for 95 percent of the foreign

exchange for Uganda. Colombia and Brazil are the world's largest coffee producing

countries, together accounting for 40 to 50 percent of the world supply. Other large

exporting countries include Costa Rica, Kenya, Guatemala, Mexico, Cote d'Ivoire,

Indonesia, Uganda, India, Ethiopia and El Salvador.

The total annual supply of green coffee beans ranged from 97 to 155 million bags

(1 bag = 60 kgs) over the period 1960 to 1994 (Table 1), with a thirty-five year average

of 124 million bags. Exports over the same period ranged from 42 to 80 million bags,

with a thirty-five year average of 61 million bags. Domestic stocks have fluctuated

widely, ranging from 25 to 86 million bags, with an average of 60 million bags.

As a result of its importance, governments in exporting countries play a critical

role in production, trade and marketing policies for coffee. Their roles include provision

of credit, quality control, monopoly over milling and marketing, transportation networks,

Table 1. World Supply and Distribution of Coffee and U.S. Imports, 1960 1994

Beginning Total Total Total U.S. Ending
Stocks Production Supply Exports Imports Stocks

Year --------------- thousands of 60 kg bags-------------

1960 59,542 65,341 125,222 42,863 16,428 65,931

1961 65,931 75,878 142,113 46,141 21,068 74,904

1962 74,904 67,781 142,973 46,905 14,773 81,295

1963 81,295 65,295 146,853 51,066 17,769 78,018

1964 78,018 52,612 130,847 41,924 16,743 72,180

1965 72,180 82,108 154,657 50,002 17,769 86,811

1966 86,811 63,298 150,529 48,996 19,317 82,216

1967 82,216 70,688 153,297 55,689 18,106 79,502

1968 79,502 63,214 143,080 53,642 19,680 69,758

1969 69,758 69,623 139,768 55,297 18,931 65,540

1970 65,540 59,202 125,228 51,694 19,408 54,126

1971 54,126 73,598 128,202 58,715 19,075 50,412

1972 50,412 77,060 127,914 61,221 17,502 49,191

1973 49,191 65,717 115,476 60,609 19,045 35,822

1974 35,822 82,651 118,941 55,387 19,204 44,350

1975 44,350 72,970 117,745 59,549 19,212 38,984

1976 38,984 61,162 100,680 56,561 18,452 25,667

1977 25,667 70,724 97,018 48,755 18,828 29,435

1978 29,435 79,018 109,131 64,612 19,462 25,057

1979 25,057 81,906 107,616 62,130 19,963 25,523

1980 25,523 86,174 112,372 59,787 20,463 32,122

1981 32,122 98,152 131,029 65,340 21,056 44,633

1982 44,633 82,074 127,440 65,454 20,620 41,366

1983 41,366 88,975 130,947 68,202 21,080 41,665

1984 41,665 90,508 132,629 72,140 22,968 37,521

1985 37,521 95,837 133,755 70,121 21,386 42,248

Table 1.-- Continue

Year Beginning Total Total Total U.S. Ending
Stocks Production Supply Exports Imports Stocks

1986 42,248 79,549 122,059 66,408 22,201 33,450

1987 33,450 103,285 137,031 67,150 23,074 46,807

1988 46,807 94,363 141,585 70,892 22,165 48,528

1989 48,528 97,286 146,072 83,321 21,536 41,215

1990 41,215 100,417 141,984 76,957 21,806 43,221
1991 43,221 104,245 147,815 80,727 19,697 47,391

1992 47,391 93,405 141,566 77,668 21,328 42,570

1993 42,570 93,538 137,140 77,609 23,997 35,534

1994 35,534 94,306 130,900 77,297 22,655 30,948

Note: numbers may not add up due to rounding-off, shrinkage and slippage.
Source: Foreign Agricultural Service, USDA.

subsidies and taxes. These policies affect the prices received by farmers. It was reported

that over 92 percent of the export price went to Kenyan farmers (Bread for the World,

1990), while only 30.3 percent went to Angolan farmers. Langham and Kamajou (1992)

estimated that only about 20 percent went to farmers in the Cameroon in the 1980's. In

other countries the percentage of the export price going to farmers averaged 58 percent.

In terms of consumption, about 75 percent of the world's coffee is consumed in

the importing countries. The major consumers in the coffee market are the U.S.,

European Community (EC), and Japan. U.S. citizens drink 430 million cups of coffee

per day, an average of 1.7 cups per capita. The annual per capita consumption (based

on green bean equivalent) in the U.S. slipped from 15 pounds to just under 10 pounds

over the period 1976 to 1990 (Bread for the World, 1990). In combination, the U.S.


(the leading consumer), EC and Japan consume over 90 percent of the world's exported

coffee. Over the last ten years the consumption in Japan has grown considerably.

By far the issue of most concern between exporting and importing countries is the

pricing policies on both sides. It is important to note that prices paid by coffee drinkers

in importing countries are substantially above the export price. It is argued that the

additional expense for the final stage of coffee processing is incurred by a few major

firms. This is estimated to be between 26 percent for instant coffee and 24 percent for

roasting of green coffee (Bread for the World, 1990). Concentration has led to an

oligopolistic market structure (Karp and Perloff, 1993; Schneider and Gallagher, 1994).

Since 1979, the top four processing firms controlled over 65 percent of the U.S. coffee

market. In addition, although many of the exporting countries have the capacity to

export roasted or instant coffee, they encounter barriers in doing so. The U.S. currently

has no import duties on any coffee. The EC, however, levies import duties on processed

coffee. Many of the previously centrally-planned economies maintain quotas. By and

large, the developing countries have difficulties competing with the market power of the

major multi-national importing firms (Schneider and Gallagher, 1994).

Problem Statement

Shifts in the world demand for coffee have not kept pace with shifts in supply,

while real prices have fallen. As a result, except in periods of occasional production

problems (e.g., frost in Brazil in 1975) which lead to increases in prices and subsequent

increases in the production of coffee in other producing countries, nominal coffee prices

have not shown wide variations. Responding to the 1975 frost, Brazil invested in its


coffee recovery by subsidizing the planting of higher yielding varieties, but competitors

had done the same. The result was that by 1981 the market for coffee had excess supply

at acceptable prices. From 1981 until 1990, coffee export prices, adjusted for inflation,

dropped by over 40 percent. At the same time consumers in importing countries

continued paying higher prices for imbibing coffee.

The concern for stabilizing prices and export revenues, and reducing over-

production over the long term, led to the formation of the International Coffee

Organization (ICO) and the signing of the first International Coffee Agreement (ICA) in

1962 (Clarke and Macrae, 1988). This agreement allowed for membership from both

consuming and producing countries in the ICO. The purpose of the ICA was to stabilize

prices by controlling supply through the use of export quotas. The ICA operates in two

ways. In certain periods the agreement is implemented with no economic provisions,

i.e., there is multilateral trade without any export quotas or price controlling

mechanisms. In this case it is said to be purely administrative. This was the case over

the periods 1975 through 1979 and July, 1986 through 1994. In other periods, the

agreement implemented economic provisions that included multilateral trading with

export quotas and price controls between producing and consuming countries. This was

the original intent of the agreement and occurred over the period 1963 through 1974 and

1980 through 1986. This system met with some success from 1963 through 1968, but

has since had severe problems. The quota system was discontinued in 1974 due to the

inability of producing and consuming countries to agree on prices and quota levels.

Prices increased to unprecedented levels, for example, $3.79 per kilogram in the U.S.


in 1977 (Akiyama and Varangis, 1989). The World Bank calculated that prices would

have risen even higher without the supplies in inventories created by the quota imposed

by International Coffee Organization/International Coffee Agreement (ICO/ICA).

Declines in prices after 1977 stimulated both the producing and consuming countries into

negotiating a new agreement with an export quota as its major tenet. This agreement

came into effect in 1980 and kept a floor under world market prices for coffee from 1980

to 1986, after which the quota system was suspended. The suspension was due to the

severe drought in Brazil in 1985 which led to reductions in forecasted trade quantities.

Prices continued to decline after this period. These declines led to several attempts to

reinstate the quota system. Finally, in the fall of 1993, the U.S. withdrew its

membership from the ICO, and 28 producing countries formed the Association of Coffee

Producing Countries (ACPC) with an export retention scheme similar to that of the old

ICO/ICA. The ICO formed in 1962 had inherent problems with the given ICA export

quota system and voluntary export restraint (VER). These included concerns that

(i) quotas tend to freeze each country's export share at historical levels without

allowing for changes in purchasers' preferences;

(ii) coffee is sold to non-ICO members at large discounts and sometimes resold to

ICO importing countries at higher prices;

(iii) importing countries commitments to producing countries tend to waiver when

stocks pile up in exporting countries, resulting in the possibility of lower prices;

(iv) there was an inflexibility in allocation of quotas among producing countries and

among coffee varieties;

(v) higher prices were paid by member importing countries and lower prices for

nonmember importing countries; and

(vi) higher prices were the primary objective rather than more stable prices.

The absence of incentives for cooperating have resulted in ICO member countries

not deciding on a new agreement with economic provisions since the spring of 1993.

The agreement reached was an administrative one (i.e., agreement without economic

provision). United States, being the major consuming market, presents an interesting

case for studying the effect of the export quota system, structural changes and demand

in the market given the fact that it has withdrawn its membership from the ICO.

Additionally, the annual per capita consumption of coffee has fallen since 1976. The

coffee industry has experienced 18 years of the ICA agreement with export quotas and

pricing controlling mechanisms, and another 13 years without economic provisions over

the period 1964-1994. For the purpose of this study the first case will be called

(ICA/BC), the agreement under binding conditions and the latter case is the international

coffee agreement under non-binding conditions (ICA/NBC). The periods are as follows:

(1) 1964- 1974 ICA/BC 11 years

(2) 1975 1979 ICA/NBC 5 years

(3) 1980 1986 ICA/BC 7 years

(4) 1987 1994 ICA/NBC 8 years


There are several problems identified in the new ICA (Vogt, 1990 and Dull, 1992

and 1994). After a series of negotiations and meetings over the period 1988 to March,

1993, a new agreement was completed. This was to become effective October 1, 1994,

and run for a five-year period to September 30, 1999. The new agreement did not have

economic provisions (e.g., export quotas). However, under Article 30 of the agreement,

there are provisions for the ICO Council to examine re-negotiation of the agreement with

measures provided to balance the supply and demand for coffee (Dull, 1994).

More importantly, the United States is not a member of the new ICA. In the fall

of 1993, the U.S. informed the ICO that it was withdrawing its membership because

funding for its continued participation was denied due to lack of congressional support

and because business entities in the U.S. coffee industry indicated a reasonably strong

preference for a free coffee market. The potential impacts of the International Coffee

Agreement and whether the U.S. made the correct decision on the basis of political and

economic rationales are empirical questions. While a trade policy variable will not be

introduced in the specification of the demand model (as it is in the export supply model),

the data will be manipulated such that statistical tests can be done to evaluate the

behavior of the market (i.e, evidence of structural changes) in the different trade policy

periods. The following testable hypothesis can be established:

Ho: There is no evidence of structural changes in export supply of coffee given the

ICA with or without the economic provisions of the export quota system.

Ho: There is no evidence of structural changes in the U.S. and rest-of-the-world

import demands for coffee given the ICA with or without the economic provisions

of the export quota system.

Ho: There is no evidence of structural changes in U.S. import demand for coffee by

type given the ICA with or without the economic provisions of the export quota


Of particular interest is the third hypothesis. If we fail to reject the null

hypothesis, then it can be stated that given trade policy, U.S. importers have treated the

market as if it were a free market regime and the ICA was not effective in influencing

the coffee market. Consequently, there was no need for the export quota agreement.

This would have further implications for the ACPC export retention scheme. On the

other hand, if the null hypothesis is rejected, the ICA was effective and therefore from

a policy standpoint violated free market conditions for coffee. This runs counter to U.S.

trade policy and therefore warrants the U.S. withdrawal from the ICO.


The general objective of this study is to assess the impacts of the ICA and the

behavior of exporters and importers of coffee given the effects of the ICA export quota

system. The specific objectives are to:

(i) estimate the aggregate export supply, and U.S and rest-of-the-world (ROW)

import demands for coffee, 1964 1994;

(ii) evaluate structural changes on aggregate export supply, U.S. and rest-of-the-world

import demands for coffee under binding and non-binding export quota periods.

(ii) model U.S. import demand for coffee by type, 1964 1994;

(iii) evaluate structural changes and importers' preferences in the U.S. coffee market,


(iv) estimate conditional expenditure and price elasticities of U.S. imported coffee by


The following rationales underscore the importance of this study. Table 2 shows

that the U.S. coffee market was valued at $3.2 billion in 1993 (Maxwell, 1994). The

value declined $1.7 billion from a high of $5 billion in 1986 when the industry peaked.

The U.S. Department of Commerce reported that the U.S. coffee roasting industry

(including coffee shipments, employment, value added, and capital expenditures) was

valued at $6.2 billion in 1989 (Table 3). Annual per capita consumption fell from 15

pounds in 1976 to just under 10 pounds in 1994. It appears at the wholesale trade level

that an ICA agreement is no guarantee that prices can be stabilized over a sustained

period. This can be seen in the years following the frost and drought in Brazil in 1975

and 1985, respectively. But this has more far reaching effects, particularly as it relates

to the structure of the market on the supply side. The fact that Brazil and Columbia

account for 40 to 50 percent of the world's production, any major natural disaster or

refusal to cooperate on their part will result in nullification of the ICA quota system or

coffee export retention scheme and, hence, fluctuations in prices as seen in 1976 and

1986. It has been suggested that the market structure is oligopoly both on the supply and

demand sides.' There are four major firms in the U.S. that account for 65 percent

This is supported by studies including Geer (1971), Andrews (1992), Maxwell (1994), and
Schneider and Gallagher (1994). However Karps and Perloff (1994) showed empricially that
on the supply side exporting countries behaviors are closer to price-takers than that of market

Table 2. U.S. Coffee Sales, 1977 1993

Year Regular Coffee Instant Coffee Total

Million Million Million Million Million Million

pounds dollars pounds dollars pounds dollars

1977 800 2,582 181 1,464 981 4,046

1978 883 2,626 183 1,602 1,066 4,228

1979 911 2,543 188 1,588 1,099 4,131

1980 911 2,765 181 1,664 1,092 4,429

1981 931 2,360 174 1,433 1,105 3,793

1982 947 2,514 170 1,436 1,117 3,950

1983 951 2,522 165 1,420 1,116 3,942

1984 948 2,690 158 1,419 1,106 4,109

1985 977 2,780 151 1,420 1,128 4,200

1986 909 3,389 141 1,569 1,050 4,958

1987 942 2,807 136 1,347 1,078 4,154

1988 918 2,610 129 1,253 1,047 3,863

1989 920 2,740 123 1,202 1,043 3,942

1990 925 2,750 119 1,165 1,044 3,915

1991 915 2,500 111 1,050 1,026 3,550

1992 900 2,300 104 1,000 1,004 3,300

1993 890 2,240 99 960 989 3,200

Source: Foreign Agricultural Services, USDA.

Table 3.


U.S. Coffee Roasting Industry Shipments, Employment,
Value-Added and Capital Expenditure, 1982 and 1987-1989

Items 1982 1987 1988 1989

Roasted coffee value of 5,827 6,400.60 6,332.40 6,167.20
industry shipments
(million dollars)

Total employment 11.1 10.7 10.7 10.5

Payroll (million dollars) (1) 303 315.6 303.1

Production workers 6.9 6.6 6.7 6.5

Hours worked (millions) {1} 13.6 12.7 13.5

Payroll (millions) {1} 170.5 173.4 172.3

Average hourly wages $10.47 $12.54 $13.65 $12.76


Employees (thousands) {1} 4.1 4 4

Hours worked (millions) (1) 8.5 8.3 8.3

Payroll (million dollars) {1} 132.5 142.2 130.8

Average hourly wages {1} $15.54 $17.09 $15.72

Value added by (1} 2,589.8 2,795.8 2,658.1
(million dollars)

(Value added/ (Industry (1} 0.4046 0.4415 0.431

Value added per {1} $392,294 $417,284 $408,938
production worker

Capital expenditures 80.5 155.2 123.2 120.9
(million dollars)

Capital expenditures per {1} $23,515 $18,388 $18,600
production worker

Source: Census of Manufacturers, U.S. Department of Commerce
Note: (1) Not available

of the coffee imported. This indicates a significant level of concentration and market

power held by these firms in the U.S. The recent passing of three trade agreements

(North American Free Trade Agreement (NAFTA), General Agreement on Tariff and

Trade (GATT) and Mercado Comum Do Sol (MERCOSUR)) begs the question of the

validity of the ICA export quota system and the role that U.S. importers may play in

preferential imports of coffee from Mexico under NAFTA and from other countries,

e.g., Brazil and Colombia.

Implicit in the analysis is the question about the economic rationale for imposing

an agreement that shifts the market into disequilibrium, thereby violating free trade. At

the same time, the major trade agreements are supporting the establishment of free trade

zones (Manzella, 1994 and 1995). In light of major trade assumptions and the vast array

of empirical trade studies done, the market structure imposed in most studies is that of

perfect competition. This study recognizes that this assumption would not be appropriate

given the structure of the market on both sides and the export policy that does not satisfy

the perfectly competitive pricing and trade patterns. The actions of the ICO through the

Cartel and the ICA export quota system have resulted in highly imperfect market

conditions in both the ICO producing and importing countries. A model therefore should

address these issues or at least show the results under different regimes given the

availability of the data.

Methodology. Scope and Organization of the Study

An aggregate export supply and demand model is developed to investigate the

effect of trade policy in the coffee industry. Then a disaggregate model was used to

assess trade policy and import demand for coffee types in the U.S. In the aggregate

model, exporters (producing countries) and importers (U.S. and rest-of-world) were

treated as firms operating in a single industry. In the disaggregate model, an evaluation

is done for each type of coffee imported by firms in the U.S. The models are static

partial equilibrium models which allow for testing of hypotheses. The major assumptions

made to facilitate the estimation of the disaggregate model are separability and two stage

budgeting to facilitate conditionality of the expenditures on coffee. The literature review

includes theoretical, descriptive and empirical studies on commercial trade policy,

international commodity agreements, import demand and market structure.

The study examines the International Coffee Organization's international coffee

agreements with and without economic provisions (export quota) over the period 1964 -

1994. The study at the disaggregate level examines the trading of the four types of

coffee as classified in the market. These include Colombian Mild, Other Mild, Brazil

and Other Arabica, and Robusta.

The study is organized in six chapters. Chapter 1 is devoted to an overview of

the world coffee market and outlines the problem statement, testable hypotheses, goal and

objectives. Chapter 2 presents a general background and market analysis of the U.S.

coffee industry. Additionally, a summary of the major milestones in the coffee market

and ICO/ICA is included. Chapter 3 provides a review of the literature. This covers

commercial trade policy and international commodity agreements, and the theoretical

framework on import demand and market structure. In addition, it provides a summary

of previous studies on U.S. coffee import demand. Chapter 4 highlights the specification

of the model and estimation procedures. The conceptual framework of the models and

the estimation methods and data sources are presented.

Chapter 5 presents the results and discussion and provides insights into the

implications of the results. Chapter 6 provides the summary and conclusions of the study

along with suggested areas for further research.



A summary discussion about the background of the U.S. coffee industry is

presented in this chapter. In general, it covers major exporters and market shares, and

pricing trends and patterns at the wholesale and retail levels. A brief account of the

marketing channels and distribution systems is provided. Additionally, the market

structure, concentration and commercial entities are identified. The major milestones in

the development of the world's coffee market and the International Coffee Organization's

coffee agreements are highlighted. These areas of information provide the requisite

background useful in the development of the analytical model and for interpretation of

the results.

Major Exporters and Market Shares in the U.S. Coffee Market

The U.S. imports coffee from over 29 countries in five regions of the world.

Coffee is imported in the form of green beans (95 percent), solubles (4 percent) and

extracts (1 percent). Importers have been able to shift to different sources of supplies,

particularly in the case of solubles and extracts, to satisfy market requirements.

However, in the case of green beans, which account for the largest volume of imports,

Brazil, Columbia, Mexico, Guatemala and El Salvador remain the major exporters to the

U.S. U.S. coffee imports accounted for an average of 25 percent of the world's total

over the period 1986 to 1993 (Table 4). This has resulted in the U.S. being the largest


destination for world coffee exports. Given the market share of exporters (by value),

higher prices are paid for Columbian Milds and Other Milds than for Brazil Arabicas.

The majority of U.S. imports of green beans are accounted for by the larger coffee

trading companies including merchants and agents. Over the last five years Columbia

accounted for 22 percent of U.S. imports, Brazil 21 percent, Mexico 14 percent,

Guatemala 8 percent, El Salvador 6 percent and the rest of the supplying countries 30

percent. Table 5 shows in detail the exports of coffee by ICO member countries to all

market destinations over the period 1987 to 1993.

Price Trends and Pricing Patterns in the US Coffee Market

The role of the International Coffee Organization (ICO) in developing its price

policy must be understood to interpret price trends and pricing patterns in the U.S. coffee

market. The ICO develops what is known as monthly and annual composite wholesale

indicator prices (ACWIP) based on supply region, grades and variety of coffee. These

are reported on a cents per pound basis in the New York spot market. As shown in

Table 6, highs in current prices were seen in the 1970's with annual average prices

ranging from $0.91 to $3.47 per pound. Annual average prices then declined to a range

of $2.53 to $3.45 per pound in the 1980's. Current prices rose 51 percent from 1964

to 1974, 98 percent from 1975 to 1984 and 18 percent from 1985 to 1994. Prices for

coffee based on type and using the ACWIP were higher for Brazil and Other Arabica.

There are several tiers of pricing at the wholesale and retail levels. Wholesale

prices for U.S. roasted coffee (coffee that is traded and packaged as regular, not gourmet

quality) have fallen steadily since the suspension of the ICA in 1989, from $6.64 per kg


in 1987 to $5.94 per kg in 1992, with a low of $4.65 per kg in 1989 (Table 7). The

average price for roasted coffee ranged from $6.34 per kg to $6.86 per kg over the same

period while soluble coffee ranged from $18.63 per kg to $20.17. An area of growth

in the industry is the specialty coffee market. This supports considerably higher

wholesale and retail prices, with specialty coffee commanding $6 to $12 per pound at

retail. Some coffees, such as the Jamaica Blue Mountain, sell for up to $25 per pound

in retail gourmet shops.2

Another factor that influences pricing trends at retail is the buffer stocks held by

consuming countries over time. Since the suspension of the ICA quota system in July,

1989, the U.S. coffee buffer stocks have risen. As of March, 1983, the U.S. alone

maintained 50 percent of the stock position of all the importing countries of the world,

despite accounting for only 25 percent of world consumption. This situation came about

as a result of the continued depressed prices for coffee traded in the United States.

Prices have remained flat as U.S. roasters and processors draw upon existing stockpiles.

Annual marketing margins between import and exfactory prices over the period 1987 -

1992 ranged from $3.56 to $3.96 per pound (calculated from Tables 6 and 7). However,

some industry analysts have suggested that the slim margin between wholesale

2 Specialty or gourmet coffee generally refers to Arabica beans grown at high altitudes which
produces an especially rich aromatic, full bodied brew. This class of coffee commands a
premium price that is based on cup quality and availability. A true specialty coffee reflects
careful selection of variety and cultivation by the growers, proper processing, grading, storage
and shipping by the mill owner and exporter, conscientious roasting, blending and grading by
the retail/wholesaler, and marketing, packaging, and presentation by the retailer that maintains
optimum freshness and appeal (Schneider and Gallagher, 1994).


Table 4. World Coffee Consumption by Importing Countries, 1986 1993

------------------------------------ thousands of 60 kg bags----------

Countries 1986 1987 1988 1989 1990 1991 1992 1993

Europe (EU)

Belgium/ 1,222 1,187 1,210 1,037 391 600 1,019 786

Denmark 939 911 872 918 865 909 959 831

France 5,067 5,404 5,384 5,290 5,203 5,557 5,614 5,506

Germany 8,707 9,572 9,677 9,881 9,079 10,477 10,771 10,706

Greece 363 500 528 581 631 387 320 96

Ireland 107 94 109 97 119 113 91 98

Italy 4,168 4,308 4,216 4,314 4,859 4,228 4,130 4,930

Netherlands 2,342 2,560 2,447 2,244 2,553 2,486 2,549 2,382

Portugal 282 450 422 448 521 523 551 615

Spain 2,224 2,106 2,342 2,592 2,713 2,652 3,044 2,728

United 2,282 2,355 2,331 2,177 2,348 2,342 2,516 2,534

Total EU 27,703 29,447 29,538 29,579 29,282 30,274 31,564 31,212

Austria 977 1,032 1,015 1,337 1,340 1,302 1,212 1,335

Bulgaria 89 95 162 85 102 17 175 0

Cyprus 15 36 49 38 34 47 37 75

Czechoslovakia 522 585 573 586 659 525 472 0

Finland 987 1,036 964 1,058 1,070 965 1,030 1,128

Hungary 746 693 825 443 574 533 588 0

Table 4. -- Continued

Countries 1986 1987 1988 1989 1990 1991 1992 1993

Norway 704 755 643 711 727 757 736 690

Poland 551 560 620 824 335 243 1,650 0

Romania 120 203 152 157 395 269 341 0

Soviet Union 913 1,215 1,144 2,639 1,080 933 700 0

Sweden 1,624 1,640 1,559 1,549 1,690 1,603 1,634 1,619

Switerzerland 714 775 855 935 908 949 995 869

Yugoslavia 897 982 829 841 1,032 638 833 0

Others 81 80 78 86 91 86 90 0

Total Europe 36,643 39,134 39,006 40,868 29,319 39,141 42,057 41,000

Algeria 598 1,849 760 1,738 1,040 1,763 1,064 0

Argentina 546 565 575 500 494 559 662 0

Australia 600 762 648 667 636 646 771 756

Canada 1,786 1,800 1,814 1,822 1,974 2,068 1,916 2,117

Fiji 1 1 1 1 1 1 1 1

Israel 244 233 317 272 272 331 370 0

Japan 4,506 4,963 5,087 5,100 5,236 6,038 5,272 5,889

Korea, North 31 53 222 75 0 0 0 0

Korea, South 349 407 513 695 842 801 942 0

Lebanon 80 130 105 97 164 264 347 0

Malaysia 4 100 323 151 150 172 184 0

Morocco 159 216 296 318 360 334 384 0

New Zealand 109 117 107 115 133 126 132 0

Saudi Arabia 285 385 305 176 259 266 219 0

Singapore 0 0 0 0 72 0 323 129

South Africa 259 245 229 268 276 280 280 0

Table 4.--Continued

Countries 1986 1987 1988 1989 1990 1991 1992 1993

United States 17,572 18,197 17,889 18,544 18,974 18,911 17,909 18,287

Others 765 979 1,301 1,294 1,150 1,438 1,649 N.A.

Total 27,887 31,002 30,592 31,833 32,033 33,998 32,334 33,000

Grand Total 64,530 70,136 69,598 72,701 71,352 73,139 74,391 74,000

Source: International Coffee Organization (Unpublished Data).

Table 5. Exports by ICO Exporting Members to all Destinations, 1987 1993

-------------------------------------------in thousand 60 kg bags --.----

Exporting 1987 1988 1989 1990 1991 1992 1993

Members entitled
to basic quota
Total Colombian (11,034) (11,033) (11,032) (11,031) (11,030) (11,029) (11,028)

Colombia 9,111 10,271 13,738 12,212 15,467 14,529 12,692

Kenya 1,292 1,678 2,020 1,649 1,399 1,412 1,412

Tanzania 631 835 948 947 826 1,009 654

Total Other (14,756) (19,089) (21,942) (17,856) (18,855) (20,861) (19,120)

Costa Rica 1,954 2,157 2,377 2410 2,275 2,662 2,056

Dominican 424 466 561 451 318 443 262
Republic _

Ecuador 1,192 1,789 1,632 1,528 1,229 1,516 1,941

El Salvador 1,878 1,693 2,591 2,018 2,176 2,991 2,065

Guatemala 2,222 2,870 3,491 2,803 3,288 4,018 3,140

Honduras 1,224 1,464 1,734 1,427 1,800 1,897 1,661

Table 5. -- Continued

Countries 1987 1988 1989 1990 1991 1992 1993

India 1,387 1,918 2,026 1,511 2,024 1,820 2,205

Mexico 2,549 3,740 4,389 3,526 3,290 3,207 3,297

Nicaragua 566 605 637 420 593 517 587

Papua New 643 1,282 1,134 776 899 1,027 1,118

Peru 718 1,105 1,370 985 963 764 787

Total (18,160) (17,895) (18,721) (20,232) (21,907) (19,253) (18,295)

Brazil 16,783 16,494 17,339 19,382 21.238 18,093 16,963

Ethiopia 1,377 1,400 1,382 849 669 1,160 1,332

Total Robustas (15,838) (18,487) (19,355) (18,287) (15,739) (18,279) (15,301)

Angola 205 129 97 61 88 49 11

Benin 44 46 0 0 0 2 0

Cameroon 1,584 1,493 2,339 1,857 1,738 926 623

Central African 268 350 187 143 92 156 147

Congo 32 18 4 1 0 1 0

Cote d' Ivoire 3,822 2,885 3,233 4,334 3,857 5,115 2,681

Equatorial Guinea 5 7 7 2 5 2 1

Gabon 27 30 4 2 2 3 1

Indonesia 4,520 6,346 6722 6,378 4,421 5,928 5,014

Madagascar 698 928 925 496 817 680 359

Philippines 440 482 164 122 47 22 133

Togo 281 287 232 181 316 214 194

Uganda 2,318 3,113 2365 2,085 2,013 2,010 2,830

Vietnam 542 868 948 1,146 1,318 2,175 2,642

Zaire 1,052 1,503 2,127 1,479 1,026 995 664

Table 5.-- Continued

1987 1988 1989 1990 1991 1992 1993

Sub-total 59,788 68,255 76,725 71,184 74,192 75,343 67,473

Members exempt
from basic quota

Total Arabicas (2,437) (2,380) (2,967) (2,493) (2,234) (2,306) (1,663)

Bolivia 113 118 158 85 96 51 81

Burundi 490 579 535 655 602 572 426

Cuba 200 202 209 152 188 126 110

Haiti 290 227 191 182 143 171 111

Jamaica 22 13 15 16 20 26 17

MaLawi 52 43 125 87 132 117 71

Panama 150 109 112 130 105 150 94

Paraguay 288 175 322 137 46 55 64

Rwanda 492 616 805 601 573 467 188

Venezuela 173 183 288 132 113 479 444

Zambia 7 4 22 27 24 27 28

Zimbabwe 160 111 188 288 194 65 27

Total Robustas (806) (1,075) (1,479) (636) (1,379) (1,168) (1,351)

Ghana 10 6 11 15 26 42 23

Guinea 89 98 121 54 66 16 27

Liberia 57 80 24 0 0 0 0

Nigeria 13 10 3 2 9 13 9

Sierra Leone 128 79 129 87 74 43 37

SriLanka 53 57 3 33 20 7 48

Thailand 446 743 1,160 433 1,183 1,042 1,196

Trinidad & 1 2 29 12 2 5 9

Sub-total 3,243 3,455 4,446 3,129 3,613 3,474 3,014

Grand Total 63,031 71,710 81,171 74,312 77,805 78,817 70,487

Source: International Coffee Organization (Unpublished Data).

Table 6. Average Quarterly United States Coffee Prices, 1962 -1994

-------------------- (Cents per pound) ---------------

YEAR 1st Qrt 2nd Qrt 3rd Qrt 4th Qrt Average

1962 71 71.2 71.1 69.2 70.8

1963 68.7 69.6 69.7 69.5 69.4

1964 78.3 84.4 84.6 84.8 81.6

1965 83.9 83.4 82.1 82.1 83.3

1966 82.5 83.2 80.6 80.7 82.3

1967 79 77.5 76.1 75.8 77.6

1968 75.9 76.8 76.5 76.4 76.4

1969 75.9 75.6 75 80.3 76.5

1970 86.6 93 95.8 95.5 91.1

1971 94.2 92.8 92.8 92.3 93.4

1972 92.1 91.2 94.4 95.1 92.7

1973 99.4 106.1 108.7 108.9 104.1

1974 114.9 125.1 131.5 129.1 122.9

1975 128.9 126.9 142.5 151.3 133.4

1976 155.5 191.6 211.6 238.1 187.3

1977 299.9 389.2 369 352.1 347.2

1978 340.3 n.a. n.a. n.a. 332.5

1979 n.a. n.a. n.a. n.a. n.a.

1980 320 321.3 301.3 277.7 310.6

1981 253.3 246.1 244 249.8 250.4

1982 260.5 256.6 254.2 253.8 256.8

1983 257.3 250.7 252 251.2 253

1984 258.9 267.1 268.6 268.5 263.9

1985 270.9 270.1 264.8 255.9 266.9

1986 380.5 359.3 324.5 327.5 344.9

1987 310.9 287.4 288 272.9 293.3

1988 276.8 287.4 285.5 292.5 284.4

Table 6.-- Continue

Year 1st Qrt 2nd Qrt 3rd Qrt 4th Qrt Average

1989 303.9 322.5 303.8 293.7 307.3

1990 289.1 305.4 303 294.1 296.6

1991 289.4 280.5 270.6 262.5 280.9

1992 263.1 265.2 249.2 236.4 257.8

1993 246.2 254.8 241.5 248 247.2

1994 251.5 234.1 245 248.1 244.68

Source: Foreign Agricultural Services, USDA.

Table 7. Average Annual U.S. Wholesale Coffee Exfactory, 1987-1992

--- -------------(Per kilogram)-----------

Type 1987 1988 1989 1990 1992

Ground $6.64 $6.41 $4.65 $6.14 $5.94
roast, all

Ground $6.49 $6.34 $6.74 $6.86 $6.53
roast in one
pound can

Soluble $18.91 $19.10 $20.54 $20.17 $18.63

Source: Foreign Agriculture Services, USDA.

and retail prices confirms coffee's use as a loss leader in most stores (Schneider and

Gallagher, 1994).

Marketing Channels and Distribution Systems

Figure 1 shows the major distribution system for coffee. There are coffee

growers in producing countries that are individual producers at the small farmer or


Wet Method Ceupng)

C, aI' I c~mr
En'tare Imn


Figure 1. Marketing Channels

Source: Schneider and Gallagher, 1994

nt Broke r

I Wholesalers

se-vice/re fL~
J n SL~o'uermret/ Coffee

plantation level. In addition, small growers are incorporated into producing cooperatives.

These farmers and organizations are supported by commodity boards that provide several

types of technical assistance. After berries undergo a series of primary value-added

activities they enter into international commerce. The first level of institutional operators

are the in-country traders. These local traders sell the commodity to agents or brokers

who then sell to processors or roasters.

New Orleans is the principal port of entry for coffee entering the U.S. This city

accounted for 28 percent of the total green coffee beans imported in 1992. Ports in New

York account for 23 percent, Laredo, Texas, 12 percent and San Francisco, 11 percent.

No duty is charged on coffee imported into the U.S.; however, under the Tariff Act of

1930 (19 U.S.C. 1319), there is allowance for duty to be charged on imports of coffee

into Puerto Rico. The duty rates imposed in Puerto Rico are $2.50 per pound for green

beans and $3.00 per pound for roasted or ground beans (Schneider and Gallagher, 1994).

Import Channels

In the importing countries, coffee is handled by agents, brokers or coffee

importers who sell the commodity to processors and roasters. The agent/broker's

function is to provide the requisite expertise to buy and sell, ship and finance the coffee

before the other buyers take actual possession. Merchants have established connections

in the areas of production through worldwide communication networks, which they use

to make rapid analyses of the market. The basic function of coffee importers includes

buying directly from the agent/brokers. Extensive buying networks are developed and

maintained by the larger roasting and packing firms. They in turn import directly


through coffee cooperatives. Import agents and brokers in large trading centers

(Hamburg, Germany and London, United Kingdom) also purchase coffee and then re-

export around the globe. Some of this is also practiced in New York, where during the

winter months when the St. Lawrence River is frozen, agents and brokers re-export and

transship coffee to Canada. Also in the U.S., warehoused coffee contracts are traded on

a futures market commodity exchange for hedging and speculation. The Coffee, Sugar

and Cocoa Exchange (CSCE) in New York City offers 37,500 pound contracts for

Arabicas coffee. These contracts call for delivery of washed arabica coffee produced in

several Central and South American, Asian and African countries. CSCE graders in

New York inspect samples sent from port districts around the world before certifying

a coffee lot for delivery. Robusta futures contracts are traded on the London Futures

and Option Exchange (London Fox) and the Paris/Le Harve Exchange. Prices on the

London Fox are quoted in dollars per metric ton and on the French Exchange in francs

per kilogram.

Ninety percent of the coffee imported into the US in 1990 was handled by

agents/brokers with the other 10 percent imported by roasters (International Trade

Center, UNTAD/GATr, 1992). The bulk (60 percent) of this coffee was sold to

roasters. There are about 51 agents and brokers in the U.S. handling the importation of

green coffee, with over 50 percent in California and New York (11 and 18, respectively).

Operations that characterize themselves as coffee importers total approximately 113, with

35 specialty coffee importers. New York dominates the regular coffee importer category

with 48 importers. The Pacific Coast is considered the center of the specialty coffee


industry with 14 importers in California and three in Washington accounting for about

50 percent of the specialty coffee importers (McCabe, 1992).

Marketing Channels

After coffee is imported and processed it is sold to a wholesaler (Figure 1). The

wholesaler then sells to the grocery trade. Wholesale prices are normally set by the large

processors. Once changes in the wholesale price quotes are made by one or more of

these industry leaders, they are often followed by other processors. In analyzing the

relationship between retail and wholesale prices, one should note that although wholesale

prices may be stable over certain periods, various types of promotional tie-in sales may

result in actual prices varying considerably from month to month.

Because of the different strata of operators in the retailing of coffee, retail prices

also reflect various pricing tiers. The final consumers get their coffee through retail

grocery companies (supermarkets), institutional outlets, specialty shops, and retailers.

There are numerous brands on an average supermarket shelf and the range of retail prices

can be large. Choices in brands include regional and private-label brands. The

institutional market includes restaurants, hotels, and fast-food outlets, public and private

facilities. Regional and local coffee processors, and wholesalers are the main suppliers

to these outlets. Coffee shops that specialize in gourmet coffee purchase green beans or

special order blends directly from importers and roasters. They generally offer a high-

priced product.

Market Structure. Concentration and Commercial Entities

The 1987 Census of Manufacturers (U.S. Department of Commerce) revealed

that there were 111 domestic coffee processing companies primarily engaged in roasting

coffee and manufacturing coffee concentrate and extract in powdered, liquid or frozen

forms. Between the census period of 1982 and 1987, the number of companies declined

by nine percent. Employment also declined by nine percent over this same period from

11,800 to 10,700 ( Schneider and Gallagher, 1994).

The coffee processing industry consists of regional roasters and packers, including

grocery companies and large national food manufacturers. Many processors are

independent roasters who prepare their own brand and products for the supermarket chain

stores. The others are national or regional food manufacturers who prepare their

company's brand of coffee. In addition, coffee is processed by small food processors

and specialty shops.

In terms of concentration in the industry, there are four large firms that account

for 66 percent of the value of shipments of roasted coffee. The largest 20 firms account

for 90 percent. Concentration in the coffee processing industry is said to increase during

periods of high green coffee prices. This is because smaller firms are believed to have

more difficulty competing on a product-by-product basis with larger roasters (Andrews,

1992). The growth in micro-roasters (250-500 bags per year) has been tremendous.

In 1969, these firms were 20 in number clustered on the west coast between San

Francisco and Seattle. In 1989, there were approximately 385 firms classified as micro-

roasters. This increase is a result of the increase in demand for specialty coffees.


The structure for the commercial roaster market has not changed much.

Maxwell, (1994) reported that this subsector is dominated by a few companies. These

include Philip Morris' General Foods with its Maxwell House Brands (33.4 percent) and

Proctor and Gamble with its Folger's Brand (31.5 percent), collectively accounting for

two-thirds of the ground roasted market in 1992. General Foods and Nestles accounts

for 67 percent of the market share for instant and soluble coffee. In general, there are

about 171 regular and 126 specialty coffee roasting establishments located throughout the

U.S. Their wide distribution suggests that these roasters are located near the market they


Milestones in the Development of the World Coffee Market and the ICO/ICA

In October, 1963, the first International Coffee Agreement (ICA) was developed

and resulted in the formation of the International Coffee Organization. The reason for

its development was to halt the rapid decline in coffee prices and reduce over production.

The main market regulatory instruments were the export quota system (EQS) and

voluntary export restraints (VER). These instruments were discontinued in 1974 because

producing and consuming countries were unable to agree on prices and quota levels.

World coffee prices were at historically low levels in the early 1970's. However,

a serious frost in Brazil in 1975 caused prices to reach a new high of $3.79/kg in 1977.

Prices thereafter declined sharply due to persistent over production and increases in

stocks. This prompted both the producing and consuming countries to enter into

negotiations for a new agreement with EQS as its main trade policy.

From October, 1980, through February, 1986, the EQS was successful in keeping

a floor under world market prices. The quota system was suspended in 1986 because of

sharp increases in prices due to a severe drought in Brazil in 1985 and low forecasted

export quantities in the 1986/87 season. After 1986, prices declined steadily. This led

to prolonged discussions in October, 1987, among the ICA members for possible re-

instatement of the quota system. Lack of agreement among member nations led to no

change occurring.

When in operation, the export quota system was used to determine the global

export quota and each exporting member's quota at the beginning of each international

coffee year (October to September). Exporting quota-members covered 95 percent of the

world's supply. Importing quota-members accounted for between 85 and 90 percent the

of world's demand. Depending on the relationship between world prices and the price

range set under the agreement, the quotas were adjusted frequently. There was no

constraint on members' exports when the EQS was not operative.

Furthermore, there were no restrictions on exports from exporting members to

non-ICA importing members (i.e., the non-quota members). New Zealand, the Newly

Independent States (NIS) and the formerly centrally planned economies of Eastern Europe

and all developing countries except Portugal and Greece are considered as non-quota

market countries. In these markets there are no consistent and reliable data on prices.

It has been reported that when burdened with large stocks of coffee, exporting ICA

countries sell to non-quota countries at discount up to 50 percent of the quota market

prices. Another factor is that of barter/counter trade transactions as practiced by these

non-quota members. The ICO has attempted to penalize these discounted sales but has

had limited success.

The presence of large stocks without an international agreement has led to coffee

prices declining 50 percent since July, 1989. In response, producing countries have

sought to cut export taxes, reduce prices to farmers and cut back on inputs, particularly

pesticides and fertilizer. A World Bank study (Akiyama and Varangis, 1989) predicted

that without the ICA, prices would climb again in 1995, and world coffee export

revenues would exceed those expected with the ICA. The study reported that a lack of

quotas favors some countries over others and would lower prices by the turn of the

century. The study questioned whether these lower export prices would lead to lower

coffee prices at retail levels.

ICO/ICA arrived at an administrative agreement in March, 1993. However, in

the fall of 1993, the U.S. withdrew its membership from the ICO/ICA. In response, 28

producing countries formed the Association of Coffee Producing Countries (ACPC). The

ACPC has instituted an export retention scheme which has shown signs of disintegration

due to numerous exemptions that are granted to various members.

In the summer of 1995, the Central American countries Brazil and Colombia

agreed to reduce the supply of coffee in order to boost world prices (Caribbean Update,

1995). The reduction is estimated to be between 3.5 and 4 million bags annually.

Industry representatives from these countries blamed "market speculators" in New York

for the recent plunge in coffee prices, while noting that stocks of coffee in consuming

countries have disappeared along with the decline in production.



The theoretical arguments for the institutional framework of international

commodity agreements are examined in this chapter. In general, the effects of the export

quota on market demand, structure and likely structural changes are presented. Several

models are presented with their results and limitations. Descriptive and empirical studies

are also presented with a view of providing justification and rationale of the approach

followed in this study. In addition, a brief summary of previous studies of coffee

demand in the U.S. is highlighted. In general, the review of the literature is vital to the

conceptual and empirical framework adopted for the estimation procedures used in this


Institutional Framework and International Commodity Agreements

An analysis of the price, quantity and expenditure effects requires that the

institutional and theoretical framework of trade agreements with and without economic

provisions be examined. Many primary non-competitive imports from developing

countries to developed countries are traded under international commodity agreements.

These include cocoa, coffee, bananas, sugar, and most spices. These agreements are

basically commercial trade policies that affect the exchange of these commodities. In

international trade there are several commercial policies that regulate the flow of goods

and services. These include tariffs, quotas, voluntary export restraints (VER), voluntary



import restraints (VIR), currency devaluation and other non-traffic measures including

licensing, pythosanitary measures, and commodity agreements. These policies may

enhance or impede the flow of goods. They impact import and export prices, the volume

of product traded and the allocation of import expenditures among competing regions or

countries. In general, these affect producers and consumers surplus both in the exporting

and importing countries. In the case of the present study, coffee has been traded under

commodity agreements since 1963.

International Commodity Agreements are placed in four classifications (Tweeten,

1992). These are buffer stocks, bilateral agreements, multilateral contracts with or

without supply controls, and export cartels. The institutional settings and trade policy

for coffee would classify as multilateral contracts.

The ICA was administered by the ICO. The focus was to control country export

quotas (but not production controls) in an effort to restrain prices within a price band

agreed upon by buyers and sellers that were signatory to the agreements. The

mechanism for implementation includes the issuing of stamps to producing countries for

their quotas and consuming countries were required to admit only ICO certified coffee

with attached stamps.

The International Coffee Organization has administered five International Coffee

Agreements since 1963 in an effort to stabilize the price of coffee (Clarke and Macrae,

1988 and Dull, 1994). These agreements were administrative with and without export


Several attempts have been made since 1902 to develop a coffee agreement (Vogt,

1990). However, it was not until after the 1929 Great Depression that prices declined

sharply. The major Latin American producing countries attempted to bond together to

protect themselves against these low prices. The Pan-American Coffee Bureau was

formed in 1937 with its headquarters in New York. Its mission was to promote the

consumption of coffee in the U.S., the largest single market for coffee. In addition, it

would try to regulate production and exports in order to stabilize prices. Prices for

coffee continued to be depressed and were pushed lower by the outbreak of World War

II, coupled with the loss of most markets. This critical situation led to the formation of

the Inter-American Coffee Agreement of 1940. This agreement provided for the issuing

of export quotas to producing countries that export to the U.S. This became the

precursor to what was to come in the post-war period.

After World War II, the development of the United Nations Organization led to

the implementation of wide spread services to solve the economic and financial

difficulties that primary commodities face in international trade. Coffee did not come

immediately under the international umbrella. During this period prices of coffee rose

to record levels due to a number of factors including the re-opening of markets and

commodity stock piling in producing countries. Production boomed in the 1950's,

resulting in the decline of coffee prices. In 1956, the Organization of American States

(OAS) sought to draft an International Coffee Agreement, but was not supported by the

U.S. Prices continued to decline and an agreement among Mexico, Brazil, Colombia,

Costa Rica, Guatemala, and Nicaragua, was implemented in 1957. This agreement was


effective in stopping the further decline in price in 1957. But reports of heavy crops in

1958 resulted in further price decline. Several studies over the period of 1958-1960

suggested the implementation of a wide range of proposals. The proposals included

import quotas, buffer stocks and production limits. By 1962, regional groupings were

formed with assistance of the colonial powers of France, Belgium, Portugal, and

England. They included the Inter-African Coffee Organization (IACO) and the Coffee

Federation of the Americas (FEDECAME). Full scale negotiation included the United

Nations, and led to the ICO based in London. ICO consisted of 71 coffee exporting and

coffee importing countries and other interested organizations. By the November, 1962,

deadline, the agreement had been signed by 54 governments, 32 coffee exporting

countries accounting for 95 percent of total world exports, and 22 coffee importing

countries accounting for 90 percent of world coffee imports.

There have been five agreements since 1963. These agreements were

administrative with and without export quotas (Clarke and Macrae, 1988 and Dull,

1994). The 1963 agreement went into force provisionally in July, 1963. The objectives

of the agreement were to:

achieve a reasonable balance between supply and demand on a basis that

ensures equitable prices and brings about long-term equilibrium between

production and consumption;

alleviate serious hardships caused by surplus supplies and excess

fluctuation in the price of coffee;

assist in increasing the purchase power of coffee-exporting countries by

keeping prices at equitable levels and by increasing consumption and;

set up a system of export quotas and controls.

Stock policies were developed with the implementation of proper statistical records. The

system lasted for ten years until it collapsed in 1973/74.

The second agreement was negotiated in 1968. While the foundation work was

completed in 1967, the oversupply of coffee caused downward pressure on prices.

Export quotas remained in effect with additional emphasis placed on the need to curb

production and to implement diversification measures in producing countries such that

supply and demand were brought into equilibrium at acceptable prices. The main

objective of this agreement was to maintain and protect the quota and controls systems,

reign in long-term oversupply problems and expand the ICO statistical database. The

Secretariat was engaged in several rifts. Two major ones were battles over market shares

of individual exporting countries and sub-plots that threatened to break up the agreement.

These plots were based on the rights of exporting members to give some form of

protective treatment to the processing of coffee into soluble (instant) form and its sales


A third agreement was negotiated in 1976. There was a break in the quota

system. Conditions in Brazil precipitated the suspension of the quota system. While the

guiding principles of the needs to stabilize prices and guarantee producing countries

adequate foreign exchange earnings were valid, the quota system agreement was not

implemented. The agreement allowed for the application of the quota when prices were


low and were to be suspended when prices were high. In addition, the agreement

segmented the markets into two types, markets where exporting countries operate under

quota and non-member markets to which all exports were free of quotas. This was the

beginning of instability in the ICO/ICA.

In 1983, the fourth agreement was developed with a view to attempt to re-

introduce the quota system. Again members who had expanded production sought to

press their claim for larger quotas. But those who had used up stocks to supply the

market argued against using stocks as a criterion for determining how quotas were

shared. A solution was found under a "status quo" market share arrangement. By 1986,

ICO/ICA entered another series of hard negotiations when weather conditions caused

supplies in Brazil to be reduced, resulting in higher prices.

In summary, the 1983 ICA failed because of discount sales to non-quota members

who in turn re-exported some processed coffee to quota members, and disagreements

over distribution of export quotas and over the issue of selectivity. Selectivity applies

to the availability of the types and qualities of coffee required by consuming countries.

The fifth and final agreement took over six years to be concluded and came into

effect October, 1994 (Dull, 1994). This agreement will run for a five-year period. It

does not have economic provisions and the U.S. is not a member. It is an

Administrative Agreement. The main features are as follows:

(i) ensure enhanced international cooperation in connection with world coffee



(ii) provide a forum for inter-governmental consultations and negotiations

when appropriate on coffee matters, to achieve a reasonable balance

between world supply and demand on a basis which will assure adequate

supplies of coffee at fair prices to consumers and remunerative prices to

producers, and long-term equilibrium between production and


(iii) facilitate the expansion of international trade in coffee through the

collection, analysis and dissemination of statistics and the publication of

indicator and other market prices and thereby to enhance transparencies

in the world coffee economy;

(iv) act as a center for the collection, exchange, and publication of economic

and technical information on coffee;

(v) promote studies and surveys in the field of coffee and;

(vi) encourage and increase the consumption of coffee.

Theoretical Framework

Market Structure and Export Quotas

The structure of the coffee market both in producing and consuming countries

depicts that of an oligopolistic market structure (Geer,1971). Conforming to the

theoretical assumptions of an oligopolistic structure with implicit conclusions, the

industries have been able to form a cartel consisting of both producing and consuming

countries (Stigler, 1964). The cartel employs the market pain approach of bargaining


power and supply control to achieve price objectives (Tweeten, 1992). However, cartels

are inherently unstable due to the following reasons.

(1) Any attempt at price fixing without production controls fails because it

develops prices initially set so high that they stimulate excess production

and result later in a decline in price, or prices are set so low they

stimulate shortages which result later in higher prices.

(2) They lack funds to hold domestic buffer stocks to support the price band.

(3) There is a lack of trust between the signatories, a lack of discipline among

sellers in the domestic countries and a lack of generosity of the buyers in

the developed countries.

(4) Unanticipated market and technological developments work against the


The result is that commodity agreements have had limited success. In general,

the formation of trade cartels have not been first best solutions (Eaton and Grossman,

1986). In addition to the above quantitative restrictions, quotas aid collusion between

firms in the producing and consuming countries (Helpman and Krugman, 1989).

Problems arise in allocating quotas and maintenance of trade shares. These problems

further impair the framework for free trade in that with the trading of a homogenous

commodity from supplying countries to one country, a quota or voluntary export restraint

(VER) allocating mechanism could intentionally and unintentionally result in price

discrimination (Kreinin and Dinopoulos, 1992).


In summary, manifested under the 1963 Agreement with the fixing of a global

quota, exports to all countries except the 30 which were designated new markets (without

quotas) had to be monitored in order to ensure a balanced market. Rigid control of the

market allowed this agreement to hold for ten years. However, a series of natural events

(for example, frost in Brazil in 1975) coupled with re-exported quantities for the "non-

quota new markets" and political upheaval in African producing countries caused the

quota system to collapse. Attempts were made to re-instate the quota system in the 1976

Agreement but the system could not be enforced. This situation still prevails, and the

1993 Agreement is administrative in nature without economic provisions.

Market Demand and Structural Changes

In the present study of structural changes in import market demand under various

trade policies, structural changes are influenced by forces and conditions either inside or

outside the market. These influence the behavior of the firms, thereby affecting the

performance of the market. Demand and supply are important market structure attributes

affecting these elements (Cochrane, 1957). Several authors have shown that structural

factors affect demand for agricultural commodities (Bain, 1942, Cochrane, 1957;

Lanzillotti, 1960; and Clodius and Mueller, 1961).

Seven structural factors within the market have been identified as affecting

agricultural markets. These are the number of operators in the market, product

homogeneity and differentiation, product durability, bulkiness of the product in

relationship to its value, ratio of overhead cost to variable costs, adequacy of grade

descriptions for buying and selling, and continuity and length of the production process.


Cochrane (1957) showed that as the number of operators within a market decrease the

probability of market collusion increases. In general, firms within an industry recognize

the advantages of collusion (either formally or tacitly). While collusion has not been

empirically shown in the coffee industry, examining the industry on the importing side

gives evidence that the number of operators has been decreasing over the years.

Schneider and Gallagher (1994) reported that the number of coffee roasting and

manufacturing firms declined by nine percent between 1982 and 1992 in the U.S. coffee


Bain (1942) and Lanzillotti (1960) illustrated that other factors, including the

number and degree of concentration of buyers and the size distribution of firms, also

influence the likelihood of market collusion. Again, while empirical evidence does not

bear this out in the U.S. coffee industry, there is evidence of consolidation of firms and

the volume of trade shares that they handle. Maxwell (1994) showed that the U.S.

ground roast market continues to be dominated by a few companies, with Philip Morris'

General Foods with its Maxwell Brand (33.4 percent), and Proctor and Gamble's

Folger's Brand (31.5 percent) accounting for nearly two-thirds of the ground roast market

in 1992. In 1980 the top four processing firms controlled 65 percent of the coffee

market in the U.S. (Bread for the World, 1990). This is likely to increase with the

acquisition of smaller firms by the larger companies, e.g., the 1989 acquisitions by

Proctor and Gamble of Maryland Club and Butternut for the Winter Park Investment

Group (Maxwell, 1994).


Structural factors cannot be considered independently according to Clodius and

Muller (1961). They argued that when firms are small in number, they differentiate their

product and use non-price competition to stimulate sales instead of collusion. Evidence

of this can be seen in coffee advertising expenditures. Industry sources (Butcher and

Singer, Inc. and Leading National Advertiser, Inc.3) reported that promotional

expenditures were $129 million for the first six months of 1991, five percent over what

was spent for the entire year in 1990. The big two, Phillip Morris and Proctor and

Gamble, accounted for $88.3 million of the total, or 68.4 percent. On the producers

side, the National Federation of Coffee Growers of Columbia increased its advertising

expenditures by over 13.5 percent to $8.3 million in the first six months of 1991 as

compared to the same months in 1990.

Product differentiation in specialty coffee may have also changed the type of

competition from price to non-price competition. This would be consistent with the

theoretical argument advanced by the above mentioned researchers. Product

differentiation is known traditionally to maintain excess profits within an industry by its

ability to act as an effective barrier to entry. Non-price competition used by firms to

maintain consumer loyalty is evidenced in the industry under study.

Several structural forces external to the coffee industry are also known to have

significant effects on the market. Whether these factors pull or push or interact to propel

the market is uncertain. The external forces consist of changing consumers' taste and

3 Butcher Singer Inc., and Leading National Advertiser Inc., are two firms that track
advertising expenditures in the coffee market (Maxwell, 1994).

preference as reflected in consumption, structural changes in the labor force, increases

in population, rising income and standards of living, the frequency of away-from-home

consumption, and the awareness of consumers to nutrition and dietary concerns due to

increased information and product research.

The National Coffee Association (1993) reported that although per capita daily

consumption increased from 1.75 cups in 1991 to 1.87 cups in 1993, the overall daily

consumption was well below the peak in the 1960's of three cups per person per day.

Changing consumer preferences have accounted for the shift in consumption. Reasons

attributed to the decline in coffee consumption in the U.S. range from increased

consumption of cold drinks (e.g., soft drinks and other beverages) to lifestyle changes

away from structured meals in the home (Schneider and Gallagher, 1994). These factors

are enhanced by the increased extraction rate in brewing and higher roasting yields from

improved roasting systems.

In terms of the relationship between consumption of coffee and changes in retail

prices, consumption was shown not to vary with most changes in retail prices. Empirical

research at the retail level showed that coffee demand is inelastic (International Monetary

Fund, 1988). However, there are two issues that must be noted, the time horizon of the

study and whether there is full price transmission between the wholesale and retail level,

particularly backward transmission. Research using data on the industry in the late

1950's into the 1960's showed that coffee demand at the wholesale level was price

responsive (Bacha, 1968; Geer, 1971; Adams and Behrman, 1976). A possible


theoretical explanation for this could be the dynamics of maturity of the market over time

and the changes in consumers' taste and preferences.

Previous Studies

Several production and marketing studies have been conducted on the world's

coffee industry. These are country or region specific in some cases but generally

worldwide in scope. These studies were conducted looking at the industry from both the

demand and supply sides. These studies can be divided into two groups, descriptive

analyses and empirical analyses. The earlier studies were descriptive in nature while in

later periods, particularly the 70's and later, most of the studies were empirical.

Descriptive Analyses

On the supply side, Smith (1974) showed the importance of coffee production in

the diversification process of the Colombian economy. After considering historical,

political, and socio-anthropological factors, he agreed with the findings of Prebish (1963)

and Krause (1965), that in the production and trade of primary products by developing

nations, the process presents certain problems. These problems are alleged to be

centered on freely competitive trade of these products which are claimed to lead to the

development of market conditions unfavorable to the producing countries. The premise

for this argument involves the unique supply and demand conditions that characterize the

production and marketing of raw material commodities. These conditions can be listed

as follows:

a relatively large number of producers of the primary commodity which

makes adjustment to market conditions difficult;


no single country is likely to control more than a small fraction of the

trade in any one commodity;

weak organizational strengths and institutional capability of the growers'

groups to exert control over the output of individual producers.

the combination of relatively inelastic short-run supply conditions coupled

with inelastic demand of the product in short-run resulting in over

production and price instability;

income elasticity of demand for raw material in developing countries less

than unity (Prebish, 1963) and;

technological advances in production of primary products leads to severe

long-run over production, surplus accumulation and declining prices and


In summary he argues that while commodity agreements like the ICA solve the

price stability problem somewhat, it creates a problem with surplus production capacity

in the major supplying countries. Therefore, commodity agreements in the presence of

excess productive capacity leads to misallocation of resources and declines in economic

efficiency. The study concludes that producing countries diversify into other enterprises

away from coffee and suggest dismantling of the ICA program.

Shelton (1976) focused on the possible effects of the instability in the prices of

a primary commodity on economic growth in Brazil. The methodology used by Shelton

was a combination of general analysis (exploration of general development theories) and

case study approach using coffee as the product. He explained two standard schools of


thought, international trade as a potent engine of economic growth, and the fraction of

gross domestic product devoted to international trade tending to decline once a country

reaches a fairly industrialized or developed state. He showed that there is no

contradiction between the two theories and concluded that one precedes the other. The

validation of the results depends on the type of primary products, the degree of

underdevelopment, and the domestic, international, political, and economic conditions.

Several studies examined the trend in consumption of coffee in the U.S., the

largest single market in the world for coffee. Each winter for the past 33 years the ICO

commissioned the winter coffee drinking study. The annual study in the winter of 1986

revealed the heightened dynamism in the U.S. coffee market. It showed that while there

is a long-term fall in the consumption of coffee on a per capita basis, the decrease is not

as significant in the total amount of coffee consumed due to increases in the overall

numbers of coffee drinkers. The study also describes the promotional efforts to change

the image of coffee in order for it to appeal to the young and modern life styles. In

addition, it suggested conducting promotional activities that will increase the consumption

of coffee away from home, e.g., workplace, institutions and vending machines. The

study showed the increase in consumption of decaffeinated coffee and changes in methods

of preparing coffee.

In 1988, the comprehensive text (Clarke and Macrae, 1988) dealing with the

commercial and technological aspects of coffee was published. Several authors

contributed to the body of work covering trading patterns, institutional settings for

commercial trade and the international standardization of coffee. In general the work


showed the political and economic importance of coffee to several developing economies

and the world at large.

The work of Vogt (1990) provided the background for the U.S. Congress and

several business interests in the coffee industry to spur the U.S. government to announce

its withdrawal of membership from the ICO/International Coffee Agreement in 1993.

Vogt suggested that the breakdown of the agreement meant less expensive, but higher

quality coffee being available to importers on world markets. He also noted that this

situation had allowed for diminished export earnings to growers. He identified the

effects of the export quotas as follows. With quotas limiting supplies, member importing

nations were more likely to bid against each other for available coffee, driving up prices.

The reverse is also true. He concluded that because the current agreement (1989) has

no export quota enforcement, no implementing legislation was necessary. However, in

the event that there is negotiation of a new agreement with export quota limiting supplies,

the U.S. participation would be subject to Senate approval and legislative measures by

both Houses of Congress. He suggested that members of Congress had several options.

These were as follows.

(i) Play the laissez-faire role of doing nothing. Any cartel mechanism that

supports managed trade would have several opponents who support free

trade and believe the marketplace provides the greatest benefit to all when

it finds its own equilibrium without government intervention.

(ii) Pass a supporting or opposing resolution to the President's effort to

establish a new ICA.


(iii) Appoint a panel that would conduct hearings to make clear the advantages

and disadvantages of the new ICA. However, this would require

additional personnel and new funds which might be a constraint given the

climate in Congress to curb deficits.

(iv) Consider special funds for producing countries' diversification programs.

(v) Consider preferential tariff treatment. Congress could reward specific

countries by levying a tariff/tax on all coffee imports from those countries

the U.S. does not support. It should be noted that currently there is no

U.S. tariff on imported green coffee.

(vi) Consider a coffee-for-debt swap. Many of the exporting countries have

heavy external debts and are combating drugs. Swapping coffee for debt

might be one way the U.S. can assist these countries.

(vii) Require stockpiling of coffee. Congress could direct the Administration

to purchase and stockpile coffee from countries the U.S. wants to assist.

This will cause the price of coffee to increase in the short-run and because

coffee is a non-competitive import, this option would not come up against

resistance from U.S. producers. The downside of this decision would be

the biological characteristics of the commodity. After six months coffee

tends to lose its flavor. In addition, purchases would increase the U.S.

budget which has other negative implications.

(viii) Consider storage facility construction assistance. U.S. funds could be

used to build storage facilities in producing countries. This would allow


for timely release of supplies into the world market. However, apart from

the loss in quality if stored too long, there is the additional problem of

where the net effect of increasing world prices would work against U.S.

importers and roasters.

Maxwell, (1994) analyzed the coffee market by looking at consumption,

recognizing that per capita consumption in 1993 was stagnant and had fallen slightly

despite favorable media attentions and the entrance of "Baby Boomers" into their middle

years when coffee drinking tends to be highest. He indicated certain significant changes

in the industry. While the consumption of decaffeinated and soluble coffee has declined,

regular coffee has become the favorite type of coffee again for most consumers.

Specialty coffee shops, gourmet food stores, "shopper club" stores, and discount stores

have gained popularity as the major shopping sources of regular coffee. Particularly in

the Northwestern U.S., espresso bars, carts, and coffee bars have become common

place. These stores are gaining market shares and show increasing profitability by

serving a variety of fancy coffees. The study also outlines institutional changes that are

likely to impact the market over the long-run. Because of the series of failures to

negotiate a new ICA, a group of 28 producing countries formed the Association of

Coffee Producing Countries (ACPC) in 1993. The action caused the market price of

coffee to show some increase. This was achieved by the group implementing a retention

scheme. Another cartel can be seen in this formation. In the fall of 1993, the U.S.

informed the ICO that it would not agree to a one-year extension of the ICA and would

no longer be a member of the ICO. This action did result in a serious economic decline


in export revenues for producing countries due to lower prices, but caused an increase

in higher quality coffee at lower prices to U.S. consumers.

Schneider and Gallagher (1994) presented a comprehensive overview of the coffee

industry as it relates to the coffee market in the U.S. On the supply side they

summarized some of the political and economic factors that affect the top supplying

regions including Brazil, Columbia, Central America, Africa, Asia, and Oceania. After

1990 the government in Brazil had relaxed its control on the industry, allowing farmers

and cooperatives to negotiate free market sales directly with exporters and their agents.

However, the government still retained control over export prices by closely monitoring

the system of Central Bank registrations. Due to its influence Brazil plays a prominent

role in the supply, price, and allocation of export quotas under ICA. Current and future

trends in the world market for coffee will be influenced by Brazil's output because it

supplies over 30 percent of the world's total supply.

In Africa, supplies are hampered by political upheavals. Because of the high

quality arabic produced by some countries (e.g., Colombia, Kenya and Tanzania), these

countries are in a better marketing position than others. However, switching of

consumer preferences to South American Arabicas could cause declining market shares

for several African countries. Production in general continues to decline due to low

international prices and in some countries (e.g., Cameroon), implicit taxation by

government policy.

In Asia and the Oceania regions, Indonesia emerges as the world's third leading

supplier of coffee behind Brazil and Colombia. In general, it is the largest producer of


Robusta. Production in the region has increased due to an increased demand for coffee

in Japan. Most plantings have included a higher percentage of Arabica. The roles of

the government consist of collection of taxes, administration of quality controls, and the

monitoring of foreign exchange receipts.

In addressing the world market situation and the future of the ICA, Schneider and

Gallagher reported that failures of the ICA caused hardship on certain producing

countries. Additionally, as returns fall the quantity of inputs used decline, which

adversely affects supplies and quality. This condition is said to be detrimental to both

the producing and consuming countries.

Dull (1994) focused on the failure of the ICA and the role of the newly formed

Association of Coffee Producing Countries (ACPC). In implementing its retention

scheme in October, 1993, it required, depending on market prices and production levels,

that the 29 members withhold up to 29 percent of their exportable production. ACPC

based its retention scheme on target price ranges of the 20-day moving average of the

ICO composite price for Other Milds and Robustas. Members whose production and

annual exports are less than 400,000 bags (bag = 60 kgs.) were exempted from the

retention scheme. ACPC monitors the market and once price exceeds the target level,

it allows the other countries to release retained coffee. In the first six-month report

released by the ACPC, over ten members were exempted from the scheme because of

the smallness in the size of their crops. Indonesia, Ethiopia and Ecuador indicated their

willingness to initiate the retention plan given their 1994 harvest cycles. Burundi and


Rwanda requested exemption because of the civil wars, while Kenya and Tanzania asked

for special treatment because their marketing was based on an auction system.

Empirical Analyses

Many empirical studies have been conducted to explain the coffee market. Most

have been modeled assuming a perfectly competitive market structure. Over the last

decade attempts have been made to model the coffee market under imperfect market

structure. The following presents a summary of some of the empirical research that has

been applied to the coffee industry after the initiation of the ICO/ICA in 1963.

Bacha (1968) constructed a policy-oriented econometric model for the

international coffee market. He evaluated feasible policy alternatives using the Brazilian

Coffee Market as the case study. Using data from 1954 up to pre-ICA (1963) and then

five years after the Agreement, he estimated the following model. It consisted of twenty-

three stochastic equations and twelve identities or market equilibrium conditions to

evaluate supply and demand sectors. The model allowed for the incorporation of three

main types of coffee (Brazil, Milds, and Robusta). It was further extended to execute

policy simulation experiments using the reduced form of the model. Certain conditional

conclusions were drawn.

(a) Brazil faces a price-elastic demand function for its coffee, counter to

results from other studies and the country stands to lose revenue by

decreasing its price.


(b) The income elasticity of demand for Brazilian coffee is smaller than that

for others types of coffee. Independent of Brazil's pricing policy, this

will result in a continuous reduction in Brazil's market share.

(c) The exportation of Brazil's market-quota without any reduction of the

current price of its coffee is contingent upon reducing the quota of its

competitors at the bargaining table of the ICO or a change in the pricing

rules of the ICA.

Vanderslice (1971) investigated the success of the ICO/ICA in dealing with the

problem of coffee production by using an oligopoly framework where the coffee industry

is dominated by governmental cartels. It confirmed the two conditions that are necessary

for an oligopoly to form under government cartelization. These were: (a) relatively

inelastic long-run demand with respect to price movements and; (b) the existence of

market conditions that prevent the entry of more than a few producers. The model was

able to show that in the transfer of resources, as occurred with the ICA, the Agreement

caused various inefficiencies resulting in losses in both producers' and consumers'

surpluses due to the artificially high price of the product. In addition, the misallocation

among producers due to the allocation by quota instead of marginal cost was recognized.

The conclusion was that although the ICA had several problems, price increases were

stimulated by oligopolistic forces that existed even before the establishment of the

ICO/ICA. The study recommended the following in order to avoid the pitfalls of such

commodity agreements:

(i) imposition of a tax on the commodity;


(ii) the establishment of a diversification fund and;

(iii) consuming countries should try to understand and know how badly the

market is functioning, and what oligopoly tactics they are confronted with

before entering into any commodity agreement.

Pollock (1971) developed a price dependent world model in an attempt to evaluate

the world price and production effects which resulted from the operation of the ICA from

1963 to 1968. The model was estimated with and without the ICA with statistical

verification. The model covered a sixteen-year period (1953-1968) using 20 consuming

countries that accounted for 90 percent of world's coffee imports. A summary of the

results is as follows.

(1) Exported quantities exceeded quotas during the 1964-68 period under the


(2) The ICA did not achieve a set of prices which was significantly different

from the prices achieved prior to the ICA, i.e., the old one-year

agreements prior to 1963.

(3) The pattern of foreign exchange earnings for exporting countries was not

significantly different in either period.

The reason offered for the above results was that the ICA quota system was not

strictly followed. The study was unable to empirically test production effects, because

the ICA had not explicitly instituted any production controls on producing member



Adams and Behrman (1976) explored the performance of relatively simple

standardized models in an attempt to capture the major features of international

commodity markets. The general specifications used in their treatment of the coffee

industry was to model : (a) supply relations for the developed market economies, the

developing economies, and the centrally planned economies; (b) demand for the same

three groups; (c) world inventory and; (d) world price determination. The study covered

the period 1956-73. The results on the supply side showed that there was no evidence

of a short-run price response, while there was evidence of long-run price response with

a gestation period of six or seven years.

The long run elasticities of demand were inelastic for developed and developing

economies, but were highly elastic for the centrally planned economies. The latter was

attributed to the non-price allocation mechanism and conservation of foreign exchange

in the centrally planned economies. The estimated relationship for deflated coffee prices

showed that the demand for coffee stocks was quite responsive with respect to deflated

prices, i.e., it was highly elastic. Fluctuations in supply and demand were absorbed in

inventory charges relative to price movements. Using dummy variables to estimate the

effects of ICA on price, it was shown that while the effect was significant, it was

diminishing around the mid-1960's and into the 1970's.

Ford (1977) developed world models to evaluate the supply, trade and demand

situation. An asymmetric adjustment model was used for Brazilian supply where coffee

trees were treated as a capital stock variable dependent on real coffee prices received by

the farmers, influence of the Brazilian diversification programs during the 1960's, the


impact of the severe frost in Southern Brazil, the gestation period between the time new

trees were planted and when the first mature harvest was obtained and coffee yield per

unit area of mature trees.

Standard economic theory was used to model demand and trade. It was further

extended to include changes in inventories. In the determination of spot market prices

three factors tend to influence these prices. These factors were Brazil's supply

dominance, the ICA and differences in prices for different grades of coffee.

Simultaneous estimations were used with a series of validation tests. The tests

showed that the production sector was the best part of the model. Stocks of trees

respond in an asymmetric way to changes in prices and are responsive to the

diversification program and the severe frost. The validation tests showed that the poorest

component of the system was the demand and trade sectors. The demand and trade

relationships were most developed for the U.S., less developed for Europe and almost

non-existent for the rest of the world. The response showed that demand for regular

coffee was inelastic with respect to it own price, real income, and substitutes (soluble

coffee and tea). While the validation tests showed the price sector performed better than

the trade and demand sectors, the results showed that spot market prices were responsive

to changes in levels of exportable production and inventories as well as the ICA.

Layman (1979) presented one of the most extensive models for analyzing the

impacts of export controls since the implementation of the ICA. This study developed

both partial and general equilibrium models to evaluate the theoretical effects of imposing

quantitative and non-quantitative restraints on exports. The model application was geared

toward the Brazilian coffee industry.

His analysis demonstrated that with the general equilibrium model, symmetry

prevails regardless of whether the controls were placed on the exported or imported

good. It also supported the known position that when a country is a significant supplier

of a strategic commodity, export controls are preferred.

Three different types of export controls where tested in the case of the partial

equilibrium model. These were an export tax, a tax on exportable production and an

export quota. The evaluation tested the impact of these controls on international price,

trade flows, domestic price, production, and consumption. The results showed that under

small country assumptions an export tax and a quota had similar effects, while the effects

of a tax on exportable production were slightly different. Under the large country

assumption, it was shown that although the magnitudes were different, market structure

does not affect the direction of the effects of imposing different types of controls. Both

export taxes and quotas have different effects from a tax on the production of exports.

The results for the application using data from 1961-75 for the world and Brazil

coffee industries confirmed that low price elasticities of demand and supply for coffee

existed as reported in previous studies. Combining the model within a monopolist

structure and imposing an export tax showed that over the period 1968-75, Brazil coffee

prices and quantities traded were influenced to a limited degree by the tax policies



Hyuha (1982) developed single-equation models to evaluate the world's demand

for coffee with particular reference to supply sources from producing countries in East

Africa. The emphasis was on the U.S. coffee market. The results showed that coffee

had an inelastic own-price elasticity of demand. The claim that small countries could

continue to supply as much as possible without affecting their revenue flow was not

supported. Evidence of this was substantiated where the own-price elasticity was less than

unity. In addition, he found that coffee was not only a normal good, it tended to be

income elastic. He supported supply management and cooperation under the ICA.

Akiyama and Duncan (1982) developed an econometric model of the world coffee

economy to evaluate the medium to long-term market outlook for coffee. The model was

constructed with nineteen equations for supply and demand.

In summary, their results showed a dismal outlook for growth in consumption and

price of coffee, with stagnating demand in the U.S. Future demand in Western Europe

was highly dependent on income growth, and pressure from substitutes (e.g., soft drinks

and fruit juices). Prices and the rate of growth of production were forecasted to decline

during the middle of the 1990's. The simulation models showed price and income

decline because of increases in production at a faster rate than growth in demand from

population and income increases.

Short run supply was highly inelastic. Long-run price elasticities were inelastic

except for Brazil and Indonesia. Price elasticities of demand (short-run and long-run)

were inelastic for all consuming countries. Income elasticities for Japan and Middle

Europe were elastic, but were inelastic for Western Europe and the United States.


Akiyama and Varangis (1989) modeled the impact of the ICA/export quota system

on the world's coffee market. The model differed from the earlier work by solving for

world prices that equilibrate export supply and import demand in the quota market. In

the earlier model price was solved as function of stocks and world demand.

Additionally, the model incorporated the vintage capital approach to the specification of

supply response. The results showed that the quota system had a stabilizing effect on

world coffee prices in the period 1981-85 period. However, the quota reduced export

revenues (in real terms) for all but the larger producing countries, Brazil and Colombia.

These countries are said to gain from the scheme because they face small or even zero

marginal export revenue for increased exports due to their large market shares.

Vogelvang (1992) developed hypotheses tests to evaluate the long run relationship

between the four indicator (spot) prices of coffee. The price formation of coffee was

investigated and the univariate properties of coffee were checked. Co-integration tests

were performed. The tests showed that certain vectors were statistically significant,

Colombian Mild and Other Mild, Robusta with Other Mild, and Robusta with Colombian

Mild. The other relationships were rejected statistically. Of interest is the relationship

of Other Mild and Brazil Arabica, although these are Arabicas, the relationship was not

statistically significant. The reason offered is that Other Mild prices are more reflective

of free-market conditions than the Brazilian Arabica, because of special deals Brazil

negotiates with leading coffee importers.

Karp and Perloff (1993) examined the oligopolistic structure of the coffee market

on the supply side. A linear-quadratic, dynamic feedback oligopoly model was used to


estimate the degree of competitiveness and the adjustment paths of the two largest coffee

exporters, Brazil and Colombia. This model was also compared to a standard static

oligopoly model and the open-loop model. The results showed that Brazil and Colombia

compete vigorously with each other in supplying the market. In addition, both countries

behavior was closer to price-takers than that of market collusion.

Bohman et.al (1996) investigated rent seeking activities and the effect of the ICA

export quotas. A theoretical model was used to show how the imposition of the ICA

quotas facilitates the creation of potential rents within coffee importing countries, and

how competition for these rents may affect coffee allocation, income distribution and

welfare in Indonesia. It was concluded that although benefits may accrue to government

treasuries and coffee farmers, rent seeking losses can outweigh the gains from the higher

prices as a result of the ICA quotas. The empirical evidence supports the argument that

there is little justification for higher prices charged to consumers in the member

importing countries as compared with the worsening effects of directly unproductive

activities (DUP) on the distribution of income in the member exporting countries.

Summary of Previous Coffee Demand Studies in the U.S.

There have been few empirical studies addressing demand for coffee in the U.S.

The first studies after the Second World War were those of Daly (1958) and U.S.

Department of Commerce (1961). These studies used time series data and revealed that

the income elasticity of demand for coffee in the U.S. was positive but was slightly lower

than it had been prior to WW II. The reasons offered included distribution of


population, the increase in consumption of instant coffee versus regular coffee and the

increased efficiency in the extraction rate of instant coffee.

Contradictory results were seen in the studies conducted later by lovasy (1967)

and Timms (1973). The results showed that the income elasticity of demand for coffee

was negative. This was attributed to saturation of demand for the commodity, and a shift

in tastes and preference.

Parikh (1973) established the hypothesis that total annual coffee consumption is

a stable function of income and that variation of imports were due mostly to variation in

inventory of these products. Demand was shown to be inelastic.

Gray (1971) showed that the own-price and income variables were limited in

explaining the downward trend in the consumption of instant coffee. The focus was on

varieties of coffee traded (Robusta, Arabicas, and Brazil's Arabica). Very high values

of elasticities were seen for income and own-price elasticities of demand.

These studies were done for time horizons in the 1960's and 1970's. The 1988

study published by the International Monetary Fund showed that both short run and long

run price elasticities of demand were inelastic, indicating that coffee demand is not very

responsive to price changes.



The conceptual framework and analytical model of the study are presented in this

chapter. The assumptions within the conceptual framework are discussed. The analytical

model is presented in two parts. First, to satisfy equilibrium conditions, aggregate export

supply and demand are specified in the framework of optimization of industry surplus,

i.e., optimizing consumer and producer surpluses within the industry as a whole.

Second, the conditional U.S. import demand for coffee by type is specified. Finally, the

estimation methods are presented.

Conceptual Framework for Aggregate Supply and Import Demand

In modeling the demand for coffee the assumption of a unidirectional cause and

effect relationship between price and quantity cannot sufficiently explain the true nature

of the market. In fact, the very nature of the policies affecting coffee implies that there

is a simultaneous relationship in the export supply and import demand for coffee.

The logical schematics of the problem can be represented in figure 2. Assume

a stable demand DD, with the imposition of a global export quota in different periods

resulting in supply curves So and S,. The equilibrium prices and quantities are po, q0,

pl, and q, in the two periods. In a non-quota regime the supply curve would lie to the

right of the quota imposed supply curves in each period, S'o to S'j. Equilibrium prices

pr ice





q q 0 q q
0 o

Figure 2. Interdependence of Price, Quantity and Quota Effects

quiant i ty


p0' and Pl' would be lower in each period than a quota derived equilibrium, with larger

equilibrium quantities qo' and q,'.

Demand and supply can be represented by the following behavioral equations with

an identity for equilibrium.

qt ao +alpt +et (1)

%t --/lo +01 p, +e2, (2)

d = qt (3)

where qd, is quantity demanded in time period t, q,, is quantity supplied in period t, and

Pt is the price in period t. It is assumed that al is less that 0 and f, is assumed to be

greater than 0. If for some reason e2, in (2) is correlated to changes in other factors

affecting q, (e.g., weather, policy), the supply curve will shift outward or inward

depending on whether e2, is positive or negative. As shown in figure 2, a shift in the

supply curve affects both p and q. The same is also true for e,,, a shift in the demand

curve. As a result of these changes there is simultaneous dependence between quantity

and price. The assumption of no correlation between the predetermined variables and

the disturbance term is therefore violated given the assumption made for the classical

linear regression model (Greene, 1990).


Recognizing the problem of simultaneity, there are other concerns that the model

must also address. These relate to the international trade framework in which coffee is

traded (especially U.S. import demand by type). The trade policy framework can be

developed for coffee as represented in figure 3.

Assuming exporting countries operate under perfect competition and non-binding

quota conditions the equilibrium price and quantity are p* and q*. The excess supply

curve (ES) in the trade sector is the horizontal difference between the domestic supply

and demand curves when supply exceeds demand. The corresponding demand curve in

the trade sector is the sum of the demand curves for the U.S. and the rest-of-the-world.

The U.S. and ROW do not have supply curves because they are not involved in the

primary production of coffee. The equilibrium price and quantity with the non-binding

quota is P,, and Qt respectively, where Pw, is the international price that importers (in the

U.S. and ROW) will have to pay to purchase the quantity Qt. By imposing a quota of

Qq, the equilibrium price is Pwq which is the price that U.S. and ROW importers would

pay to buy the quota quantity Qq. Pwq is greater than p,, i.e., the equilibrium price

importers face when the quota is binding is greater than the equilibrium price when the

quota is non-binding.

International trade will yield one price (net of transportation costs and duties)

where world imports must be equal to the world exports. All markets must clear and

equilibrium is where the excess demand curve intersects the excess supply curve.

prrce ES' ES

CO quent I ty

Exporting Country-Bloc


Oq Ot quant i ty

Trade Sector


I Il

lus quantity
Iniortlng Country (U.S.)

Figure 3. Trade Policy Framework for Coffee



In this case area abed is the quota rent which goes to the exporting country and

area bce is dead weight social loss resulting from the imposition of the quota. In the

case of welfare, there is an expected loss to the consumers in the importing country. The

welfare loss can be decomposed into distributional effects, where there is a reduction in

the amount of coffee consumed as a result of higher prices and the volume-of-trade effect

is lower due to a fall in the volume of coffee traded.

Modeling Aggregate Export Supply and Import Demand of Coffee Traded

In the survey of empirical work in international trade, the research can be divided

into two broad categories (Kholi, 1991). The first deals with studies related to the

verification of different aspects of trade theory, especially related to the Hecksher Ohlin

Samuelson (HOS) model.

The second category covers those studies that deal with the estimation of import

and export functions. These studies examine the effects of relative prices and volume

of trade, based on various specifications, and have been sector specific, global or

country-wide. The earlier models were endemic with the following problems. They

were based on little or no trade theory, and elasticities that differed significantly from

unity were difficult to interpret. In addition, the aggregation of domestic variables

appears to be excessive. Also, the use of single equation techniques was inappropriate.

The linear and log linear functional forms were said to be inadequate, and too little

attention was paid to the definitions of the explanatory variables.4

4 Although these problems have been recognized, most of the models estimated are still
specified as log linear functional forms. However emphasis has been placed on the definition and
inclusion of explanatory variables based on a priori reasoning and theory, and estimation using


As a result, models are now being developed on the foundation of trade theory,

however certain assumptions have to be invoked in order to facilitate applied work. For

example, in modeling conditional U.S. import demand at the disaggregate level in this

study requires the assumption of separability of preferences. This allows commodities

to be partitioned into commodity groups, such that preferences within groups (say food

as in this case) can be described independently of quantities in other groups (Deaton and

Muellbauer, 1980, p 122).'

General Export Supply and Import Demand Functions

To establish the functional forms of the export supply and import demand, the

optimization of industry surplus (consumers and producers surpluses) is examined within

the international trade framework as shown in figure 3. The objective function

maximizes industry surplus subject to market clearing conditions. Using the Lagrangean

(L) to state the problem, assuming no transfer costs, yields the following

L=[Jd (q) dq S (q,) dq,] +M(Q q)-X (Q -q,) (4)

systems of equations. The models are estimated as log linear forms and the definition and
inclusion of explanatory variables are based on a priori results and international trade theory.
In addition, they are estimated as systems of equations. The aggregate model is nonlinear and
a nonlinear estimation technique was used.

This concept will be further elaborated in sections dealing with the development and
estimation of the disaggregate U.S. import demand model.


where qd is the quantity demanded in the importing country bloc, q. is the quantity

supplied by the exporting bloc, Q is the available quantity on the market, and must be

equal or greater than the quantity demanded and less than or equal to the quantity

supplied, d(qd) and s(q5) are the inverse or price-dependent demand and supply functions,

respectively. The shadow prices for supply and demand are PL and X, respectively.

The Kuhn-Tucker condition, can be used to solve for the export supply and import

demand equations, and the international clearing price.

"a =d ( qd) S0 (5)

Ua (q ) =0, qd>O(6

aL (6)

U -S(q,) +X
aL (8

(q,) =0, q,2:o (8)


(Q ) =0, ifQ O, thenX =Ij (10)


Equations (5) and (7) are the first-order derivatives showing import demand and export

supply. In addition to prices, other economic variables also affect export supply and

import demand. Given the functional forms with the inclusion of non-price variables

and the error term, the unknown parameters can be estimated.

Aggregate Supply Equation

The first order conditions of the aggregate model can be used to specify the

structural export supply equation,

q.,= f ( Sp,, INV,-,, TP,, D, DPRI, ) (

where q,, is quantity of all coffee exported worldwide in period t in pounds, spt is the

price (New York spot market price) the group of export firms face in period t in cents

per pound, TP, is total production, INV,, is the carry over stock or inventory of all

coffee at the beginning of period t in pounds, D is (the proxy for the export quota) a

dummy variable (D=1 when the export quota is binding and D=0 when it is non-

binding) and DPRI, is an interaction of the policy (D) and price variables.

Aggregate Import Demand Equations for the U.S. and Rest-of-the-world.

Similarly, the following equation represents the U.S. import demand for all


qus = f ( sp,, USPCDI,, TEAP, COCP,)



where qua is the quantity of all coffee imported into the U.S. in period t in per capita

pounds, sp, is the price (New York spot market price) import firms face in period t in

cents per pound, USPDI, is the U.S. per capita disposable income in period t in U.S.

dollars, TEAPt is the price of tea in period t in cents per pound and COCPt is the price

of cocoa in period t in cents per pound. Income and prices are deflated by the consumer

price index (CPI).

The import demand for the rest-of-the-world can be expressed as

qRow, = f ( spt, WPCGD P,, TEAP,, COCPt) (13)

where q1owt is the quantity of all coffee imported by the rest-of-the-world in period t in

per capita pounds, WPCGDPt is the world per capita gross domestic product in time

period t in constant U.S. dollars, and spt, TEAP, and COCP, are as defined in equation

(12). Additionally, the identity holds where the quantity supplied is equal to the quantity

demanded in the U.S. and rest-of-the-world,

%t = qust + qRowt" (14)

Modeling Conditional U.S. Import Demand for Coffee by
Type and Importers' Preferences

Demand for a commodity depends on the price of the commodity in question as

well as prices of substitutes and complements, income/expenditure levels, and other

significant economic variables, e.g., population. A two-stage budgeting process might


be used to describe this demand structure. For this budgeting process, Consumers first

decide on how much to consume of a particular bundle of goods. Then, they decide how

much of each product in that bundle to buy. For the present study, once the total

consumption level of the bundle (market demand for coffee) has been determined, the

allocation among different supplies (product demand) has to be made.

Coffee like many other product categories consists of products that are

distinguished by types. The types within this product category are not considered perfect

substitutes. U.S. import demand is developed from a strongly separable (additive) utility

function which can be written as

U = f(v(Q0) + u(q)) (15)

where Q0 represents the aggregate consumption of a numeraire good, q ia a vector of

quantities for different types of coffee, and v and u are sub-utility functions.

With the assumption of separability of preferences, coffee can be partitioned as

a group of goods so that preferences within that group can be described independently

of the quantities in other groups of goods (Deaton and Muellbauer, 1980, p 122-125).

The coffee subutility function can be further written as


u(q) =U (qcoL, qom, qUA qRO) "

where q represents the vector for consumption of the different types of coffee,

Colombian Mild (qcOL), Other Mild (qo., Brazil and Other Arabica (qBA) and Robusta


Given this assumption, the quantity of different coffee types purchased can be

written as a function of the group expenditure and prices within that group. Formally,

the U.S. import demand function for a particular coffee type is6

qusit = f (pt Et) (17)

where qusit is the per capita quantity of coffee type i (Colombian Mild, Other Mild,

Brazil and Other Arabica, and Robusta) imported into the U.S. in time period t in per

capita pounds, pt is the vector of prices for types (p.... p4) of coffee in time period t in

cents per pound, and E is the expenditure on all coffee imported into the U.S. in time

period t in per capita U.S. dollars.7

6 These demand functions are derived as Marshallian Demand curves using the Lagrangean
technique to impose the budget constraint. Additionally, this research is conducted at the import
level, using shipping point data instead of consumer level data. Hence, the results obtained are
applicable to shipping point or wholesale level.

When separability is assumed, the expenditure variable could be endogenous
(LaFrance,1991). This possibility of simultaneous equation bias can be tested. In the present
research the Hausman specification test (Hausman, 1978) was used to test for possible
endogenous expenditure. The chi-square statistics with seven degree of freedom were 3.80, 3.25,
0.027 and 0.022 for the different types. The tests indicated a failure to reject the null
hypotheses at the 10% level of statistical significance, reinforcing the assumption that there is
no expenditure endogeneity. Several works on import demand have justified the incorporation
of expenditure instead of income in the demand function (M. Goldstein and M.S. Khan, 1985;
and A. Brillembourg, 1975).


The consumption of coffee is driven partially by the importers behavioral patterns

and market demand. Both economic and non-economic factors stimulate any deviation

from these consumption patterns. Behavioral patterns toward routine purchases of a

product are frequently referred to as habit persistence. While coffee purchases do not

show complete habit persistence, some element of it exists and must be considered when

determining strategies to influence demand and/or project expected demand for future


Let q%, be (actual) per capita consumption of coffee type i in time period t, with

Pt and E, being factors influencing demand. To incorporate habit persistence, we

assume that the relative change in actual consumption is dependent on the desired level

of consumption relative to the actual consumption in the previous period. This is

represented as,

qt = (%qit )Y (18)
qt -~ -

where q,. is the desired level.

Incorporating the independent variables, the desired quantity (cht.) can be

expressed as

q i f ( p , E ) .(19 )


Substituting (19) into (18) for ci," and applying logarithmic transformation, the

conditional import demand for coffee can be expressed as

in q=-yin(f ( p, ,E, )) +(1--7) in % (20)

where In qt, is a one period lag of the dependent variable.

Using the boundaries 0 < -y! 1, if -y = 1, then in equation (20) there is no habit

persistence and the importation of coffee is dependent only on p, and Et. If -y = 0, habit

persistence is perfect and importation in this period is equal to importation in the

previous period; here Pt and F, have no impact on the import demand.

Estimated Equations and Proposed Models

The following structural equations for aggregate supply, inventory and demand

for coffee were estimated as double logs with the identity holding.

lnq.,=ao+a, In spt+a2 in INV,-, +a3 InTP,+a4D+a5 (D in sp, ) +e,, (21)

in qus, =#o+f, in sp1t+t2 In USPCDI,+3 in TEAP,+g4 in COCPt+et (22)

in qRowtyo + 1 in spt+ y2 in WPCGDP,+y3 in TEAPt+,74 in COCP+e3, (23)


q = qust + qROw


The a's and O's and -y's are coefficients to be estimated. Based on equation (20) the

following equation for conditional U.S. import coffee demand by type was estimated

in qusit f6o+6, in E,+ 62 in p,+ 63 in qjt -I +vii (25)

The 6's are coefficients to be estimated with 6i being a vector of coefficients and 6j3

equal to 1 -y.

Estimation Methods

The structural equations for the aggregate export supply (equation 21), and

import demands (equations 22 and 23) were estimated as a system. The total quantity

exported is expressed as a function of the New York spot market price, the annual

beginning inventory level, total production, a dummy variable as a proxy for trade policy

and an interaction variable consisting of the dummy and the price. The New York spot

market price was included based on evidence from previous research showing its

significance in influencing the market and its use in guiding the ICO price indicator

(Ford, 1977; Vogelvang, 1992). It is quoted on the basis of Unwashed Arabica which

is produced by Brazil, the dominant producer and trader. Akiyama and Varangis (1989)

suggested the inclusion of carry over stocks (inventory) and total production as

explanatory variables in the supply model. Inventory levels are expected to affect release

quantities or supplies. Increases in inventory levels are expected to result in increased

release quantities and result in a decline in prices and vice versa. Similarly, total

production will also influence release quantities or supply. The dummy variable was

used as a proxy for the export quota policy to examine shifts in the intercept and price


slope. The dummy variable was interacted with the price variable to examine any policy-

induced change in the price slope of the regression. A binding export quota will reduce

supplies in the market.

Aggregate U.S. import demand was specified with per capita quantity of all coffee

imported as a function of the New York spot market price, U.S. per capita disposal

income (constant U.S. dollars), and prices of tea and cocoa to capture expected

substitution relationships. The spot market price was included because import demand

is estimated at the wholesale level.

The rest-of-the-world demand equation was specified as a function of the New

York spot market price, world per capita gross domestic product and the prices of tea

and cocoa. The spot market price was included because it represents the international

price (single price) for coffee. World per capita gross domestic product (constant U.S.

dollars) was used as a proxy for income to examine the causal effect it is likely to have

on demand.

Quantities were expressed on a per capita basis.8 The spot market price, tea and

cocoa prices were deflated by the consumer price index. The system was estimated as

double log model system of equations. Based on the conceptual framework and

economic theory, a simultaneous equation estimator was deemed as the most appropriate

approach to be used to estimate the aggregate export supply and demand equations. The

identification process for a nonlinear simultaneous system is quite different from that of

8 U.S. population figures were used for U.S. import demand. Population figures based on
region were used for the Rest-of-the World import demand (Urban and Nightingale, 1993).


a linear system (Fisher, 1976). In order to satisfy the identification process, the

nonlinearity of the variables must be considered. In this study the nonlinearity of the

variables is the result of the export supply and import demands connected by a nonlinear

identity. This allows for additional variables from the nonlinear condition to be used to

identify the equations. In addition, because of the nonlinearity, the closed form

analytical solutions for the reduced form equations do not exist, but can be solved

numerically (Greene, 1990). Order (necessary) and rank (sufficient) conditions were

satisfied in the identification of the above equations. Each equation is over identified.

In a system with over identified equations there is more information than is required for

the estimation of the parameters. An appropriate estimator must be employed to take

into account this situation and to yield estimates that are asymptotically efficient and

consistent. The appropriate system estimator used for this model is nonlinear three stage

least squares (NL3SLS). The TSP econometric package (Hall et al., 1995) was used to

generate 3SLS estimates. The instrumental variables consisted of all the predetermined

variables, their squares and cross products, and the constant term (A. Gallant, 1987).

Evidence for structural changes based on different trade policy regimes were evaluated

in the aggregate model.

The equations for conditional U.S. import demands for coffee by type (equation

25) were estimated as a second system under the assumption of cross equation correlation

of the disturbances. Each equation in the model was specified as a function of the

overall expenditure on coffee, prices for all types and a one period lag on the dependent

variable (used to incorporate the importers' preferences by type). The four demand

equations for the different coffee types were estimated in double logs. Quantities were

converted to a per capita basis. To impose homogeneity, prices were normalized by one

of the coffee prices ( i.e., Robusta price), with expenditure also normalized by the

robusta price and expressed on a per capita basis. The equations were operationalized

by adding the respective error terms. Parameter estimates are expected to be consistent

and asymptotically efficient for the model. Given the assumption of cross-equation

correlation of the error terms, the system of equations was estimated using Seemingly

Unrelated Regression Techniques (SUR) (Greene, 1990). TSP econometric package was

used in estimating the models. Structural change was also evaluated using dummy

variables for different policy regimes. The Wald statistical test was used to evaluate

structural changes based on different policy regimes. The restrictions of adding up and

symmetry were not imposed due to the double log specification of the model (Deaton and

Muellbauer, 1980, pp 17 and 45).9

9 In demand analysis certain restrictions are imposed on the derivatives of the demand
function rather than the function themselves (Deaton and Muellbauer, 1980). These are used
to restrict the parameter space and to allow the function specifications to conform to the
properties of the demand function. In general, the restrictions of adding-up and homogeneity
are imposed as a consequence of the assumption of a linear budget constraint and, symmetry and
negativity are derived from the second order condition. However, models can be specified
without the imposition of one or more restrictions. For instance, in the specification of double
log demand models adding-up and symmetry cannot be imposed. In the international trade
literature, double log models are used to estimate import demand without the imposition of
restrictions (A.Intalianer and G. d'Alacantara, 1986). The current research evaluated several
specifications for the disaggregated import demand model. This was because the data on prices
were suspected to show moderate to severe multicollinarity as checked by the examination of
the condition number (701) and condition index (26) (see Greene 1990 p 35, 280-281). In
addition, Dickey-Fuller tests (degrees of freedom 30, 1) were done on all the endogenous and
exogenous variables. The tests revealed that the endogenous variables (quantity values of 4.85
to 6.56) were stationary and exogenous variables (price values of 1.6 to 1.48 and expenditure
value of 2.61) were non-stationary. The double log model with homogeneity imposed, yielded


Annual data for the period 1964-94 were obtained for use in the analysis. The

primary data sources were the International Coffee Organization, USDA Foreign

Agricultural Service, the U.S. Bureau of Labor Statistics, U.S. Department of Labor and

Bureau of Census, Economic Report of the President, and the World Bank.

Time series data for imported quantities, prices, annual total production and

inventory were obtained from the International Coffee Organization (unpublished data)

and the USDA Foreign Agricultural Service (United States Department of Agriculture,


World per capita gross domestic product was taken from World Bank Data Tables

(World Bank, 1995). U.S. per capita disposable income, the consumer price index (CPI)

and the U.S. population were taken from the U.S. Bureau of Labor Statistics, U.S.

Department of Labor and Bureau of Census, and the Economic Report of the President

(1986). Data on world population by country and region were obtained from Urban and

Nightingale (1993).

the reported estimates. However, caution must be exercised in the interpretation of those
estimates, and especially inferences with policy implications, given data problems discussed.



Empirical Results

The results are presented in this chapter for the aggregate world trade model and

the disaggregate model for U.S. import demand by coffee type. Based on statistical

tests, structural changes both in the aggregate and disaggregate models were suggested.

Results are presented based on these tests and from the pooled sample. A summary of

the results is presented with a discussion about the implications for the industry.

Aggregate Export Supply and Import Demand For Coffee

Table 8 presents the estimated coefficients of the aggregate structural model and

their significance for the pooled sample. The estimates of the explanatory variables are

reported, and are evaluated at statistical significance levels of 5% and 10% (i.e., a =

0.05 and 0.10). The export supply curve was upward sloping, i.e., the sign on the

estimated parameter for the spot market price was positive, indicating that as price

increases the quantity supplied also increases. Similarly, coefficients were positive for

inventory and total production.

All explanatory variables in the export supply equation were statistically

significant at the 5% level, except for the policy and the interaction variables. The

results suggest that supply is price inelastic with a 1 % increase in the price of coffee

causing an increase in the quantity of coffee supplied by 0. 13 %.


Table 8.

Parameter Estimates for the Aggregate Structural Model for Export
Supply, Aggregate Demand for Coffee in the U.S. and Rest-of-the-world
Over the Pooled Sample (1964 -1994).

Variables' Aggregate Aggregate U.S. Aggregate Rest-of-
Export Supply Coffee Demand the-world Demand
Intercept 14.21 20.04 21.11
(2.25)"a (2.42)" (1.03)"
Spot market Price 0.13 -0.17 -0.13
(SP) (0.02)- (0.06)- (0.05)"
Inventory (INV,,) 0.11 N.lb N.I.
Total Production 0.13 N.I N.I.
(TP) (0.07)"
World Per Capita N.I. N.I. 0.19
GDP (WPCGDP) (0.10).
U.S. Per Capita N.I. 0.32 N.I.
Disposable Income (0.12)"
Policy (D) -0.05 N.I. N.I.
Policy*Price 0.02 N.I. N.I.
(DPRI) (0.07)
Tea Price N.I. 0.06 -0.03
(TEAP) (0.10) (0.07)
Cocoa Price N.I. 0.08 -0.08
(COCP) (0.06 (0.05)

1 Variables were estimated as logs except the dummy(D). The variable DPRI is the
dummy times log of price.
" Standard errors are presented in parentheses below the parameter estimates.
*Estimate is statistically significant at 5 %level of significance.
-Estimate is statistically significant at 10%level of significance.
b N.I. indicates this variable was not included in the corresponding equation for


Inventory or carry over stock was statistically significant at the 5% level. The

estimated coefficient for this variable shows that a 1 % increase in inventory will result

in a 0.11% increase in quantity supplied. Similarly, total production was statistically

significant at the 5% level with its estimated coefficient value at 0. 13. This result

indicates that a 1% increase in total production of coffee results in a 0.13% increase in

quantity supplied.

The estimated aggregate demand curves were downward sloping with negative

coefficients on the spot market price variable, i.e., the quantity of coffee demanded

declines as price increases as suggested by theory. The signs of the coefficients on

world per capita gross domestic product (world income) and U.S. per capita disposable

income (U.S. income) were positive, indicating that quantity demanded increases as

income increases. The increase in income denotes an outward shift in demand at given


All parameter estimates in the aggregate U.S. coffee demand equation were

statistically significant at the 5 % level, except for those on the prices of tea and cocoa.

The own-price elasticity was -0. 17 for coffee in the U.S. This indicates that a 1%

increase in the price of all coffee results in a decline in the quantity demanded in the

U.S. by 0.17%. If we assume equi-proportional effects between the wholesale and retail

prices, symmetric price transmission, then a 1 % increase in price at those levels will

affect the quantity demanded in the same proportion (0.17 %). These results suggest U.S.

aggregate coffee demand is price inelastic, implying that while there may be close

substitutes across coffee types, coffee in general has no major substitutes.


In terms of U.S. per capita disposable income, the sign indicates that as income

rises the quantity of coffee demanded increases. The coefficient was statistically

significant at the 5% level. The results showed that a 1 % increase in U.S. income

increases demand in the U.S. for all coffee by 0.32%, indicating that coffee demand has

grown at a slower pace than changes in income.

All explanatory variables for the rest-of-the-world (ROW) demand equation were

statistically significant, except for the tea and cocoa price variables. The demand curve

was shown to be downward sloping. The results indicate that a 1 % increase in price

causes ROW demand for all coffee to decline by 0.13%. Again coffee is price inelastic.

The parameter estimate for world per capita gross domestic product (world

income) showed that as world income rises, the quantity of coffee demanded increased.

The coefficient was statistically significant at the 10% level. However, coffee is income

inelastic, indicating that ROW coffee demand has grown at a slower pace than changes

in income. The estimate showed that ROW coffee demand increased by 0.19% with a

1 % increase in income.

Table 9 presents the aggregate supply and demand parameter estimates by trade

policy regimes (export quota binding vs quota non-binding periods). The data were

partitioned by the trade policy variable D, into quota binding versus quota non-binding

periods and partition-specific aggregate supply and demand estimates were obtained. The

export supply was upward sloping in both periods, i.e., the sign on the two parameter

estimates for the spot market price were positive. However, the coefficient on the spot

Table 9.


Parameter Estimates for the Aggregate Structural Model for Export Supply, Aggregate
Demand for Coffee in the U.S. and Rest-of-the-world with the Export Quota Policy

Binding (D= 1) and Non- Binding(D=0),(1964 -1994).

Variables' Aggregate Export Supply Aggregate U.S. Coffee Aggregate Rest-of-the-
Demand world Demand

(D= 1) (D=0) (D= 1) (D=0) (D= 1) (D=0)

Intercept -2.97 15.38 -7.47 25.06 3.74 17.93
(4.33) (3.27)- (5.22) (4.39)- (2.48) (1.09)"

Spot market 0.08 0.14 -0.27 -0.36 -0.05 -0.14
Price (SP) (0.09) (0.03)- (0.13)- (0.09)Y (0.13) (0.08)"

Inventory 0.15 0.01 N.I.1 N.I. N.I. N.I.
(INV,.) (0.08)- (0.07)

Total 0.04 0.20 N.I. N.I. N.I. N.I.
Production (0.15) (0.11)-

World Per N.I. N.I. N.I. N.I. 0.43 0.21
Capita GDP (0.28) (0.23)

U.S. Per N.I. N.I. 0.72 0.82 N.I. N.I.
Capita (0.50) (0.43)-

Tea Price N.I N.I. 0.09 0.04 0.04 -0.13
(TEAP) (0.21) (0.17) (0.16) (0.14)

Cocoa Price N.I. N.I. 0.23 0.22 -0.05 0.02
(COCP) I_1 1_(0.14) (0.18) (0.13) (0.12)

Variables were specified as logs expect the dummy (D) and DPRI is the dummy times log of prices.
a Standard errors are presented in parentheses below the parameter estimates.
*Estimate is statistically significant at 5 %level of significance.
Estimate is statistically significant at 10%level of significance.
N.I. indicates this variable was not included in the corresponding equation for estimation.


market price was not statistically significant (a = 0.05) during the export quota binding

period. Similarly, the sign on the coefficients for inventory and total production were

positive, but the inventory coefficient was statistically significant (a =0. 10) only when

the export quota was binding, and the total production coefficient estimate being

significant (o = 0.10) only when the quota was non-binding. This indicates that during

the binding quota period an increase in carry over stocks will stimulate an increase in

supply, while when the quota is non-binding an increase in total production (i.e., fresh

out from the field) will cause an increase in the quantity supplied. It can be deduced that

during the binding quota period inventory levels influenced supply decisions, while

during non-binding quota period total production influenced supply decisions.

The results showed that supply during the non-binding quota period was price

inelastic at 0.14, with a 1% increase in the spot market price resulting in a 0.14%

increase in shipments of coffee. Also, a 1 % increase in total production increases

exported quantities 0.20% during the non-quota binding period. During binding quota

periods, a 1 % increase in carry over stocks or inventory level causes a 0. 15% increase

in export quantities supplied.

The demand curves were downward sloping with negative coefficients on the

price variables for U.S import and rest-of-the-world demands in both trade policy

regimes. These indicate that the quantity of coffee demanded declined as price increased

given the trade policy regimes. The signs of the coefficients on the variables used as

proxies for world and U.S. incomes were positive for both the non-binding and binding

quota periods. These indicate that quantity demanded increased as income increased.


Parameter estimates in the aggregate U.S. coffee import demand were

statistically significant (a = 0.05) for prices in both trade policy regimes, and for

income (a = 0.10) during non-binding quota periods. An inverse relationship was seen

with respect to the quantity demanded and the spot market prices in both quota binding

and non-binding periods. Aggregate U.S. coffee import demand was price inelastic in

both trade policy regimes. The price elasticities were -0.27 and -0.36 in quota binding

and non-binding periods respectively. That is, a 1% increase in the price of coffee

results in a decline in total quantity demanded by 0.27% and 0.36% in quota and non-

binding periods respectively.

Aggregate U.S. coffee import demand was income inelastic in the non-binding

period. The results show that with 1% rise in U.S. per capita income, coffee demand

increases by 0.83%. It can be deduced that coffee demand will grow at a slower pace

than changes in U.S. income.

Rest-of-the-world demand was price inelastic. The parameter estimate was -0.14,

indicating that a 1 % increase in the price of coffee will decrease the demand for coffee

in the rest-of-the-world by 0.14%. The parameter for world income (world per capita

gross domestic) was not statistically significant during either trade policy regime.

Likewise, for the rest-of-the-world demand, as well as U.S. demand, the coefficients on

tea and coffee prices were not statistically significant during either trade policy regime.

Test for Evidence of Structural Changes in Aggregate Model

Tests for equality of parameters in the binding and non-binding export quota

periods were done for the aggregate export supply equation, and U.S and rest-of-the-


world import demand equations. The Wald statistical test for the export supply equation

yielded a chi-square value of 13.02 with four degrees of freedom, greater than the table

value (9.49) at the 5% level. Hence, the null hypothesis was rejected, indicating that

there is structural change in export supply given the ICA agreement.

The Wald statistical tests results for the U.S. and rest-of-the-world import demand

equations gave a chi-square value of 24. 10 with ten degrees of freedom, which was

greater than the table value (18.31) at the 5% level. The null hypothesis was rejected,

indicating that there is structural change in U.S and rest-of-the-world import demands

given the ICA agreement.

Conditional U.S Import Demand by Coffee Types

The parameter estimates and standard errors of the structural model for U.S

import demand by coffee types are presented in table 10 and represent short-run estimates

of elasticities. Estimates of "long-run" import demand elasticities may be obtained by

dividing the short-run import demand estimates by one minus the coefficient value on the

lagged dependent variable, after dropping that variable. All of the estimated demand

curves were downward sloping with respect to the own price variable. The coefficients

on the coffee expenditures were positive, indicating that the amount of consumed coffee

increased for each coffee type as expenditure increased or demand shifts outward as

expenditure rises.

Table 10. Parameter Estimates for the Structural Model for Conditional U.S. Import
Demand of Coffee by Type (1964 1994)

Variables' Colombian Other Mild Brazil and Robusta
Mild Other

Intercept 8.41' 9.13* 10.72* 9.10"
(0.54)' (0.65) (0.68) (0.72)
Expenditures (E) 1.03' 0.91" 1.03" 1.15*
(0.09) (0.10) (0.09) (0.12)
Price Colombian -0.01 1.22" -1.11' -1.65"
(PCOL) (0.28) (0.34) (0.39) (0.42)
Price Other Mild -0.28 -1.33' -0.46 -0.82""
(POM) (0.37) (0.44) (0.50) (0.51)
Price Brazil and -0.74" -0.80' -0.26 1.40'
other Arabicas (0.30) (0.34) (0.39) (0.43)

Price -0.91' -0.88' 0.88' -1.12'
Robusta (0.28) (0.32) (0.37) (0.38)
Importers 0.17" 0.13 -0.04 0.15"
Preferences (Q,-,) (0.07) (0.08) (0.08) (0.08)

'Variables were estimated in logs.
'standard errors are in parenthesis.
'Estimate is statistically significant at 5%level of significance.
"Estimate is statistically significant at 10%level of significance.

Colombian Mild. Coffee classified within this group is produced mainly in

Colombia, Tanzania and Kenya. These countries are the main suppliers of coffee to the

U.S. market. The expenditure elasticity was significantly different from zero (a =

0.05); the sign was consistent with expectation. The expenditure elasticity estimate was


slightly greater than unity (1.03) indicating that a 1% increase in expenditures will

stimulate a 1.03% increase in the total quantity of Colombian Mild consumed. Own

price elasticity was not statistically significant (a = 0.05). This suggests a weak

relationship between own price and the quantity of this type of coffee consumed. All

the cross price elasticities had negative signs indicating that Colombian Mild has a

complementary relationship with other coffee types. The price variables for Brazil and

Other Arabicas (-0.74) and Robusta (-0.91) were statistically significant (a = 0.05).

This result is consistent with trading activities in the market where Brazil and Other

Arabica, and Robusta are used as blending coffees with Columbian Mild. The results

suggest persistence in the import demand for Colombian Mild." The lagged-quantity

coefficient (0.17) was statistically significant at the 5% level. Current consumption is

influenced significantly by previous consumption, as reflected by this variable.

Other Milds. The main producers and sources of other mild imports into the U.S.

are Central and Latin American Countries, India, Dominican Republic, Haiti, Jamaica,

Burundi and Malawi. However, most of the ICO members growing this type of coffee

are exempt from the basic export quota (Table 5). The expenditure elasticity estimate

was statistically significant (a = 0.05) at 0.91, indicating that a 1% increase in

expenditure results in an increase in total demand for Other Milds by 0.91%. This

coffee type also dominates the high end of the market, especially with the rapid

proliferation of specialty coffee stores in upscale metropolitan areas. The own price

10 This result was supported through the evaluation of the importers preference variable in
the model using the Durban-H statistic. The calculated Durban-H statistic was 0.65 for
Colombian Mild, 0.71 for Robusta indicating no evidence of serial correlation.