IJRPAhY '
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

ESSAYS ON SPECIALTY COFFEE PROCUREMENT

presented by

M. LAURA DONNET

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Agricultural Economics

 

 

    

Major Professor's Sign urev

XflV/fl?‘

Date

MSU is an affinnative-action, equal-opportunity employer

- -.--— --A- ..-.-o---o-----.-.-c-—-o--—.-------n-n-.-.-.-u—.-.-.-----1-.-

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

lo8040§

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6107 p:lClRC/DateDue.indd-p.1

 

 

ESSAYS ON SPECIALTY COFFEE PROCUREMENT

By
M. Laura Donnet

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Agricultural Economics

2007

ABSTRACT

ESSAYS ON SPECIALTY COFFEE PROCUREMENT

By

M. Laura Donnet

Specialty coffees distinguish themselves by the quality of the coffee beans from what it is
made. The demand for specialty coffees is growing consistent with the general trend in
food upgrading (more quality less quantity). Specialty coffees are a differentiated product
with complex and variable quality. This poses the challenge of coordinating the supply of
premium quality beans with the consequent opportunity of capturing the specialty
demand and increasing revenue of supply chain participants. This dissertation studies the
coordination of procurement in the supply chain; in particular, product differentiation and
valuation and the choice of procurement strategies in three separate papers.

The first paper analyzes the determinants of specialty coffee prices traded at e-auctions
by estimating a hedonic price function on 541 coffees from the ‘Cup of Excellence’. The
analysis tests the hypothesis that specialty coffee prices are determined by both sensory
and reputation attributes, as evidence from other specialty products suggests. Results
show that specialty coffee prices are influenced by both sensory attributes, capture in the
quality rating, and reputation attributes captured by quality rankings, country of origin,

coffee variety and quantity available of a coffee lot.

The second paper analyses the economic impact of e-auctions under conditions of
uncertainty by comparing the optimal strategies in a benchmark case solely based on
relationships to a case in which auctions can be utilized, this paper analyzes the economic
impact of auctions under conditions of demand uncertainty. Results indicate that auctions
are especially beneficial for smaller masters and in more uncertain environments.
Moreover, auctions greatly aid in the discovery process of new sources of product.

The third paper proposes a ‘differentiation measure’ for calculating the degree of
differentiation vis-a-vis quality ratings at the procurement level of the specialty coffee.
The empirical analysis is based on cross-entropy analysis from information economics on
653 observations from the Cup of Excellence and 59 observations from the Q auction.
The measure is a proposition to measuring value creation in supply chains.

The dissertation argues that specialty coffee roasters are achieving coordination of supply
chains with scope for simultaneously increasing consumers’ satisfaction and producers’
income. This explicit, high level coordination of the complex coffee supply chain signify

a profound change of the management role and the potential benefits to all participants.

Copyright by
M. LAURA DONNET
2007

ACKNOWLEDGEMENTS
I would like to thank Dr. Weatherspoon for his great job as my dissertation adviser. I
appreciate the care and sense of humor that made working and learning most positive and
enjoyable. Thanks for stimulating and supporting my ideas and for the training in writing,
organizing, communicating and networking.
I am grateful to Allan Schmid for supporting my learning in the broadest sense during
these five years and inspiring me to follow my own path. Thank you for fulfilling the
major professor duties and making me meet all the program requirements. I am indebted
to Dr. Hoehn for encouraging me to do the dissertation on the specialty coffee topic and
providing important analyses on e-auctions which are at the core of this work. Also
thanks for his help on improving my outlines and narratives.
My gratitude to Dr. Jeitschko for kindly committing to working with me to do research
and writing it. I appreciate his confident teachings to develop the specialty procurement
model and the paper; his contribution made this dissertation a certainty and favorably
impacted my thinking.
I would also like to thank the special support from the following professors: Laura
Cheney for an enjoyable and enlightening TA, Dr. Black for his teachings in information
class and during RA, Dr. Crawford for his guidance on the program procedures and
decisions, and Dr. Peterson for leading my learning on agribusiness and presenting our
work at IAMA.
I am grateful to the Department of Agricultural Economics for the assistantship funding

and the Graduate School for the Dissertation Completion Fellowship.

Special thanks to Pamela for her support and encouragement during this time and for
checking regressions and paragraphs so many times. Tasting the specialty coffees would
not have been so amusing without your presence. Specialty thanks to Lourdes for
cheering up this dissertation and her motivations and discussions about coffee. Finally, I

am grateful to my parents, brothers, Tere and Maria for their constant support and care.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES ............................................................................................................ x
INTRODUCTION ........................... ‘ ................................................................................... 1
Background and Research Context ................................................................................. 1
Research Goals, Gap and Relevance .............................................................................. 3
Methodology ................................................................................................................... 5
Contributions of this Dissertation ................................................................................... 7
Organization of the Dissertation ..................................................................................... 9
PRICE DETERMINANTS IN TOP QUALITY E-AUCTIONED SPECIALTY
COF F EES ......................................................................................................................... 10
Introduction ................................................................................................................... 10
Hedonic Approach and Applications to Specialty Food Products ................................ 14
E-Auctions of Specialty Coffee .................................................................................... 19
Specialty Coffee Hedonic Model .................................................................................. 24
Data ........................................................................................................................... 25
Functional Form ........................................................................................................ 27
Explanatory variables ................................................................................................ 29
Interpretation of Results ................................................................................................ 30
Conclusion .................................................................................................................... 34
THE IMPACT OF E-AUCTIONS IN ADJUSTING PROCUREMENT STRATEGIES
FOR SPECIALTY COFFEE ............................................................................................ 37
Introduction ................................................................................................................... 37
Framing and Modeling Procurement Decisions ........................................................... 41
Supply Relationships and E-Auctions in Specialty Coffee Procurement ..................... 44
The Analytical Framework ........................................................................................... 46
The Model ................................................................................................................. 46
Relationship procurement without the availability of E-auctions ............................. 49
Optimal Procurement with the Use of E-auctions .................................................... 50

The optimal use of the e-auction ............................................................................... 51

The use of relationships when e-auctions are available ............................................ 52

The Importance of E-auctions as Part of an Optimal Procurement Strategy ................ 54
The probability of using the auction ......................................................................... 55

The impact of auctions on relationship procurement ................................................ 58

The impact of demand uncertainty on profits with the optimal use of auctions ....... 60
Conclusion .................................................................................................................... 62
PRODUCT DIFFERENTIATION THROUGH QUALITY RATINGS IN SPECIALTY
COFFEE: A CROSS-ENTROPY ANALYSIS OF E-AUCTIONS DATA ..................... 64
Introduction ................................................................................................................... 64
The Specialty Coffee Quality Rating ............................................................................ 67
The Cross-Entropy Measure and Applications ............................................................. 69
Cross-entropy measure with probabilities ................................................................ 69

vii

The Relationship between Measuring Income inequality and Product Differentiation

................................................................................................................................... 71
Quality Differentiation Measure in Specialty Coffee Procurement .............................. 72
Empirical model ........................................................................................................ 72
Data ........................................................................................................................... 74
Decomposition of the Cross-entropy of value and quality rating ............................. 75
Results ....................................................................................................................... 76
Interpretation of the Differentiation Measure and Management Implications ............. 77
Conclusion .................................................................................................................... 82
CONCLUSION ................................................................................................................. 84
APPENDIX ....................................................................................................................... 87
A. The Coffee Supply Chain ......................................................................................... 87
Coffee growing and processing ................................................................................ 88
Trade and Procurement ............................................................................................. 89
Roasting .................................................................................................................... 91
Retailing .................................................................................................................... 92
Consumption ............................................................................................................. 94

B. Definition of Specialty Coffee ................................................................................. 96
In Green Coffee Beans .............................................................................................. 96

In Roasted Coffee ..................................................................................................... 97

In the Coffee Beverage ............................................................................................. 98

C. The Emergence of Specialty Coffee ......................................................................... 99
Changes in Roasting and Retailing ........................................................................... 99
Changes in Consumption ........................................................................................ 100
Results in Product Differentiation .......................................................................... 101

D. Detailed derivations and proofs of propositions for second paper ......................... 102
REFERENCES ............................................................................................................... 108

viii

LIST OF TABLES

Table 1: Example of Coffee and Grower Information Available to Potential Bidders of

the CofE .................................................................................................................... 21
Table 2. Comparison of Coffee Supply Chains. Volume and Prices of Dry Coffee Beans
in 2005 ...................................................................................................................... 24
Table 3: Descriptive Statistics CofE E-Auctions: N=541 ................................................ 27
Table 4: Box-Cox Test of Functional Form for the Specialty Coffee Hedonic Price Model
................................................................................................................................... 28
Table 5: Chow Test of Differences between Linear Specialty Coffee Hedonic Price
Models ....................................................................................................................... 29
Table 6: Specialty Coffee Hedonic Model Estimates and Implicit Prices of Attributes.
Dependent Variable: log (Price) ............................................................................... 31
Table 7: Results of the Quality Differentiation Measure the CofE and Q Specialty Coffee
Auctions .................................................................................................................... 77
Table 8: Degrees of Differentiation by Quality Rating Groups in the Cup of Excellence 78
Table 9: Degrees of Differentiation by Quality Rating Groups in the Q .......................... 8]

ix

INTRODUCTION

“The nature of the international coffee market has changed dramatically over its five centuries of
existence. So has control of that market, which moved in the sixteenth century from the producer
to the exporter, and to national and international government institutions, and finally today to a
few vertically integrated multinational firms. To simply speak of “the market” or “market forces”
is to ignore the evolving nature of that market and the role of human agency in creating it. But the
central fact continues to be that over 90 percent of all coffee is exported by relatively poor tropical
countries and a similar percentage is imported by rich, temperate countries”

(Clarence-Smith and Topik, 2003)

Background and Research Context

This dissertation is about one of the most exciting developments in the food industry in
modern times. New ways of consuming coffee that focus on quality, differentiation and
value-adding attributes have created a specialty coffee market segment that has grown
dramatically since its formative years in the 19805 (Roseberry, 1996; Ponte, 2001).
Specialty coffee is a growing market segment in an otherwise declining industry. The
tremendous growth of this differentiated industry requires the full incorporation and
improvement of producers in a highly demanding and coordinated supply chain.
Coordination is crucial between coffee producers and masters since “specialty coffee
distinguishes itself first and foremost by the quality of the raw material” (Knox and
Huffaker, 1997). The focus of this dissertation is on the procurement strategies that are
unique to specialty coffee, in particular e-auctions and supply relationships.

The specialty food industry has placed coffee together with other differentiated food
products including wines, beers, and cheeses, among others. One major difference with
these food products, however, is that specialty coffee is the result of global supply chains

in which consumers are largely located in the US, Europe and Japan and producers in

poor agriculturally-based developing countries. The consistent supply of premium quality
coffee beans from its origin in coffee growing countries is crucial to sustaining the
specialty coffee chains. Simultaneously, the impact on rural poverty in developing
countries is not well understood. This represents a challenge to traditional producers who
were used to selling to intermediaries in anonymous commodity supply chains. At the
same time, this also represents a unique opportunity to show that the intrinsic quality and
value of specialty coffees are created at the production level in the growing countries.

Specialty coffee entails the need for trade and procurement strategies different from the
mainstream supply chains that deliver regular coffee'. Mainstream coffee supply chains
are characterized by a large number of intermediaries and a system of market relations
that give greater market power to the import side than to the export side resulting in a
proportionally greater share of total export earnings accruing to importing companies
(United Nations, 2003). In the commodity coffee, the grading of coffee beans is the basic
measure of value. The mainstream coffee trade is organized around a grading hierarchy
according to type (Arabica or Robusta), primary processing (wet or dry), country of
origin, and grade (Ponte, 2001). However, beans that are relatively uniform in appearance
would be graded into the same category, the problem is that the beans can vary
significantly in the cup (Bates, 1997). In addition, most defects regarding appearance of

the beans generally have little impact on cup quality (Barker, 2004). Because of the

 

' ‘Mainstream’ —supply chains and markets— is used in the global value chain literature to refer to the
prevailing arrangements by which coffee is produced, traded and consume. ‘Regular’ coffee is the blended

and homogeneous product appealing to a standardized taste of the mass consumer.

inability of buyers to assess quality, commodity traders offer prices that reflect the
expected quality of the crop; the result is a weakening of the incentives to incur the costs
of producing premium quality coffee by devoting above average resources to its
production (Bates, 1997). i

In contrast to the mainstream industry, the assessment of the quality attributes in the cup,
acidity, body and flavor, is of foremost importance in the specialty industry. Specialty
coffee buyers evaluate quality in the cup, transmit quality information to producers and
remunerate them according to quality. This is a breakthrough in the transaction of green
coffee beans and the basic idea of procurement strategies in specialty coffee. There are
two specific systems or procurement strategies unique to specialty coffee: relationships
and e-auctions. Both relationship and e-auction coffees— “strive to achieve a market-
based system that allows the transmission of quality information to the source, and
therefore a payment in relation to quality. They are market-based systems, as opposed to
certification systems, such as the organic and fair trade, because quality “stands on its
own” and is assessed directly by the buyer who remunerates according to this evaluation
(Ponte 2002).

Research Goals, Gap and Relevance

The goal of this dissertation is to examine the procurement approaches that are distinctive
to specialty coffee supply chains and their valuations, incentive and strategic properties
and implications. These procurement approaches are relationships and e-auctions and
their characteristic is that they transmit quality information and remunerate according to
it. In this sense, these procurement approaches represent a breakthrough in coffee trading

and represent a relevant alternative to traditional procurement. This is important because

_ _..‘.—.n..__.- . .-

failure to distinguish and remunerate high value and differentiation at the specialty coffee
production level constrains the capacity of coffee producers to make the necessary
investments to maintain and enhance quality. Consequently there is a risk of the supply
chain collapsing due to an ever declining quality. In addition, innovations at the producer
level are improbable since risk taking is not rewarded. Without continual reinvestment,
coffee producers miss the growth opportunity from the value creation in differentiated

supply chains. All of these imply that in the medium to long run there is a decline in ~
1'
producers’ income and increasing rural poverty. /./

Previous research has addressed the opportunity to increase coffee producers’ income
from the differentiation in specialty coffee from the political economy perspective of

value chain analysisz. Kaplinsky and Fitter (2004) argue that coffee producers’ prices and" .‘ I

‘-
“1 ..__
-._u-

income is increased fi'om the systematic application of knowledge at all stages of they;

- ' ‘1. a
w 'v V.

supply chain to create value and competitive advantage. Lewin et al (2004) discuss“ \ i I
possible ways in which some coffee producers can take advantage of specialty markets to B

increase their income. They point to the difficulties of small producers to access specialty

markets and maintaining and improving quality according to the changing characteristics

7
of specialty coffee markets. Fitter and Kaplinsky (2001) outline the fact that the capacity \j \ ,I

",4

to meet increasingly complex demands from consumers and buyers in coffee does not

necessarily mean that the returns to differentiation accrue to poor producers.

 

2 Global value chain analysis is a framework that highlights the relative value of the activities required to
bring a product or service from conception through the different phases of production, delivery to final
customers and final disposal after use, to analyze underlying factors that affect governance, employment

generation and growth capacity, etc..

Oberhansli (2001) argues that the possibility of producers capturing price differentials
will depend on consumers recognizing valuable characteristics produced at the green
coffee production level. Also, Kaplinsky and Fitter (2004) detail the importance of .- -
erecting barriers to entry at the production level to sustain competitive advantage in the
highly competitive environment of specialty coffee production. These political economy
studies discuss the macro context and governance of supply chains and firms operating in
them but do not go into detail of how to interpret increased information flows, how coffee
producers create value or what other strategic decisions they can make to sustain value

and income. Studies at the micro and case study level, Bacon (2005) discusses the impact '

.
_,..-.-4— -‘

of premium prices paid to organic and fair-trade certified cooperatives on coffee
producers’ livelihoods. The author also discusses the perceived benefits by producers on
this alternative trade networks in relation to the reduction of production costs,
diversification of marketing channels, learning processes and relationship networks.
Methodology

This dissertation examines the developments in specialty coffee from the active point of
view of supply chain participants to create value for the industry through differentiation
(paper 1), procurement management (paper 2) and the making and use of quality
evaluation procedures (paper 3). Our primary analysis is at the roasters’ level because
they are the captains of the specialty coffee supply chains; they link the dynamic
consumers’ market with the technologically, informationally, managerially and
geographically challenged coffee producers. The supply chain depicted in Figure 1 shows

the economic stages and as well as the players that deliver the coffee product to the final

consumers. The focus of this dissertation is on the transaction between coffee producers

and masters in the supply chains that deliver a specialty coffee product.

A Smallholder

......
no."
-----

Parchment coffee

  
 
 
  

 

 

  
 
  
 

   
    
   
   

. . Coo rative
Green coffee Dry Milling Farm: group Producing
Plant agent countries
Procurement Strategies
- Spot market
' - Auction
- Relationship
\- - Integration
Roasted coffee Roaster ............................... Consuming
............. countnes

 

V

   

   
 

Retailer
Supermarket

       

Coffee shop

Brewed coffee Restaurant

 
 

Figure l: The Coffee Supply Chain. Source: Adapted from (Ponte, 2001)

The background information for the dissertation comes from industry literature,
communication with industry participants and academic literature. The industry literature
includes trade journals and magazines, publications of the Specialty Coffee Association
of America (SCAA) and publications and websites of specialty coffee toasters. Most of
the industry contacts that provided information on the specialty coffee chains were the
Cup of Excellence and the Q auctions which were contacted at the SCAA Annual

Meeting in Charlotte, North Carolina, 2006. lnformally talking with industry participants

has been of foremost importance to learning about the most important facts, making sense
of the information and developing a more rounded idea of the industry. Specific
information obtained through this source is appropriately cited throughout the
dissertation. The academic literature used mainly pertains to the field of global value
chains (Ponte, Daviron, Gereffi, Lewis and others).

This dissertation uses theory and methods germane to economics and management. The

_ , ._...-.._- ._1-.1- ——-.- — H n.._..‘
.__r

__.. M2...
first paper is conceptually framed in the determinants of price in specialty food products.
The method is hedonic price analysis applied to a data set of 541 specialty coffees from
the Cup of Excellence. The second paper uses organizational economics to frame the
choice of procurement strategies based on supply relationships and e-auctions. The
theoretical modeling relies on a newsvendor approach adapted to the profit maximization
function of specialty coffee roasters. The third paper relates to the strategic management
concepts of differentiation and economic concept of quality evaluation to proposes a
‘differentiation measure’ for calculating the degree of differentiation vis-a-vis quality
ratings at the procurement level of the specialty coffee. The empirical analysis is based
on cross-entropy analysis from information economics. The data contains 653
observations from the Cup of Excellence and 59 observations from the Q auction.
Contributions of this Dissertation

This dissertation offers an analysis of specialty coffee procurement from the point of
view of economics and management. In this sense it offers an additional point of view to
previous analysis for example in the fields of sociology and anthropology (see for
example, Roseberry, 1996; Mathieu, 1999; Samper, 2003) and political economy (see for

example Kaplinsky and Fitter, 2004; Daviron and Ponte, 2005). The dissertation provides

conceptual and analytical frameworks and evidence to understand specialty coffee in the
context of specialty food products, procurement strategies and value adding in the supply
chain.

The novelty of the topic in the economics and management disciplines has the advantage
that one can utilize tools developed in other areas and products to our case. This is the
case of the hedonic analysis that we apply to specialty coffee for the first time in paper 1.
The contribution of the hedonic study of specialty coffee prices is that it provides
evidence to support that the price of specialty coffees is not only related to the product
material quality (sensory attributes) but also to the product symbolic quality (reputation
attributes). The study complements the analyses of other specialty food products,
especially wine. The idea of marginal prices of attributes is important for supply chain
participants to make better decisions regarding marketing possibilities and resource
allocation.

The contribution of this dissertation is however not limited to applying developed tools to
new products and data set. A unique contribution of this dissertation is the ‘original
adaptation’ of a modeling and an econometric approach in papers 2 and 3 respectively.
The first is an adaptation of the classical newsvendor approach to the analysis of
procurement strategies in specialty coffee. The original contribution is that it proposes a
theoretical model of evaluation of the trade-offs and complementarities among
procurement strategies in specialty coffee. Another interesting feature of the study is that
it covers a common ground for strategic and management operations with the broader

economic organizational forms. The outlook is that similar and additional problems can

be addressed in an analogous way. An application is that the theoretical model can be put
to work empirically with firm data and provide direct assistance to management.

The originality of the ‘differentiation measure’ is the adaptation of an approach from the
field of information economics and income inequality to measures product differentiation
for which there are no previous measurements as such. The differentiation measure
provides an addition to the field of price analysis, e.g. different from the hedonic analysis.
The measure is a proposition to measuring value creation in supply chains. It could be
further applied for other food products and provide for comparative analysis between
different marketing channels within an industry and between industries.

Organization of the Dissertation

The dissertation consists of an introduction, three independent papers and a conclusion.
The introduction provides the motivation, questions, the methodology and contributions
of this research. The central body includes the three papers that are intended to stand
alone individually. Research extensions and conclusions offer a final thought on this
work. The appendix contains background material on: A. the coffee supply chain, B. the
emergence, and C. The definition of specialty coffee. Appendix D shows detailed

derivation of the specialty coffee procurement model.

PRICE DETERMINANTS IN TOP QUALITY E-AUCTIONED SPECIALTY
COFFEES

Introduction

The specialty coffee industry experienced an explosive growth in the 19903 driven by the
masters and retailers’ investments in product innovation and market segmentation, hence
increasing value added in the coffee supply chain (Ponte, 2001; Lee, 2006). The US
specialty coffee market has grown from $1 billion in 1990 to $11 billion in 2006,
according to the Specialty Coffee Association of America (SCAA), the National Coffee
Association (NCA) and Mintel group industry reports (Hillside Agricultural Program,
2002; California Libraries Association Conference, 2005; Lee, 2006). The only
operational definition for specialty coffee is, those “made from exceptional beans grown
only in ideal coffee-producing climates; they tend to feature distinctive flavors, which are
shaped by the unique characteristics of the soil that produces them” (SCAA, 2007)3.
Traditionally, expert tasters, usually the buyers and/or roastmasters of coffee firms,
define the quality that determines whether a coffee fits into the specialty category or not.4
As an industry, there is an attempt to encode information about these subjective gustation

experiences into ‘certified’ quality systems (Ponte and Gibbon, 2005), such as the coffee

 

3 This definition emphasizes that specialty coffee stands for an outstanding cup quality and does not include
classifications which refer to ‘immaterial’ or ‘credence’ attributes such as organic, fair-trade, and bird-
friendly. This definition for specialty coffee is the one being increasingly accepted in the industry.

4 “Like an excellent chef or wine maker, such people typically apprentice for many years with more
experienced tasters, gradually developing a taste memory and sense of the range of qualities that can be
expected by tasting hundreds of samples from each major origin country over many crop cycles, at many

degrees of roast, singly and in combination with other coffees” (Knox, 2006).

10

standards by the SCAA, the Cup of Excellence cupping form, and the quality rating in the
Coffee Review. The prices paid according to this quality information by specialty coffee
buyers to growers are the crucial element of supply chain coordination in these market-
based supply chains (Ponte, 2002).

Coffee is a product that lends itself to differentiation. There is a wide diversity of flavors
and aromas that emerge from different coffee growing soils and climates, tree varieties,
and cultivation and processing methods, all of which impact quality. The natural variation
of coffee places it in a widening spectrum of differentiated food products which include
wines, beers, and cheeses among others (Roseberry, 1996). Measuring quality of
specialty food products is complex as they need to be ‘looked and felt’ (de Figueiredo,
2000). Complexity arises from the variability of the sensory experience from person to
person (quality is subjective) and from product to product (specialty products are
different from one another). The practice to assess coffee quality in the specialty industry
is the cupping. Analogous to wine tasting, cupping is the sensing of aromas, flavors and
body through olfaction, gustation and mouthfeel, respectively (Lingle, 2001). The
specialty coffee industry has operationalized quality assessment through the convention
of a cupping form that rates coffees with 100 point scale.

Price discovery is even more complex given that quality is subjectively determined.
Prices of specialty food products depend on both sensory as well as reputation attributes
(de Figueiredo, 2000). Complexity in appraising sensory quality characterizes a situation
of imperfect information in which reputations provide extra information to supply chain
participants to ascertain the value of the product (Nelson, 1970; Kreps and Wilson, 1982;

Shapiro, 1983; de Figueiredo, 2000; Carriquiry and Babcock, 2004). Reputation quality is

11

the recognition awarded to the names of specific farms, origins and conditions and
methods of production known to consistently produce coffee with desirable attributes
(Samper, 2003). Similar to the concept of terroir in wine production, high quality coffees
can be produced only in certain places. There are origins that have proven their ability to
grow great coffees due to altitude, latitude, soil and other conditions while other origins
have not demonstrated such ability (Holly, 2005). For example, Colombian (as well as
Guatemalan, Costa Rican, and Kenyan) coffees earned longstanding reputations for
producing high quality coffee (Samper, 2003).5 Moreover, similar to grapes, the genetic
strain of the coffee plant and its unique adaptation to the environment profoundly
influence the character of the coffee cup (Arvidson, 2003; Davids, 2004). Specialty
coffees are usually obtained from several varieties of Coffea arabica: the traditional
cultivars of Typica and Bourbon, the more recent Catuai variety and the heirloom
Pacamara are outstanding producers of specialty coffees”. Growing methods also
impact quality, fertilization, harvest and processing practices affect the characteristics in
the coffee beans (Samper, 2003). Therefore, origin and the individual conditions of
altitude, soils and climate, the variety of the coffee plant and the cultivation methods
constitute potential reputation attributes. This paper examines the market values of

specialty coffee attributes, especially those associated with the sensory evaluations of

 

5 The international classification system has remunerated this reputation by separating Colombian Mild
Arabicas are distinguished from Other Mild Arabicas.

5 Presentation at the SCAA 17‘’1 Annual Conference and Exhibition, panel ‘Exploring Distinctive
Characteristics and Virtues of Coffee Varietals: the Bourbon and Pacamara case.’

7 Holly (2005) Specialty Coffee Definition.

12

specialty coffee buyers and production reputations of coffees traded in top quality e-
auctions. Although the volume traded at e-auctions is still very small, the coordination
based on the remuneration of quality attributes is of great importance for developing the
top quality segment of the industry. Moreover, the value associated with distinguishing
among coffee groups becomes important for the whole industry as it continues to move
from a commodity to a differentiated beverage.

We hypothesize that specialty coffee prices are influenced by both sensory and reputation
attributes since specialty coffee is a highly differentiated specialty food product. We
expect quality ratings from the sensory evaluation of coffees in the cupping competitions
to have a significant impact on prices, particularly, since the specialty coffee market is in
the early stage of its life cycle, in which the taste experts at the cupping competitions
have a crucial role in defining sensory quality. Despite the exhaustive procedure for
measuring quality developed by the specialty industry, our conjecture is that prices are
not a function of this measure alone but that preferences supported by reputations are
likely to be observed in their bidding behavior. This observation was made by Boccaletti
(1999) for food products with designation of origin, where he argued that for reputations
to work as a quality signal, buyers should have developed skills in inferring the -
sensory— attributes associated with these reputations. Moreover, since specialty coffee
buyers are interested in passing quality information on to the consumer, they are likely to
value the communication capacity of reputations for use in their marketing strategies.
Understanding what determines the price of specialty coffees is crucial to developing a
sustainable supply base for the specialty coffee industry and to pass the benefits of

growth on to the growers. On the one hand, what buyers pay for the beans informs

13

growers about the consumers’ quality preferences so they can direct production practices
towards obtaining the desired attributes. On the other hand, the transmission of price to
consumers informs them about the resources involved in producing the attributes they
desire. Contrary to the mainstream coffee markets, in which roasters disclose as little
information on quality as possible to growers and consumers, the original configuration
of the specialty coffee industry was based on transparency, where consumers and masters
engaged in a mutual definition and redefinition of quality and remuneration accordingly
(Ponte and Gibbon, 2005; Daviron and Ponte, 2005). Thus, specialty supply chains
(Lewin et al., 2004) are seen as a possible way to increase the income of some coffee
growers. The possibility of growers capturing the price differentials depends on
consumers recognizing the link between desired traits and their underlying determinants
at the production level (Institute of Development Studies, 2001).

The paper proceeds with a review of the literature on hedonic price models and their
application to specialty foods such as wines. Next, we describe the specialty coffee e-
auctions. Then we present our empirical specialty coffee hedonic price model, followed
by the presentation and discussion of the results. In the final section we recap what has
been learned and address limitations and further research.

Hedonic Approach and Applications to Specialty Food Products

An approach to studying the relationship between prices and quality attributes is the
hedonic analysis of prices. The analysis consists of modeling the price of individual
products as a function of various attributes that can be encountered in them in order to
estimate the implicit or marginal prices of such attributes. If the estimated implicit price

is not significantly different from zero, then the attribute is not valued by consumers.

l4

Possibly, the earliest reference to this approach is Waugh (1928). Waugh analyzed
individual prices of asparagus, tomatoes and cucumbers using their physical attributes
such as color, size of stalks, shapes, maturity, uniformity, and similar factors in order to
determine consumers’ marginal valuations of these attributes. The quality attributes
analyzed by Waugh were all sensory. He viewed this type of analysis as of great practical
value for growers to adjust production to the market demand.

Later, the approach advanced from developments in demand theory for quality attributes
and the statistical methods applied to the hedonic regression. In 1966, Lancaster
presented a theory of consumer utility based on preferences with respect to characteristics
rather than with respect to goods. This approach allowed goods to be viewed as a bundle
of characteristics desired by consumers. In 1974, Rosen presented a model of product
differentiation based on the hedonic prices of attributes empirically estimated as the
relationship between the observed prices of products and the specific amounts of
individual attributes associated with them. Hedonic price functions relate to both the
underlying supply costs and the underlying demand preferences for individual attributes
and hence they do not identify the structure of either (ibid.). Hence, a significant and
large implicit price may not reflect consumers’ high valuation of an attribute, but rather a
high cost incurred in producing that attribute. However, if the supplied quantity was
exogenously determined, the regression of price on quantity may be used to estimate the
demand elasticity (Schultz, 1938).

Great advances have been made in the application of hedonic price functions to specialty
products such as en the case of wine. Hedonic price analysis has been extensively applied

to determine the attributes that significantly influence prices in wine. In an early study,

15

Oczkowski (1994) argued that it is important to understand implicit prices of attributes
because of their marketing implications, they help: producers to evaluate the
implementation of practices leading to attaining a particular attribute; and consumers to
evaluate their wine purchases relative to average values of attributes made available
through wine guides or newspapers (ibid.). The author further illustrates with an
Australian wine example, he found that quality rating, cellaring potential, grape variety,
region, vintage, and producer size determined price.

In subsequent papers, authors generally distinguished among sensory, chemical,
objective, and climatic attributes. Sensory attributes were subjectively assessed measures
like ‘aromatic intensity’, ‘aromatic complexity’ and ‘finish’; chemical attributes were
technical measures, such as sugar and acid levels; objective attributes were those that
appeared on the bottle label and easily recognizable by the consumer, including grape
variety, origin, vintage year, rankings, and climatic attributes; and climatic were those
which measure the effects of weather on the grapes used in wine production (Oczkowski,
2001). Most studies indicated that objective attributes, which consumers can identify far
easier than sensory attributes, were the most influential. The general finding was that
sensory attributes influence the quality assessments of experts but because they were less
evident to consumers, had little to no influence on price (Troncoso and Aguirre, 2006).
Combris et al. (1997) found that Bordeaux wine prices were largely determined by the
variables appearing on the bottle label, including color, vintage, and appellation of origin,
but not by sensory variables. The authors provided an explanation in terms of imperfect
information; consumers can more easily assess objective attributes compared to sensory

attributes, hence their purchasing decision was influenced by the former. Lecocq and

16

Visser (2003) also found for Bordeaux and Burgundy wines that the objective attributes
appearing on the bottle label, ranking, vintage and appellation, were largely responsible
for determining wine prices while sensory attributes had a small influence.

Some studies found quality ranking to have the most influence in determining price.
Landon and Smith (1997) highlighted the important role the wine classification system
(cru, grand cm, and premier cru) played in Bordeaux wines. Combris et al. (2000) also
found that the most influential attributes for Burgundy prices were the ranking of grand
cru, premier cru, communale and regionale, and the vintage (year of bottling).

Some studies concentrate on the impact of wine experts’ reviews on the prices paid by
consumers. Wine reviewers rate wines using a 100 point scale and describe their
gustatory experiences which they make available to consumers in wine guides and
magazines as well as in non-specialized publications. Schamel and Anderson (2003)
found that the magazine for premium wines from Australia and New Zealand, “James
Halliday”, was highly significant on determining wine prices. Ali et al., (2005) analyzed
the impact of the renowned wine critic Robert Parker on Bordeaux primeur wine prices,
i.e. the prices determined by the winery owners when the wines are still very young. The
authors compared Bordeaux wines which prices have been established with and without
knowing the Parker grade. They found that, on average, the impact of the expert
valuation was 2.80 euros per bottle of wine. Specialty coffee buyers, unlike consumers,
are themselves coffee experts who use their skill to offer coffees with a unique taste
(Knox, 2006). Thus, the quality rating in hedonic wine studies measures the impact of
experts’ opinion on prices whereas in this coffee study, the quality rating is the variable

that measures the coffee sensory quality.

17

Combris et al. (2000) provide two types of arguments to explain why quality grading by
experts was not an irrefutable measure of quality. The first view was that tastes converge
to a consistent quality measure as long as there were a sufficiently large number of juries,
in which case, the grading would also be an ‘objective’ quality indicator. Quality
measures that do not reach an irrefutable measure of quality do so because they do not
have enough graders. In contrary, the second view states that tastes were irreducible, so
grades represent a compromise between different opinions which are not representative
of the actual subsets of preferences and hence their weakness. However, this does not
mean that there was a conflict between the two sets of variables, sensory (those
describing the taste experience) and objective (those appearing on the bottle). Objective
attributes boost prices by highlighting some of the effects of sensory variables (Combris
et al., 2000). Lecocq and Visser (2003) argued that the debate on how much weight
should be placed on the importance of sensory versus reputation results in a ‘puzzle’ for
pricing specialty food products.

With respect to the origin, Landon and Smith (1997) also highlighted the influence of the
origin, which they considered a collective reputation variable. The authors argued that
consumers form their predictions of the quality of an individual firm by using
information on the quality of similar firms. Troncoso and Aguirre (2006) found that
valley of origin was the most important price determinants of Chilean wines in the US
market, along with grape variety. From those results, they stated that the most important
management decision in wine production was the choice of alternative vineyard locations

and varieties at the moment of planting.

l8

In this paper, we draw on the previous wine studies to identify and construct explanatory
variables for coffee prices. From the wine studies we have the concept that specialty
coffee prices may be influenced by attributes analogous to wine experts’ quality ratings
and rankings, regions, and wine varieties. We also draw on the wine hedonic studies to
interpret the implicit prices of attributes and infer implications on the supply chain
participants. Although the cases of wine and specialty coffee are analogous in many
ways, there are some differences that need to be taken into account. The most important
difference is that while wine studies use retail price data and information on attributes
from wine reviews and consumers’ guides, the data set for this paper is from a
competition-auction at the processing (roasting) level in the supply chain. We perform
our analysis at the processing level because roasters are leading the definition of quality
and the education about specialty coffee attributes. As specialty roasters preserve the
identity of the grower in their marketing strategies and consumers become more educated
about the origin characteristics, we can expect to be able to analyze the preferences for
attributes at the consumers’ level.

E-Auctions of Specialty Coffee

E-auctions are a quality competition-auction process in which the highest quality coffees
are selected at a national level and then sold through intemet auctions (Ponte, 2002). In
this paper we analyze data from the Cup of Excellence specialty coffee competition and
auction. The CofE ran the first specialty coffee auction in 1999 as part of the Gourmet
Project (funded by the lntemational Coffee Organization, the lntemational Trade Centre
and the UN Common Fund for Commodities) and continues to be organized by the

Alliance for Coffee Excellence (ACE). Since then, trade has been increasing at the CofE

19

and the prices formed at the auctions have significantly influenced the structure of the
specialty coffee market (United Nations, 2003). As the appreciation of and remuneration
for quality increases, not only at e-auctions but also in private trade and supply
relationships, the relevance of this type of auctions as a price discovery and reference
system for the specialty coffee market becomes more important.8

The CofE schedule is organized to following the coffee harvest in the participating
countries (Figure 2). This corresponds to once a year in the case of Bolivia, Brazil, El
Salvador, Honduras and Nicaragua, and twice a year in the case of Colombia. Coffee
growers submit samples of their coffee lots that are prepared and codified for
participating in the cupping competition. The competition consists of three stages: a pre-
screening stage to pre-select the best of all entries and two cupping stages where national
and international jurors successively evaluate the pre-selected coffees using the CofE
cupping form. The categories included on the form are: absence of defects, cleanness of
cup, sweetness, acidity, mouthfeel, flavor, aftertaste and balance (Howell, 2005). The
assessment on each of the categories is added to yield a final score or quality rating. The

juries’ individual assessments are averaged to give one final quality rating for each

 

8 Auctions have the role of providing a way for market participants to obtain approximately the market

value for the items being traded (Ashenfelter, 1989).

20

 

SAMPLES SUBMISSION
- Growers’ participation is free
- Minimum size of 15 bags per lot
- Preparation of samples for cupping competitio

l

PRES-SCREENING
- Jurors select 150 coffees through visual inspection and cup attributes
appraisal
- Selected farmers move their lot to a warehouse
- Auditing firm samples each lot and prepares samples for cupping

l

NATIONAL JURY
- Host country cupper
2-3 days - First round of cupping: select 80 coffees
- Second round: select around 60 coffees scoring 80 and above

1

INTERNATIONAL JURY
- Day 1:jury calibrates coffees to a score of 80 points
5 days - Days 2 and 3: jurors cup all coffees coming from the national
- Day 4: second round with top coffees (20-45 depending on the auction)
- Day 5: the top ten scoring coffees are cupped and ranked

1

PRE-AUCT ION
- Bidders registration for getting samples and participating at the
auction
- Exporters send sample to registered bidders
- Award winners’ ceremony

AUCTION
- Buyers bids on-line
- Winning bidder and price defined '

l

POST-AUCT ION
- Buyer sends shipping instruction
- Exporter invoices and collects
- Exporter pays producer
- Exporter pays organizer cost recove I

30 days

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 day

 

 

 

 

 

 

 

Figure 2: CofE Auction Schedule. Source: own elaboration.

coffee. The coffees are ranked according to the quality rating and the final winners are

awarded the Cup of Excellence. Immediately following the competition, all information

21

is gathered on the coffees and growers and is made available to the potential buyers
(Table 1). In addition, samples of the coffee lots are shipped to potential buyers for them
to cup personally. During the auction, the lot numbers are displayed on the auction
website. The bidding is ascending and bids are shown to all participants. Buyers can see
their bids on the screen or can check later via e-mail how their bids compare with those
of other bidders (United Nations, 2003). After the winning bid is determined, the auction

manager facilitates the arrangements for payment and delivery.

Table 1: Example of Coffee and Grower Information Available to Potential Bidders

 

of the CofE
Coffee Characteristics Property Characteristics
Agricabv BV Calama
Variety: Typica Farm: Marka
Processing Pulped, naturally fermented,
System: mechanically washed, sun drying Farmer: Juan de Dios Blanco
Lot Size: 16 bags Rank: 1
lntemational Jury
Score: 93.52 City: Calama
Orange honeysuckle fiagrance, Regronz. Calama
lemony tang, chocolate orange, Country. Bolrvra
improved as it cooled, deep dark Farm Size: “-00 Hectares
chocolate, sweet floral and cherry Coffee
blossom, complex, rich and growmg
delicate at the same time, sweet area: 4-50 Hectares
citric, very clean, great balance, Altrtude: 1200 masl
Jury Descriptions: caramel, butterscotch flavor Certification: None

 

Source: 2005 Bolivia Cup of Excellence® Winning Farms - Auction December 1, 2005

The CofE provides a unique data source on —sensory— quality rewarding coffee trade. The
information on specialty coffee prices would not exist without these public specialty
coffee auctions. The majority of specialty coffee roasters market coffees of the highest

quality by procuring beans from the Cup of Excellence and offer CofE coffees in their

22

assortment. The CofE is where leading specialty coffee firms discover new qualities and
new values of specialty coffees thus increasing product and price differentiation.
Differentiation is possible because the quality information is disclosed to many
prospective buyers who show their valuations thus increasing transparency not only for
other buyers but also for growers and consumers. The price discovery at these auctions
can be a benchmark for private transactions outside the auction system, which is the case
of coffees traded within direct supply relationships”.

The annual trade volume at the CofE is very small relative to the volume traded in the
commodity and other supply chains including Starbucks and the organic and fair-trade
certifications (Table 2). However, comparing the CofE to any of these supply chains is
misleading since they are essentially different products, which can be readily seen from
the price differences with respect to the other supply chains. The coffees traded at the
CofE are representative of the ‘truly’ specialty coffee market in the sense of the SCAA
definition and the results apply to the highest quality segments of specialty. As the
broader coffee market moves towards differentiation and appreciation of individual
attributes, we deduce that our study has relevance for future research on price

coordination in the coffee supply chain.

 

9 Direct relationships are another trade system based on coffee quality in which roasters build long-term
relations with growers, directly transmitting quality and price information; many relationships are formed

following the auction between participant masters and growers.

23

Table 2: Comparison of Coffee Supply Chains: Volume and Prices of Dry Coffee

 

 

Beans in 2005
Supply chain Coffee Volume (bags) Price (S/Ib)
Commodity coffee' Brazrllan and Other 27 million 1.0229
Naturals Group
Colombian Mild . .
Arabicas Group 11 million 1.1573
0"” M‘” A’ab'cas 20 million 1.1486
Group
Certifications Fair trade 2.2 million2 1.263
Rain Forest Alliance 0.71 million2 no minimum
Utz Kapeh 1.5 million2 no minimum
Organic 1.5 million no minimum
Starbucks CAFE Practices 581,818" 1.285
Internet auctions Q auction 9,0836 1.426
Cup of Excellence 4,0606 4.1286

 

lMonthly averages of ICC Indicator prices for Other Mild Arabicas Group.

2Canes-Chase, Judith. Presentation at the SCAA 17"1 Annual Conference and Exhibition, panel
“State of the Industry Report: Global Supply and Demand Outlook”.

3Fair Trade minimum price. Fair trade website. Coffee. http://www.fairtrade.net/coffee.html
4Starbucks 2005 annual report.

5Starbucks press release.

°Average 2005 auctions: Q (Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua) and
Cup of Excellence (Bolivia, Brazil, Colombia, El Salvador, Honduras, Nicaragua)

Specialty Coffee Hedonic Model

Following the hedonic approach, we propose that the price paid by buyers is a function of
the information that is available to them when submitting their bids. Thus, the price of the
i-th specialty coffee is a function of the value the buyer attaches to its attributes Z1,-
(j=1,...,m), i.e. both sensory and reputation attributes according to our hypothesis. Thus,

the hedonic regression for specialty coffee has the following form:

24

ffPi) = [30+ zjf(Zij) [3:] + zjf(Ziw) Biw+ 5i
where i=1,...,n are the observed specialty coffees, P,- is the price of the i-th coffee, Z,-

are the j-th variables representing sensory attributes and Zw are the w-th variables
representing reputation attributes and 8,- is an independently distributed error term with

mean 0 and variance 02. The functional form will be tested not imposed, hence f (P;) is
either the identity function f (P,) E P,- or the natural logarithm function f (P;) E In P;.

Similarly, f (Zij) and f (Ziw) are the identity function, the logarithm function or a dummy

variable which takes on the value 1 if the characteristic j is present for the i-th
observation or 0 otherwise. We use a Box-Cox test for guidance to determine which
variables need to be transformed for appropriateness of the functional form. We test for

joint significance of dummy variables and then estimate the hedonic regression using
ordinary least squares (OLS) robust estimation. The estimated [3]- are the implicit prices

for the attributes.
Data

Our data set includes 541 observations from 21 CofE e-auctions. The purchase prices for
specialty coffee range from $1.20 to $49.75 with an average of approximately $4 per
pound. The sensory and reputation explanatory variables are the information on the
coffee lots and the production characteristics available to potential bidders from the CofE
website which includes one sensory variable, 1) the quality rating in a 100 point scale; the
following reputation variables: 2) quality ranking, a dummy variable indicating if the
coffee came in the first, second, third, fourth or lower than forth place in the competition;

3) country of origin, a dummy variable indicating whether the coffee is from Bolivia,

25

Brazil, Colombia, El Salvador, Honduras or Nicaragua; 4) coffee tree variety, a dummy
variable indicating Bourbon, Catuai, Caturra, Pacamara, Typica or Other variety; 5)
available quantity of coffee, the size of the coffee lot in bags; and two macro variables: 6)
competition year, a dummy variable indicating 2003, 2004, 2005 and 2006; and 7) the
price of commodity coffee, the monthly averages of the lntemational Coffee
Organization (ICO) for the corresponding auction month and coffee groups"). All of these
variables are exogenous and available to buyers prior to the auction bidding.

Table 3 shows the summary statistics of the variables in the specialty coffee hedonic
model. The quality rating ranges from 80 to 96 with an average of 86.70. The reputation
variables include the quality ranking, the country of origin, and the tree variety. Since
buyers get to cup the coffee first hand from the samples submitted by the auction and are
taste experts, we take the quality rating given by the competition panel of juries as a
proxy for the buyer’s individual assessment of the coffee sensory quality. The quality
ranking is constructed by ranking the coffees from first to last according to their quality
rating and it captures the relative placing of a coffee within the competition. Thus, the
ranking captures the impact of the assessment of the panel of experts on the coffees.
Country of origin and variety serve to characterize the market status of coffees in terms
of a relative differentiation within the specialty coffee sector much in the same way as in
the wine industry, e.g. Bordeaux versus Burgundy or Merlot versus Cabernet. The size of
the coffee lot is the number of bags that are sold together, indicating the total availability

of a particular coffee. Year indicates to which harvest the coffee belongs. Commodity

 

'0 Brazilian Natural Arabicas for Brazil, Colombian Mild Arabicas for Colombia, and Other Mild Arabicas

for Bolivia, El Salvador, Honduras and Nicaragua.

26

prices provide a benchmark from which to obtain a specialty premium relative to

commodity coffee.

Table 3: Descriptive Statistics CofE E-Auctions: N=541

 

 

Mean]
Variable Frequency Std. Dev. Min Max
Dependent Variable
Price $/lb 3.97 3.15 1.2 49.8
Sensory Variable
Quality Rating score 86.70 2.80 80.3 95.9
Reputation Variables
Quality Rankings
First 0.04 0.18 0 1
Second 0.04 0.18 0 1
Third 0.04 0.18 0 1
Fourth 0.04 0.18 0 1
Lower 0.86 0.35 0 1
Countries
Bolivia 0.06 0.24 0 1
Brazil 0.19 0.40 0 1
Colombia 0.15 0.36 0 1
El Salvador 0.20 0.40 0 1
Honduras 0.17 0.3 8 0 1
Nicaragua 0.23 0.42 0 l
Varieties
Bourbon 0.25 0.43 0 1
Catuai 0.19 0.39 0 1
Caturra 0.38 0.49 0 1
Typica 0.08 0.27 0 l
Pacamara 0.06 0.23 0 1
Quantity Bags 21.25 10.98 9.0 122
Macro Variables
Commodity Price $/lb 0.94 0.24 0.6 1.4
Harvest Years
2003 0.21 0.40 0 l
2004 0.23 0.42 0 1
2005 0.37 0.48 0 1
2006 0.19 0.39 0 1

 

Functional Form
Since the results of the hedonic approach often depend critically on the functional form,

we follow Rodriguez (2002). We first use the Box-Cox procedure to give guidance on

27

which functional form describes the hedonic relationship between price and the
continuous explanatory variables more appropriately. Then, we estimate the functional
forms suggested by the Box-Cox results with ordinary least squares (OLS). Lastly, we
select the most appropriate functional form according to the performance of the
estimation. From the Box-Cox test results we reject the linear form and find no evidence
to reject the log-linear and log-log specifications (Table 4). Since both functional forms
yield very similar estimates. We are selecting among the acceptable functional forms. We
follow Halvorson and Pollakowski’s recommendationll and choose the log-linear form
for quality rating and the log-log form for quantity and commodity price for more

meaningful interpretation of their units.

Table 4: Box-Cox Test of Functional Form for the Specialty Coffee Hedonic Price

 

 

Model
Box-Cox testl Functional form
Variables to apply functional form Identity2 Log}—
707 .83 1.08
Left Hand Side Only (0.000) ( 0.298)
Both Left and Right Hand Side (same coefficient)
707.52 1.15
Quality rating (0.000) ( 0.284)
703.31 0.98
Commodity price (0.000) ( 0.322)
735.41 1.85
Lot size (0.000) ( 0.174)
731.03 1.79
Commodity price and lot size (0.000) (0.181)

 

1X2 test statistic of the lambda coefficient in the functions flP)=P7’ and flZ)=Zk

p—values in parentheses, N=541
Ho: lambda = I, or no transformation

3 Ho: lambda = 0, or log transformation

 

” To resort to theoretical grounds if possible or to convenience in dealing with the problem at hand.

28

Explanatory variables
We follow a Chow procedure to test the joint significance of the groups of explanatory
variables: quality rankings, harvest years, countries of origin, and varieties. From the test
results we find that all groups are jointly significant at the 1% level of statistical
significance (Table 5). Thus, we are able to confirm that the proposed model
specification, containing the groups of variables quality rankings, countries of origin,
coffee tree varieties, and harvest year is appropriate. The reference variables for the
categorical variables are: lower than Fourth, Brazil, Bourbon, and 2003 for ranking,

country of origin, tree variety and harvest year respectively.

Table 5: Chow Test of Differences between Linear Specialty Coffee Hedonic Price

Models

 

Group of variables tested R-squared Chow testl
Unrestricted model 0,66

Restricted models

0.59 26.76
Quality rankings (0.000)
0.63 15.29
Years ( 0.000)
0.60 18.35
Countries ( 0.000)
0.65 3.06
Varieties (0.001)

 

1F test statistic
p-values in parentheses
N=54l

29

Interpretation of Results

Table 6 shows the robust OLS estimation results of the specialty coffee hedonic price
model and the implicit price of the significant variables”. The sensory variable quality
rating; the reputation variables: First,» Second and Third quality ranking, El Salvador,
Honduras and Nicaragua origins and quantity, and the macro variables commodity price
and year 2005 are significant at the 1% level. The reputation variables Fourth ranking,
Colombia origin and Pacamara variety are significant at the 10% level.

The quality rating is highly significant indicating the importance of experts’ evaluation in
determining coffee prices. This is not surprising given the structure of the e-auction; the
bidding is largely based on the previous evaluation in the cupping competition. In
addition, the jury members are usually buyers and thus the willingness to pay expressed
at the auction coincides with the preferences expressed in the quality rating. The price
premium for a one point increase in the 100-point quality rating scale is 7.7%; i.e., a 5
point increase in the quality rating translates into almost 40% increase in the price. Given
that quality ratings in the CofE supply chain range from 80 and 95, this represents a fairly
high price differentiation range with respect to sensory quality. Whether this 7.7% impact

per percentage point provides a great incentive for the producers to adopt quality

 

’2 In the log-linear function, the coefficient of continuous variables, multiplied by 100, can be correctly
interpreted as the percentage variation of the dependent variable in relation to a unitary change in the
variable in question. Correspondingly, the percentage impact of the dummy variables can be calculated as:
exp (Bj - 0.5 s.e. fij) -1, multiplied by 100, being s.e. the standard error of the coefficient Bj. We use the F

test to test for joint significance of dummy variables’ groups.

30

Table 6: Specialty Coffee Hedonic Model Estimates and Implicit Prices of

Attributes. Dependent Variable: log (Price)

 

 

Coefficient S.E. Significance Implicit price
£41)
Sensory Quality rating (%) 0.077 (0009) * ** 7.7
Reputation Quality rankingsl
First 0.877 (0.104) **"‘ 153.2
Second 0.318 (0.104) * "‘ "' 44.8
Third 0.312 (0.078) *** 42.0
Fourth 0.139 (0.076) * 19.4
Country of origin2
Bolivia -0. 148 (0.096)
Colombia -0.145 (0.079) * - 10.0
El Salvador -0.191 (0.055) "* ~15.1
Honduras -0.448 (0.056) * "‘ "' -34.3
Nicaragua -0.262 (0.063) * "‘ * -20.6
Tree variety3
Catuai -0.056 -0.054
Caturra 0.049 -0.069
Typica -0.002 -0.075
Pacamara 0.158 (0.095) * 22.8
Other 0.002
Log Quantity (%) -0.375 (0.039) *** -0.375
Macro Year4
2004 -0.084 (0.066)
2005 -0.326 (0.107) *"' * -23 .9
2006 -0.046 (0.093)
Log Commodity price (%) 0.589 (0.153) * ** 0.589
Constant -3.982 (0.741) * **
Observations 54 l
R-squared 0.67

 

Robust standard errors in parentheses

* significant at 10%; " significant at 5%; *** significant at 1%

l Dummy variables indicating if ranking 1, 2, 3 or 4, respectively. Base group is rank between 5 and 43.

2 Dummy variable indicating country of coffee origin. Base group is Brazil.

3 Dummy variable indicating coffee tree variety. Base group is Bourbon.

4 Dummy variable indicating auction year. Base group is 2003.
improvement practices depends on the production effort involved in each particular case,
i.e. the individual grower’s cost function.
The quality rankings are the variables that have the greatest impact on specialty coffee

prices. The price premium for winning the first prize in the cupping competition is 153 %

31

relative to the places lower than fourth. The second, third and fourth places also have a
highly significant impact of 45%, 42% and 20% respectively. These implicit prices
indicate that competition winning coffees are highly valuable for specialty buyers.
Winning a competition is a quality signal that can be easily conveyed to consumers.
Indeed, CofE roasters are using the CofE winners’ logo on their coffees and educating
consumers about cupping competitions and outstanding coffees. This is an extra value in
addition to the value that the coffee already has from a high quality rating. This indicates
that specialty coffee rankings have an important marketing value throughout the supply
chain and that roasters are eager to purchase and capitalize on the quality competitions in
general and on the first, second, third and even forth places in particular. Moreover, note
that these premiums are above and beyond the premium paid for just being qualified as a
specialty coffee versus commodity coffee, indicating that the differentiation within
specialty coffees is likely to be meaningful for consumers similar to what rankings are in
wines -grand cru, premier cm and cm. The market’s emphasis on relative performance in
which relatively small or no differences between absolute quality result in substantial
differences in incomes for the first place winners allows this market to be characterized
as a ‘winner-take-all’ market (Rosen, 1981; Frank and Cook, 1995).

When examining the country coefficients, Bolivian specialty coffees attain the same price
as the Brazilian coffees. Columbian, El Salvadorian, Nicaraguan and Honduran specialty
coffees, however, receive prices that are 10%, 15%, 20%, and 35% lower, respectively.
Similar to the rating and ranking, this result suggests that country of origin provides

valuable information to buyers that they might have an interest in transmitting to

32

 

 

alll

ind

 

consumers so they can also value the information on origin in general and certain origins
in particular.

Buyers also pay attention to the information about the variety of the coffee plant.
Pacamara has a positive implicit price of 24% with respect to the reference variety
Bourbon, indicating the preference of buyers for this type of tree. Similar to the other
reputation variables, the name Pacamara conveys an image of high quality that buyers are
willing to market to the final consumer. Catuai, Caturra, Typica and Other (varieties with
smaller representation) present no significant differences with respect to Bourbon,
indicating that these varieties are equally valuable as specialty coffees.

The reputation variable, quantity availability, is negatively associated with price. A one
percent decrease in quantity increases price by 0.38 percent. This indicates that specialty
coffee buyers place a higher value on coffees that are available in smaller quantities, i.e.
the specialty industry values exclusivity. We suggest that the marketing venue of this
attribute is on ‘limited editions’ of a coffee product. From the point of view of the coffee
grower, the coefficient of -0.38 represents the elasticity of demand facing the growers,
which indicates a price reduction less than proportional for each additional unit of
quantity available, i.e. growers can increase revenue by increasing quantity in the
specialty coffee market.

When examining the year variable, specialty coffee prices in 2004 and 2006 do not
present differences with respect to 2003. However, prices in 2005 were 24% significantly
lower on average. We consider year to be a macro variable that captures broader supply
and demand factors influencing coffee prices. Similarly, the commodity price estimate

indicates that the specialty coffeepremium is influenced by the commodity price level

33

during the auction month. For every one percent increase in the commodity price there is
a 0.58% increase in the specialty price. For specialty coffee producers, this means that
they are subject to the price fluctuations of the supply and demand factors in the broader
coffee market. The coefficient of 0.58 specifically indicates that while producers can
expect higher prices when the commodity price goes up, they can also expect that the
premiums, i.e. the difference between the two markets, is proportionally smaller when the
commodity price is higher.

Conclusion

Our hedonic analysis of the CofE data indicates that prices are determined by both a
sensory attribute, the quality rating determined by experts, and several reputation
attributes, which include the first places in the cupping competition winners, country of
origin, and incipiently tree variety. These results can be interpreted as representative of
the highest quality segments of specialty coffee and indicative of the trends in coffee
differentiation in broader coffee markets. In addition, the macro level variables
controlling for the commodity price and the competition-auction year were significant.
From the results we offer four main conclusions: 1) implicit prices for sensory and
reputation attributes can be distinguished in explicit prices of specialty coffees, 2)
reputation attributes constitute quality signals, 3) quality rankings have the dominant
impact on price, and 4) implicit prices of attributes at the producers level make coffee
growers visible to consumers and vice versa. An elaboration on these four key findings is
give below.

First, the hedonic equation of specialty coffees traded at the CofE e-auctions shows that

both sensory and reputation attributes determine the overall price of specialty coffees, as

34

expected for specialty food products. Moreover, the combination of attributes is such that
high sensory quality plus high reputation quality results in extraordinarily high coffee
values. A combination of all outstanding attributes results in super-premium prices, as
exemplified by the Brazilian and Colombian winners in 2005.

Second, reputation attributes convey information on quality that although related on the
sensory quality it is distinguishable from it. While the quality rating refers to the quality
of aroma, flavor and body, the quality ranking recognizes the consensus process to
achieve it and of the value of the quality certification thereby attained. The reputation of
the origin gives information that enhances the sensory experience by giving meaning and
a foundation to build on the concept that the coffee has been grown where conditions and
practices are best for growing coffee. This idea that there are unique combinations of
conditions that might be difficult to reproduce is what lies behind the marketing concept
of ‘limited edition’. The implication for supply chain coordination is that reputations can
be used as identifiable quality signals by buyers in their marketing strategies to
consumers.

Third, of all attributes quality ranking is the single most important factor in fostering
economic impact. At these relatively early stages of development of the specialty coffee
industry, these premia provide a super-incentive for stressing quality thus creating value
for the industry. To firrther support this statement, there is evidence that buyers at
specialty coffee auctions are trying to push the industry towards further differentiation
and greater value. This level of premia generates a symbolic content that enhances the
coffee product and creates excitement for all participants in this specialty supply chain.

This targets the marketing venue that one gets more pleasure by knowing that one is

35

drinking ‘one of the best coffees in the world’ as opposed to just drinking a ‘highly rated
coffee’. Furthermore, this type of value emphasizes that quality is really interesting in
relative terms and subject to the experts’ consensus for defining and measuring it. Such a
price structure is the salient feature of the ‘winner-take-all’ markets.

Fourth, disclosing the information on the sensory and reputation attributes of these
coffees to consumers draws attention to the role of coffee growers as crucial participants
in the final product quality. In turn, implicit prices inform producers about the product
characteristics that are most attractive to consumers, thus providing the producers with a
feedback signal of the demand. This suggests a much greater direct participation of
growers in specialty coffee as opposed to the commodity and other supply chains, such as
Starbucks and fair-trade. Overall, the two-way information flow through implicit prices
enhances the integration of production methods into the specialty coffee supply chain and
sustainability of the industry.

Finally, we expect new results from hedonic analysis of specialty coffee prices as the
industry evolves and more data becomes available. New specialty coffee regions may
become important as production programs result in successful specialty coffee production
or as the micro-climates change as the global weather changes. Varieties may also
become more desired as masters and consumers develop coffee connoisseurship.
Specialty coffees might be or become preferred by characteristics that we have not
included in this study (aroma, flavor, acidity and body). In particular, from this study it
remains to be analyzed whether altitude and the weather during the coffee growing
season are significant quality signals as is the case for wines. The impact on prices of

certifications and coffee processing at the grower level also remains to be tested.

36

THE IMPACT OF E-AUCTIONS IN ADJUSTING PROCUREMENT
STRATEGIES FOR SPECIALTY COFFEE

Introduction

Specialty coffee is associated with excellence in the quality of the underlying product
materials. This excellence is expressed in the taste, aroma, mouthfeel and other sensory
attributes (Daviron and Ponte, 2005). Specialty coffees are a differentiated product that
is freshly roasted, is usually released in limited quantities and is marketed to the
consumer with origin and processing information. Specialty Coffee supply chains require
the producers to utilize special cultivation and production practices to produce beans that,
once roasted, will give an outstanding flavor profile in the cup (Davids, 2002; Arvidson,
2003; Mabbett, 2006).l3 Procurement is the key function that coordinates the interface
between roaster and producer to identify, develop and sustain quality in the specialty
coffee supply chain. In this article we address the procurement approaches that center on
the necessary product material quality and therefore are unique to specialty coffee supply
chains. We examine both the traditional approach of supply relationships between
growers and masters; and the more recent development enabled by information
technologies (11') of electronic auctions (e-auctions).

The specialty coffee industry emerged when small roasters started to differentiate the

coffee product and utilize innovative business strategies to capture new markets for

 

'3 The original definition of specialty coffee ties a coffee’s flavor profile to the plant variety,
altitude, soil composition, fertilizer, sun, shade, rain, cultivation and harvesting methods all determine this

profile. This definition is what we have in mind for the purpose of this paper.

37

coffee drinking (Roseberry, 1996; Ponte, 2001). Specialty coffee consumers demand high
quality and differentiation, i.e. they want coffee that tastes better than regular coffee and
displays different flavor profiles according to the variety of coffee origins. Hence,
maintaining a fairly broad assortment of products is important for specialty coffee firms.
On the other hand, since individual consumers have idiosyncratic preferences over a
differentiated product, a specialty coffee firm may find itself targeting a particular niche
segment. For example, some consumers prefer a bright coffee (higher acidity) while
others prefer it flat (lower acidity), either the roaster carries both types (differentiation
strategy) or specializes in one (niche strategy). A result of this differentiation is the
fragmentation of the coffee industry, with many small firms competing for small shares
of the overall market. Concurrent with this development, e-auctions have emerged as a
means for masters to procure beans.

The objective of this paper is to investigate the role of e-auctions in complementing
traditional supply relationships in coordinating supply and demand in specialty coffee
supply chains using a newsvendor model. In particular, we concentrate on the impact of
auctions in facilitating the adjustment of the roaster’s procurement strategy through the
possibility of making use of demand information updating. The classic newsvendor
model is a suitable starting point for analyzing the procurement problem in specialty
coffee for a number of reasons. First, stochastic demand mirrors the roasters uncertainty
when planning in the growing stage of the specialty coffee market. Second, the model
takes into account shortage costs, which fit the differentiation objective of the specialty

coffee business, i.e. satisfaction of individual customers is important and the inability to

38

meet demand is costly. Third, the model accounts for the perishability, which is
convenient due to the importance of freshness for specialty coffee.14

Our analysis recognizes that roasters design their procurement strategy at the beginning
of the season based on previous years and current information, anticipating that they will
sell an approximate number of coffee units based on their market size and market
variability. Supply relationships with producers are established early in the season when
demand is still uncertain since the specialty coffee firm needs to identify and contact an
adequate supplier, negotiate the terms of the relationship and initiate a contract with them
in advance of the coffee harvest (Figure 3). Underestimating demand results in a cost to
the roaster associated with lost sales and lost customers. Overestimating demand results
in a loss of value of the excess beans that will not sell at the full price. If possible,
roasters would like to adjust their inventory such that if demand is greater than expected
they would adjust their procurement strategy. This is possible whenever there is the
alternative of procuring from an auction later in the season when the coffee beans are

harvested and processed.

 

" Once coffee is roasted, it is difficult to preserve the volatile compounds that make up the coffee aroma,
which are the hallmark of specialty coffees. However, there is no consensus on how fresh roasted coffee
must be. Retail roasters tend to favor a shorter optimum life-span (3 to 7 days) for the best taste while
wholesale roasters contend that the optimum period of time can be greatly expanded into weeks or even

months with appropriate packaging technology (Holly, 2005).

39

A roaster ’s procurement strategy planning season with
demand information update over time

 

 

 

A

fi \

l I l i
Establish a supply Coffee E-auctions Coffee roasting
relationship with harvest and sales
producer
LESS MORE
INFORMATION INFORMATION

Figure 3: Timing of roaster’s procurement decisions in the coffee season

At the beginning of the season, the roaster must decide how to use the alternative
procurement forms available to them. It may be that the firm can only purchase through
relationships or through relationships supplemented by auctions. In both cases the roaster
has to decide which quantity to procure through the relationship. If auctions are not
available, the optimal decision balances the cost of buying too much versus the cost of
being short. If the roaster has the possibility of utilizing the auction, the decision balances
the tradeoff between the costs of using each of the two procurement approaches. Based
on our model, we discuss the benefits of having e-auctions in addition to relationships
under different market circumstances.

This article proceeds with a literature review on supply chain procurement decisions that
we use to frame and model our problem. In the second section we describe supply
relationships and e-auctions in specialty coffee and their economic properties. In the third
section we present the specialty coffee procurement model and the benchmark case that
considers relationships as the sole means of procurement. The fourth section contains the
model solution including the optimal use of e-auctions and use of relationships when

auctions are available. The fifih section discuses the importance of e-auctions as part of

40

an optimal procurement strategy. In the final section, we give a brief summary and
conclusion. Detailed derivations of the model results are provided in the Appendix.
Framing and Modeling Procurement Decisions

Three bodies of literature are relevant to our analysis. The first concerns the classical
question in organizational economics of markets versus hierarchies or the choice of
organizational forms; the second arises from strategic management; and the third arises
from operational management. In the presence of uncertainty, which emanates from
imperfect information, bounded rationality, strategic behavior and incomplete markets,
different supply chain organizational forms emerge to reduce transaction costs, improve
value creation and incentives (Milgrom and Roberts, 1992; Foss et al., 1999; Williamson,
1999; Foss, 2003). Defining the alternative organizational forms in which a given
transaction can be organized and illustrating the tradeoff between them is the goal of the
theory of the firm (Gibbons, 2005). Our paper builds on this body of literature, we view
procurement decisions as pertaining to the design and choice of organizational forms for
coordinating the economic exchange. We emphasize the complementarities between
relationship procurement that entails committing to a decision before the uncertainty is
resolved and the costly modification of a decision through an auction when uncertainty is
resolved.

The main focus of the strategic management literature, as it relates to our work, is to
determine when a particular organizational form may be pursued given a number of
conditions (Pyke and Johnson, 2003). The choice of procurement strategy depends on
several critical variables, hence, usually no single strategy dominates another in all

dimensions (Bunn, 1993; Elmaghraby, 2000; Grey, 2005). Therefore, procurement

41

managers must understand the critical variables that affect the decision and evaluate the
performance of each procurement strategy used in the context of the specific institutional
conditions (Elmaghraby, 2000; Grey, 2005). This literature is concerned with questions
that applications of information technologies increasingly pose in procurement, including
when to use auctions and how auctions complement supply relationships (Elmaghraby,
2000; Moss et al., 2003; Pyke and Johnson, 2003; Swaminathan and Tayur, 2003; Cattani
et al., 2004; Bunduchi, 2005; Grey, 2005). Our analysis draws on this literature to assist
in selecting the relevant factors that affect the procurement decision in specialty coffee
and outlining economic properties of relationships and auctions.

The two papers most related to ours are Peleg et al. (2002) and Chang et al. (2005). In
terms of operationalizing the procurement management questions, Peleg et al. (2002) also
use a newsvendor model. They compare three procurement strategies: first, a short-term
contract with an online supplier chosen on the basis of the lowest cost; second, a long-
term relationship with a single supplier; and lastly a combined strategy that uses both.
The authors compare the three strategies on the basis of their expected costs. They
calculate the minimum discount that a supplier should offer in the long-term relationship
to be preferred to the lowest cost online supplier. The authors further explore the effect of
increasing the supplier base on the price distribution and its relation to the variance of
demand. We follow Peleg et al.’s use of the newsvendor model to study the choice of
procurement strategy. However, the main difference between our approach and theirs is
that in theirs the manufacturer chooses between the alternatives of a relationship, an
auction or a combination of the two simultaneously when facing demand uncertainty. In

our model, the decisions of procuring from relationships and auctions are sequential, with

42

uncertainty being resolved between the timing of the two. Thus, we focus on the benefit
that the auction adds to the procurement strategy as a complement to the relationship.
Moreover, we tailor our analysis to the peculiarities of specialty coffee procurement,
whereas Peleg et a1. do not specify an industry context.

Chang et al. (2005) also apply the newsvendor model to examine and compare traditional
procurement and e-procurement to quantify the effect of the Internet on supply chain
procurement management. They model two forms of traditional procurement: a vertically
integrated supply chain and a decentralized supply chain. Transaction costs are negligible
in the vertically integrated supply chain, whereas in the decentralized supply chain, buyer
and seller have competing objective functions. They found that expected profits are
greater in the e-procurement strategy if the supplier provides return policies (unit
refunds). They also found a maximum cost for IT procurement, above that point they
would not participate in the e-market because its expected profits would be lower than
that in the traditional procurement strategy. We present a similar analysis based on the
comparison of optimal quantities and expected profits. However, we compare the
classical newsvendor, for the case when the relationship is used exclusively, with an
extended model, which is a combined relationship and auction strategy. Our model, thus,
extends the classic newsvendor in two ways: first, by allowing for different states of
information concerning demand and second, by considering a strategy that covers more

than one period.ls

 

'5 See Khouja (1999) for a survey of the traditional newsvendor model with extensions.

43

Supply Relationships and E-Auctions in Specialty Coffee Procurement

Relationships and e-auction procurement focuses on the evaluation of sensory quality
through cupping and remuneration of producers accordingly. In the cupping, roasters
evaluate the intrinsic quality of the coffees by testing samples of the roasted and brewed
coffee and assessing its aroma, flavor and mouthfeel according to established procedures.
The characteristic and common feature of these procurement approaches is the
transmission of quality information to the source (producers)(Ponte, 2002).

Relationships have been the traditional procurement approach for roasters concerned with
quality (Castle, 2006). They consist of long-term interactions in which roaster and
producer work closely together to create a distinctive cup of coffee for the final consumer
(Griswold 2000; lntelligentsia 2004). Producer and roaster exchange information about
the desired coffee quality attributes and how to obtain them; the roaster ‘interpret’ the
distinctive flavor profile that the grower produces from the unique coffee origin
characteristics and cultivation practices (Foley, 2004; lntelligentsia, 2004; Daviron and
Ponte, 2005; Castle, 2006). The costs of a relationship are initially high but their benefits
are accumulated over time. Incentives, more than price, are provided by means of
informal agreements, trust and mutuality. The result is that relationships allow the
consistent supply of high quality coffee beans year after year from a limited number of
producers.

In relationship procurement, roasters incur expenses in traveling to producing countries to
evaluate and select those farmers whose product matches the desired characteristics and
to better understand their business partner's operations (Griswold, 2000; lntelligentsia,

2004). Upon identifying potential growers, roasters are faced with negotiating the terms

44

of the exchange and, upon successful contracting, monitoring the quality of the product.
Established relationships usually involve pre-financing and marketing arrangements
depending on the degrees of involvement from roasters (Ponte, 2001; Daviron and Ponte,
2005).“5 Once the relationship has been established, roasters have secured a coffee supply
at a fixed price in advance of the season which saves them time and expenses and allows
them to concentrate on roasting and marketing activities (Griswold, 2000; Daviron and
Ponte, 2005; Castle, 2006).

In auctions, coordination is achieved through competitive bidding (i.e. "arms-length" or
market coordination) between buyers and sellers so there is a threat of losing business to
a competing supplier/buyer. In specialty products, there is scope for this market-type of
coordination provided that complex quality information can be embedded in standards,
labels and certification procedures (Ponte and Gibbon, 2005). This is what specialty
coffee auctions pursue. The evaluation and selection of coffees using a standardized
cupping form and procedure allows transparency in the revelation of quality. Moreover,
the provision of coffee samples to the potential bidders prior to the auction, allows
quality assessment at a relatively low cost. E-auctions increase the supply of specialty
coffee through quality training and improvement and the incorporation of previously
unknown producers in the countries which host the auctions (United Nations, 2003;

Daviron and Ponte, 2005).

 

'6 In ‘relationship coffees’ the level of involvement is highest and masters and grower work jointly to
define and develop the coffee product quality and the tasting experts of the firms are directly involved in

coffee production (Griswold, 2000; lntelligentsia, 2004).

45

E-auctions offer the possibility of finding high quality at low search cost as the upfront
investment for using this procurement approach usually entails only a modest auction fee.
This is particularly valuable for small roasters for whom it is more costly to find the right
quality of coffee year after year (Daviron and Ponte, 2005). Auctions disclose quality
information (including the pro-auction samples to potential buyers) that reduces
transactions risks (Daviron and Ponte, 2005). While these factors result in higher prices
paid to growers, auctions increase the flexibility of adjusting the procurement strategy of
the roaster by allowing access to additional supply if needed.
The Analytical Framework
We begin our analysis by presenting the model. followed by solving a benchmark case in
which e-auctions are not available which is the classic Newsvendor case. Next we extend
the model to include e-auction and solve.

The Model
We model a specialty coffee supply chain consisting of a grower/producer and a roaster.
The roaster sells a consistent coffee product in the final demand market at a given
expected price p per unit of a uniform quality level of the coffee beans procured. The
desired quality of coffee beans are obtained either through a relationship with the grower
or at an auction in which many growers offer their product. In the case of the use of a
relationship, the roaster commits to procure an agreed-upon quantity xR from the
producer. For the benchmark analysis, this relationship is the only procurement method
available. In the complete strategy, the manager can augment the total procurement

quantity by purchasing additional coffee units at an e-auction.

46

In the relationship, the roaster incurs variable costs cR per unit, which correspond to the
price paid to the producer for the coffee beans plus roasting and packaging costs per unit.
In addition, the roaster incurs fixed costs in the relationship KR, which correspond to the
transaction costs of finding a qualified producer, negotiating the terms of the exchange
and monitoring quality.

In the auction, variable costs are 6,; per unit, representing the roaster’s price expectations
at the auction at the time that relationship decisions are made.‘7 Fixed costs are KA which
include the auction fee. We assume that the (expected) auction variable cost is higher
than the relationship variable cost, CA > cR, because transaction in the relationship are
reduced over time (risk and uncertainty).

The selling price, salvage value and all costs are known to both the roaster and the
producer at the beginning of the season. Each unit of coffee not sold in the first period
has a salvage value of v. All unsatisfied demand is subject to a shortage cost s. We
maintain the natural newsvendor assumption for a non-trivial solution that the selling
price is larger than the unitary cost and the unitary cost is larger than the salvage value,
i.e. product is perishable, p > cA > C]; > v.

The risk neutral roaster determines its procurement strategy based on the maximization of

expected profit, given uncertainty in the final demand market. Specifically, the roaster

 

'7 Prior price expectations can be based on past experience, but also on anticipated properties of the coffee
sought (on this see Donnet et al., 2007). For more on price formation in auctions see Wolfstetter (1996) in
general and Jeitschko (1998, 1999) for applications to agricultural and fisheries wholesale markets. For
simplicity, we deliberately abstract from the price formation process in this article in order to highlight the

tradeoffs and complementarities between the two procurement methods under consideration.

47

faces random uniformly distributed demand for the final product with mean a and

dispersion 17. The realization of demand D is positive for the range[1_),D], so that
Q=,u—nandD=,u+77.

The sequence of events is illustrated in Figure 4. At the beginning of Stage 1 the roaster
decides how many coffee units to purchase from the relationship, xR, while facing
uncertain future demand. After this decision is completed the roaster observes the
realization of demand, subsequent to which the firm can go to an auction and get an
arbitrary amount of additional coffee units, xA, to supplement its supply obtained through
the relationship. Only after this decision, is the product brought to the final demand
market and profits 7r (xx xA) are realized. The firm chooses its procurement strategy to
maximize expected profits Err (xR, xA).

We calculate the optimal quantity and expected profits for two possible procurement
strategies: First, we consider the traditional case in which auctions are not available and
supply chains rely on relationships exclusively (this is the benchmark case). Second, we
consider the complete strategy involving access to both relationships and e-auctions.
Throughout the paper a superscript * denotes the optimal solution to the full model,
whereas a superscript B denotes the optimal (benchmark) solution relying solely on

relationships without access to auctions.

48

 

 

 

A A
f i f N
I l 1 l \ t
I l l l 7
Commit toD is realized Purchase xA from 7:13 realized
purchase xR from auction

relationship

Figure 4: Sequence of events in the specialty coffee procurement model

Relationship procurement without the availability of E-auctions
We first examine the benchmark strategy in which the firm can only utilize a supply
relationship. This is the classical newsvendor problem in which the manager’s decision is
to balance the tradeoff between the costs of procuring excess coffee units and the costs of
not meeting the full demand. At the end of Stage 2, if the procurement quantity is larger
than the realized demand, the (overage) profit is the simple profit associated with selling
D units, augmented by the salvage value of the excess inventory (xR — D). If instead the
procurement quantity is smaller than the realized demand, the (underage) profit is the
revenue minus costs from selling xR units minus the shortage cost of not meeting the

excess demand (D — xR). That is,

pD+v(xR—D)—cRxR—KR if _D_<D<xR
”805ml”: (1)
(p—CR)xR—S(D-xk)‘KR If xR<D<D.

Given our assumption on the distribution of demand, the firrn’s expected profit, when
there are no e-auctions to adjust the procurement strategy after the realization of demand,

is expressed as:

49

Err” (xR) =i{-l(p+s—v)xk2 —l(p--v)Q2 +(cR —v)xRQ
277 2 2
_ 1 __2 (2)
+(p+s—cR)xRD—ESD }—KR.
Taking the first order (sufficient) conditions for profit maximization and simplifying, the
optimal quantity to procure can be expressed as a convex combination of the upper and
lower bound of the possible states of demand, i.e., the summation of the relationship

contribution margin multiplied by the maximum possible demand and the relationship net

cost multiplied by the minimum possible demand over the revenue from sales:

,=___(,...-.,)E+__(.,-.) Dames, (3)

xR (p+s—v) (p+s—v)—- MR MR _

where MPR is the marginal profit of selling an extra unit at full price at the relationship
variable cost, MCR is the marginal cost of not selling an additional coffee unit procured
through the relationship and MR is the marginal revenue of selling an additional coffee
unit at full price. Substituting the optimal procurement quantity into the expected profit
function and consolidating gives the optimized expected profit expression for the

benchmark case,

 

E7r8(psc c vK p77)=l[(p+S-CR)—(CR-v)]2 +l(p+S-CR)2
’ ’ R, A, ’ R, ’ 4 (PH-V) 2 (pH-V) (4)
l (CR-v) 1

l
————— -— — - — - -K.
2(p+s-v)# 4(p+s v)77+2(p S v),u R

Optimal Procurement with the Use of E-auctions
The complete model is solved through backward induction. For the strategy with the
availability of an auction we begin by examining the decision of utilizing the auction at

the beginning of Stage 2 when the realized demand for the firm’s product, D, is known

50

and the firm has already secured the quantity xR through its relationship arrangement.
Given the optimal continuation profits determined in Stage 2, we then turn to examining
the optimal order quantity from the relationship at the beginning of Stage 1 that
maximizes overall expected profits.
The optimal use of the e-auction

At the beginning of Stage 2 the firm knows the quantity procured through the relationship
xR and the demand for its product D. If the relationship order quantity is larger than the
realized demand, xR 2 D, no further decision is made in the procurement strategy. The
firm’s profit is given in the top branch of Equation (1). However, if the order quantity is
not sufficient to meet realized demand, xR < D, the firm decides whether to supplement its
coffee supply by going to the auction. If the firm chooses not to go to the auction,
shortage costs of s are incurred per unit of excess demand. Hence, the profit expression at
the end of Stage 2 is given by the profit margin multiplied by the procured units,
diminished by the relationship fixed costs and the shortage costs incurred on the units in
excess demand. F ormally,

(p-cnxi. _1<,, -s(D-x.). (5)
In contrast, if the firm chooses to go to the auction, in addition to the profit from the
relationship, there is a profit for selling the excess demand at the auction margin, p — cA,
while incurring additional auction fixed costs of KA. Since the cost of procuring
additional units at auction, cA, is smaller than the selling price, p, it follows trivially that
conditioned on incurring the fixed cost of attending the auction, K4, the firm optimally
chooses to purchase the entire shortfall D — xR at the auction. The profit expression at the

end of Stage 2 for this event is, thus,

51

(P-CR)xR-KR+(p-CA)(D-XR)-KA- (6)
The decision to go to the auction balances the trade-off between equations (5) and (6).
The firm optimally chooses to participate in the auction and purchase the excess demand
if the profit from participating is larger than the cost of not participating,
— s(D — x R) S (p — CA )(D — x R) — K A . Rearranging the inequality, the marginal profit for
participating has to be larger than the marginal cost of participating, i.e.,

K A S (D — xRXp + s —cA). This implies a demand threshold 3? above which it becomes

profitable to use the auction

 

x(xR):=————+xR = +xR (S D), (7)

where MPA is the marginal profit of selling an additional unit at a full price at the auction
variable cost. In other words, the shortfall must be sufficiently large in order to warrant
incurring the fixed costs of participating in the auction to supplement supply and replace
shortage costs with additional profit.
The use of relationships when e-auctions are available

As just sketched, given the past decision on engaging in the relationship, xR Z 0, and the
realization of demand, D, the firm’s profits at the end of Stage 2 fall into three cases. In
order to ascertain the optimal relationship strategy, the roaster must determine expected
profit across the three cases, given a choice of size of the relationship quantity.

To recap, in a first case, the realized demand is larger than the relationship quantity
procured and thus no further decision is required. Profits equal the revenue from realized
demand plus the salvage value of the excess units minus unitary costs for the procured

units minus the fixed costs incurred in the relationship. In the second possible case, the

52

realized demand is larger than the relationship quantity procured but the benefits from
participating in the auction are insufficient to off set the fixed costs incurred, so the firm
chooses not to participate. Profits consist of the margin from the procured units through
the relationship minus the relationship fixed costs minus a shortage cost of each unit of
excess demand. The third possible case occurs when realized demand is larger than the
relationship quantity and the benefits from using the auction are sufficiently large to
cover the fixed costs incurred from utilizing the auction. Profits result from the
relationship margin over the relationship procured units plus the auction margin over the

excess demand minus the fixed costs of the two alternatives:

rpD+v(xR—D)—cRxR—KR if _D_<D<xR

7r.(xR,D)=<(p—CR)xR—s(D—xR)—KR If xR<D<36 (8)

 

t(p—CR)xR +(p—CA)(D-xk)_KR ”K14 if 35 < D<B.
Integrating over the possible realizations of demand and consolidating, the expected

profit at the beginning of Stage 1 from the strategy that includes using e—auctions is:

a I I _
En (WEI-{71cm-(v-c.)x.2+(c.-c.)x.o

I "2 I 2
Tab-CA)” _'2'(P"V)l_) (9)
2
+lK—A—K,E+K,x,,}
2 p + s — CA
— KR.
The firm maximizes expression (9) by optimally choosing the relationship quantity xR.

The first order condition gives the optimal relationship quantity in anticipation of being

able to access an auction at the beginning of Stage 2:

53

._(_,___c —-c,,) (cR- ___)vD K, AVC— MCRD+ K,

x (c, -v)D (c,—v)— (c,--v)= MC, +MC,— MC,‘ (10)

 

Notice that the optimal relationship quantity consists of three terms. The first term is the
ratio between the relationship cost advantage over the auction and the auction marginal

cost, MC A := c, —v, multiplied by the maximum demand. The second term is the ratio

between the relationship and the auction marginal costs multiplied by the minimum
demand. The third term is the ratio between the auction fixed and marginal costs. The
first two terms give a weighted average of the end points of the demand support, while
the final term augments this convex combination increasingly in the cost of attending the
auction.

Substituting the optimal procurement quantity (10) into the expected profit expression

(9), optimized expected profit is given by

1[(CA —CR)-(CR -V)I2 1

 

 

E”.(p’s9cR9C/1’v’KR’KAHu’U)=_ (C _v) U—Z(cA-v)77
+1(CA- CR)2_ (CR—vyfl
+2 (CA-V) (11)
I
:Eip—Callh‘EiP-Vlfl

_1__2 KAzl- (p+s— v) _
KL(c,—v)(p+s— c,):L K” KR.

The Importance of E-auctions as Part of an Optimal Procurement Strategy

As shown above, in the optimal strategy when auctions are available, the firm’s decision
concerning the extent of the relationship is based on the relation of marginal costs
between the two alternative procurement forms. In the benchmark strategy sans auction
(the newsvendor problem), the decision of the firm is based on the relation of marginal

profits and marginal costs of the relationship alone.

54

We now consider how the optimal relationship strategy and overall firm profitability is
affected by the availability of the e-auction. In particular, we examine the changes
induced in the relationship procurement quantity due to the auction option, and compare
the expected profit between the benchmark and the complete strategy. In so doing, we
also consider how the impact of the e-auction varies with the roaster firm size; proxied by
the firm’s expected demand ,u, and the demand uncertainty, captured by the dispersion
parameter n. From this comparison we evaluate the benefit of accessing e-auctions in
specialty coffee procurement and how this is affected by firm size and uncertainty.
The probability of using the auction

Given the distribution of demand and the critical threshold} , above which it is
profitable to attend the auction, one can calculate the ex ante probability that the auction
will be attended to procure additional coffee units to meet any demand exceeding the

units procured through the relationship. Letting P{A}denote the probability that the

auction is utilized, one finds

 

P{A}=P{D23‘c’}=(cR-v)—(p+l::c )(p+2;—v)=MCR_§Afl

 

Thus, the probability of attending the auction is the difference between two terms. The
first term is the marginal cost of the relationship. The second term is the product of two
factors, one referring to the auction cost and the other referring to revenue from coffee
sales. The cost factor is weighted by the demand variability and the revenue factor is

weighted by the auction margin. The first cost factor is the ratio of the auction fixed costs
~ . K A . . .
to demand uncertainty, —2— , and the revenue factor 18 the ratlo of marglnal revenue to
77

55

. This can also be expressed as the inverse of the auction

 

marginal auction profits,
A

gross profit margin, GPMA, which is the ratio between marginal profits and marginal
MP, ~
revenue, W , and measures the gross profit earned on sales.

Note that the probability of attending the auction is increasing with the marginal cost of
the relationship, and decreases when auction fixed costs are not expected to be diluted
from a large demand variability and when the margin earned is large relative to the
revenue from selling an additional coffee unit. This last factor can be expressed as the
inverse of auction gross profit margin (Figure 5). Thus, the probability of attending the
auction decreases with the inverse of the auction gross profit margin. The product of the
two auction factors accounting for costs and revenue can be interpreted as the marginal
profit from participating in the auction which is the same as its opportunity cost, i.e. the

cost of not participating in the auction.

56

 

Probability of Attending the Auction as a Function of Demand Variability

 

0.. '.

Probability

 

_—__—-~_._.____ _4

 

 

2n
Demand Variability

 

 

 

L —l — high auction marginal profit + medium auction mrg'nal profit low auction marginal profit L

 

 

Figure 5: Auction attendance as a function of demand variability and gross profit

margins

Proposition 1 The probability of attending the auction is positive if the marginal cost of
purchasing an additional coffee unit fiom the relationship is larger than the auction

5. I

o ortuni cost, i.e., MC > .
PP W R 217 G P MA

 

Proof: The proof follows readily from Equation (12).

The fact that the probability of attending the auction is higher for roasters whose
relationship marginal costs are higher gives additional insight. Recall that relationship
marginal costs are given by MCR = cR — v, so higher relationship marginal costs includes
both the case of increased unitary cost in the relationship cR as well as the case of facing a
lower salvage value v.

The former indicates that auctions are more appealing to roasters who are unable to
secure enough benefits through the relationship so as to reduce risk and costs, either
through negotiations or through long-term (possibly joint) investment that lower

transactions and production costs and thus decrease CR. As for the latter, i.e., a larger

57

salvage value v, our model indicates that auctions are more appealing to roasters who
emphasize freshness and short product cycles as well as those willing to pay a premium
for high aroma; compared to those with a low preference for freshness and aroma.
Overall these differences correspond to differences between retail and wholesale roasters;
or between top quality specialty masters and average quality roasters.
In addition, the ex ante probability of attending the auction is higher for roasters who
have a larger auction gross profit margin and who can expect to dilute fixed auction costs
from expected variability of demand—corresponding to new products and new market
niches for example.
For the remainder of the analysis we assume non-trivial parameter values such that there
is a positive ex ante probability that the roaster attends the auction.

The impact of auctions on relationship procurement
Let Ax}; := xRB — xR* denote the change in the relationship procurement quantity induced
by the inclusion of the auction strategy. That is, AxR is the difference between the
relationship procurement quantity with and without the availability of the auction.
Substituting for the Optimal quantities, given in Equations (3) and (10), the change in the
relationship quantity from the no auction to the auction strategy depends on the marginal
cost of the relationship, the auction marginal profits and marginal and fixed costs and the
marginal revenue from coffee sales:

(p+s— c,)(cR— v)— -MCRQ— MP, MCRD+ K,
(p+s— v) (c,-v) (c, K-v) MC, MR MC,

 

 

AxR = (CR — ‘02"

(13)
Proposition 2 The availability of the auction reduces the optimal size of the relationship
procurement quantity, xR“ < xRB..

58

Proof. Given the assumption that there is an ex ante positive probability of attending the

(p+S-C,XCR -V)

auction, Equation (12) implies thatK, < (p ) 277. Therefore, using
+s—v

Equation (5) while recognizing that 2r].is simply the length of the support of the uncertain

<(c,-cR)—+(cR—v) +(p+S-CA)(cR-v) __
(.,-.)D (cl—vl-Q (p+s—v>(c.—v)(” 2"

demand, it follows that x,‘
However, simple algebra reveals that the right hand side of this inequality is equal to xRB,
given in Equation (3). Hence xR" < xRB. Q. E. D.

The optimal quantity to procure from the relationship is smaller when e-auctions are
available versus when e-auctions are not available. Proposition 2 states that less quantity
will be purchased through the relationship and any sufficiently large excess demand that
results will be fulfilled by participating in the e-auction. The factors that are driving the
reduction in relationship exchanges are the perishability of the product and the ability to

dilute auction fixed costs with auction profit margin.

Proposition 3 The change in the relationship procurement quantity, from the benchmark

 

. . . . , dAxR
to the strategy wrth auctions, increases wrth uncertainty, (1 > 0.
77

Proof. The change in quantity procured through relationships with respect to uncertainty

 

 

 

 

is given by,

dAxR = deB _dx,‘ = (p+s-cR)—(c,—v)_(c,-cR)-(cR—v) =2(p+s-c,)(cR—v)>
dry dr7 dn (p+s—v) (c,—v) (p+s—v)(c,—v)

QED.

When demand uncertainty increases, relatively more quantity is procured through the use

of auctions versus relationships, i.e., the role of the auction for specialty coffee

59

O

procurement is more important when demand uncertainty is higher. Therefore one
expects to find more procurement through auctions by roasters who operate in growing
markets versus roasters in mature markets. Moreover, as new specialty coffees are not
introduced to the market, there is considerably greater uncertainty about their demand
than there is for established specialty coffees. Consequently, new specialty coffees are
likely to be procured through auctions rather than relationships. This is consistent with
the evidence from the industry thus far, in which auctions started to being used for
emerging specialty coffee origins (Brazil, Nicaragua, El Salvador)."3

The impact of demand uncertainty on profits with the optimal use of
auctions
We define Art as the difference between the expected profit in the strategies with and

without e-auctions:

Arr(cR,c,, p,v,s,K,):= En‘ —E7r" (14)
. . . . . . dArr
Proposrtron 4 The Impact of e-auctlons decreases With the srze of the firm, 7 < O,
[1

whenever there are positive shortage costs from excess demand, i. e., if s > 0.
Proof. The impact of the change of firm size on the increase in profit due to the

introduction of auctions is given by,

 

'8 More recently, coffees from the already emerged origins of Costa Rica and Guatemala have been traded
in auctions; auctions are not used for trading coffees from origins with an established reputation as
specialty producers, such as Hawaii (the Kona coffee) and Jamaica (the Blue Mountain). For emerging,

emerged and established origins see for example Davids (2004).

60

dAEn _ dErr' _ Err”
dp dy d,u

 

2%(p-CR)-%(p+S-CR) =-%s< 0,\7’s > 0.Q.E.D.

Proposition 4 states that the positive impact of auctions is larger for smaller versus larger
roasters if shortage costs are important. Shortage costs give the economic implication of
losing a sale, which entails more than merely the lost revenue from the unsold item. For
example, it also accounts for the fact that a customer and their future purchases are lost
because they are not satisfied with the roaster’s business. This result emphasizes the role
of auctions in developing customer’s satisfaction and gaining customer’s loyalty in the
differentiated specialty coffee market. Auctions are more important when customers’
loyalty matters, which is especially the case for small firms that compete in a
differentiated market. As a corollary, the model indicates that auctions are more valuable
in fragmented versus consolidated industries. Thus, if there is imitation of corporative
strategies that allow consolidation, auctions become less important.

Proposition 5 The profitable impact of auctions increases with demand uncertainty,

dArr
do

>0.

Proof. The impact of auctions is measured by the increase in expected profits due to the

use of the auction. This increase is affected by demand variability as follows,

dAEn=dErr'_dE7rB =[(c,-cR)—(cR—v)]2_ 1 K 2 (p+s—v)
do do do (Cl-V) 4o2 A (cl-VXpH-CA)
(p+S-CRXcR-v)
+
(_p+s-v)

 

 

 

 

61

Solving and rearranging, the condition for the derivative to be positive is that there is a

positive probability of attending the auction, implied by Equation

 

(12); K, < (pH-CAXCR 'v)2n. QED.

(p + s — v)

The positive impact of auctions on the roaster’s profit is higher when demand uncertainty
is higher. This situation is present in new specialty coffee markets as new products whose
potential for capturing new customers is yet to be tested and the potential for losing initial
customers is high. The importance of auctions may diminish as these market segments
mature, i.e. when consumption is established for routine customers and there are no new
consumers to capture.'9
As alluded to above, the model lends itself to additional interpretations of the role of
auctions in specialty coffee markets, namely as facilitators of the discovery process.
Propositions 4 and 5 jointly indicate that new (i.e., small) market segments, with
unknown potential (high variability) are possibly best dealt with by means of procuring
quality beans in auctions. That is, auctions can serve as tools in exploring possible market
niches which upon maturation (larger, more stable demand) can also easily be
accommodated with traditional supply relationships. I

Conclusion

This paper analyzes the importance of e-auctions in the procurement strategy of specialty

coffee roasters. In establishing relationships, the roaster typically commits to purchasing

 

‘9 The analysis is confined to the particular model and market structure we are examining. Clearly auctions
may work well for other reasons and play important roles in other contexts, even in mature markets, as is

the case in many well-established action markets for agricultural and fisheries products.

62

a quantity at the beginning of the season when demand is uncertain. E-auctions can be
incorporated into the procurement strategy later in the season when more information
about demand is available to the roaster. The focus of our analysis concentrates on the
importance of e-auctions for adjusting the procurement strategy once information on final
demand has been updated. We compare a benchmark strategy in which no e-auctions are
available with a combined strategy in which e-auctions are available to complement the
specialty coffee firm’s prior supply obtained through relationships. The benchmark
strategy is a classical newsvendor problem in which the procurement manager balances
the marginal profits of selling an additional unit and the marginal costs of not selling an
additional unit. The combined strategy is an extended newsvendor model in which the
procurement manager balances the marginal costs of the two procurement forms.

Our results indicate that e-auctions matter the most to small roasters (rather than large
wholesalers) who concentrate on top end quality and niche markets. The model further
suggests that newer markets are best opened and explored by utilizing the auction system,
and fragmented markets may also benefit more from the addition of auctions as
alternative methods of procurement when compared to more consolidated markets for
coffee.

By facilitating the procurement of new specialty coffees in an expanding market, auctions
are important in the discovery process that is essential to product innovation and

maintaining competitive advantages in the specialty coffee industry.

63

PRODUCT DIFFERENTIATION THROUGH QUALITY RATINGS IN
SPECIALTY COFFEE: A CROSS-ENTROPY ANALYSIS OF E—AUCTION S
DATA

Introduction

Today nearly 90 percent of consumers understand that coffee is a differentiated product,
like wine and cheese (Beaver, 2006) 20. Through product innovation and differentiation
masters and retailers have created value added in what is known as the specialty coffee
industry (Roseberry, 1996; Ponte, 2001)". The industry offers a wide variety of coffees
differentiated by their origin and the variety of the coffee plant, the roast and form of
preparation, the packaging and the ambience of coffee shops and the supermarket aisle.
In this paper, we concentrate on the quality rating of specialty coffees as an index of
product differentiation and a price signal to growers. Quality ratings summarize the
opinion of expert tasters about the sensory quality of a product. Evaluation through
quality ratings is a successful differentiation strategy in specialty products that require
connoisseurship for a full appreciation of quality, as it is exemplified by wines. We
investigate the relation between quality ratings and price in specialty coffee using cross-
entropy analysis of quality ratings and prices of specialty coffee traded at e-auctions. We

address the following questions: 1) is there vertical differentiation across specialty

 

2° Specialty coffees are defined by the Specialty Coffee Association of America as the highest quality green
coffee beans roasted to their greatest flavor potential by true craftspeople and then properly brewed to well-
established standards (Holly, 2005).

1 Product differentiation refers to the strategic decisions that seek to distinguish the product from

competing products, to enhance its value and the revenue it generates (Moulton et al., 2001).

64

coffees according to the quality rating, 2) how effective is the quality rating as an index
of vertical differentiation, 3) how informative is the price signal to the grower with
respect to the quality rating, and 4) how strong is the quality rating for remunerating
quality.

Product differentiation refers to the strategic decisions that seek to distinguish the product
from competing products, to enhance its value and the revenue it generates (Moulton et
al., 2001). Product differentiation by quality ratings is a vertical differentiation, as
opposed to horizontal, because they order products according to their quality from the
highest to the lowest, i.e. it is possible to say that one good is "better" than another
(Piana, 2003). Specialty coffee chains are of particular interest because coffee buyers
(exporters, importers and roasters) pursuing a differentiation strategy are likely to
coordinate quality issues more closely with suppliers relative to buyers in the commodity
chains. This ‘explicit coordination’ suggests that producers can get more informative
price signals from upstream buyers in the specialty chains (Gereffi, 2005). A 100-points
quality rating has been adopted to summarize different categories of the product sensory
evaluation in a number to convey a straight forwards and comparable idea of quality. The
Specialty Coffee Association of America (SCAA) developed a cupping form that seeks to
serve all segments of the specialty coffee industry (Songer, 2004). By standardizing the
cupping, the SCAA intend to prevent the loss of meaning of the term ‘specialty’ from the
addition of milk, water, syrups and others (Daviron and Ponte, 2005).

The quality rating can constitute an important signal to coffee growers to make decisions
that integrate them more efficiently in the coffee supply chain. Specialty coffee

competitions and auctions, such as the Cup of Excellence (CofE) and Q, present the

65

important feature of disclosing market information to all buyers and sellers. This is
particularly important in the case of growers who typically do not know such detailed
quality and price information about their produce (Ponte). An appropriate price signal is
one that conveys the market informatibn —consumers’ preferences and valuations— so as
to allow producers to make decisions about the allocation of their scarce resources in a
way that maximizes their profits. The consequences of not receiving an informative price
signal from the primary demand are to the disadvantage of farmers”. The market fails to
provide an incentive to farmers to make the necessary investments to maintain and
enhance the high quality that is necessary to sustain the whole specialty business with the
consequential risk of the collapse of the supply chain due to an ever declining quality. In
addition, new innovations at the producer level are improbable since returns on
investment and risk taking is not rewarded. Without continual reinvestment, farmers miss
the growth opportunity from the increased value of high quality and differentiated coffee
markets. All of the above implies that in the medium to long run there is a decline in
farmers’ income and increasing poverty if a clear price signal is not received.

Most analysis of differentiation of specialty food products utilizes hedonic modeling of
prices. Hedonic analysis has been applied extensively in wine (see for example Combris
et al., 1997; Lecocq and Visser, 2003; Schamel and Anderson, 2003). These studies
analyze wine characteristics that have an individual impact on the price. Authors
distinguish between characteristics that refer to the sensory evaluation by professional

tasters (e.g. aromatic intensity and suppleness) and characteristics that relate to

 

22 Note that even with good information many coffee producers may not have the capacity for a suitable
response due to their limited resources, as well as lack of viable income alternatives in many poor rural

areas (Lewin et al., 2004).

66

production conditions and methods and appear on the wine bottle (e.g. origin, grape
variety, and processing method). Applying hedonic analysis to specialty coffees e-auction
data, Donnet et al., (2007) found that coffees at the procurement level in the supply chain
are differentiated by coffee ratings, rankings, quantity available, and country of origin. In
this paper we present an addition to hedonic analysis that focuses on the vertical
differentiation by quality ratings. While hedonic analysis evaluates the marginal impact
of various variables on price, the cross-entropy measures how strong is correlation
between price and one particular variable, quality rating.

The following section offers a discussion of quality ratings and specifically how it is
determined in specialty coffees. Next, we present the entropy measure and its application
to measuring income inequality which we replicate for measuring product differentiation.
Following that section, we present the empirical model, data, decomposition by quality
rating groups and results. We finalize with a conclusion.

The Specialty Coffee Quality Rating

Quality ratings are measures of the intrinsic quality of food products evaluated through
sensory analysis. Sensory analysis of coffee is called cupping and consists of the
olfaction, gustation, and mouthfeel of the coffee components in the cup (Lingle, 2001).
Cuppers interpret and record the sensations perceived using flavor attribute descriptors,
numeric representations of intensity and additional comments such as “like it/hate it”
(Songer, 2004). Because it is likely that discrepancies arise with the rating of most
products, especially aesthetic ones, two conditions are considered for the validity of
ratings (Benjamin and Podolny, 1999). First that there is a convergent judgment across

independent evaluation, i.e. different experts should somehow agree on the quality of one

67

coffee. Second, the evaluation must derive strictly from the sensory attributes of the
product itself and not from external cues such as origin, producer, or price (Benjamin and
Podolny, 1999).

Quality ratings have been successful a successful marketing strategy. The use of the 100-
point scale gave an important boost to the wine industry during the 1980’ and 1990’s and
wine tasting has become a profession. The success of the 100 point scale over other
scales is attributed to the easiness to convey a straight forward and universally
understood idea of quality (McCoy, 2005). Interested consumers use the quality rating
displayed in store and published buyers’ as a guide for their purchases. Ratings given by
renown experts have proved to be powerful in influencing consumers’ choices and
willingness to pay for wines as it is the case of the ‘wine emperor’ Robert Parker
(McCoy, 2005; Rivlin, 2006). Knowing the influence of experts’ ratings on consumers,
retailers learn about ratings in advance of consumers and utilize this information to stock
highly rated wines (McCoy, 2005).

In coffee, the system of rating coffees in the lOO-point quality rating has been
implemented by individual tasters who write coffee reviews for buyers’ guides, such as
the Coffee Review. The generally accepted cut-off point for specialty coffees is above 80
points. The quality rating is also used in cupping competitions of specialty coffee e-
auctions. Quality ratings are typically determined by blind tasting the roasted and brewed
coffee samples. The coffee aroma, sweetness, acidity and flavor are described. There are
combined terms that refer to several sensory experiences at once, like balance and
complexity. The different categories are added to give a score. The number is

accompanied by a verbal description of the coffee, for example “heavy body, low acidity,

68

evaluation categories: cleanness of cup, sweetness, acidity, mouth-feel, flavor, aftertaste,
balance and overall. Each category scores in a range of 0 to 8 and half points are possible
(e.g., 7%). The individual score in each criterion are added to give a sub-total. If the
coffee presents any defects, i.e. negative or poor such as ‘phenolic’ and ‘ferment’ a
discount is applied giving the raw score. The raw score plus 36 gives the final score. The
maximum attainable for a coffee is 100.

The Cross-Entropy Measure and Applications

Cross-entropy measure with probabilities

The cross entropy measure is based on the concept of entropy in information theory. The
concept was initially developed by Shannon (1948) who proposed entropy as a way of
measuring the value of the information contained in a message. Kullback-Leibler
developed the cross-entropy (CE) which measures the unpredictability of an event E,-
given its prior and posterior probabilities of occurrence p,- and q,- of a set of n mutually

exclusive events E = {Ei,...,En} (Theil and Fiebig, 1984; Golan, 2002; Soofi, 2002). The

cross-entropy I (q : p) is a logarithmic measure of the discrepancy or distance between

the two probability distributions,

" qr
1(qrp)= Z qilog— (1)
i=1 Pi

The interpretation of the cross-entropy as an information quantity depends on the
distributions involved in the context of the problem under consideration. The measure

provides no implication about the content of the message; it may be used as a measure of

information loss or gain due to change from p,- to q,- (Soofi, 1994). The higher the

69

discrepancy the more unexpected the event and therefore the higher the value of the
information contained in the initial message.

The cross-entropy measure, similar to other information measures, has simple
aggregation properties that allow the decomposition of the total entropy into a between-
group information and a within-groups information (Theil and F iebig, 1984). Both prior

and posterior probabilities can be aggregated into groups g so that:

Pg = 2m and Qg = Zn (2)
i e S i e S
Applying the cross-entropy measure 1(q: p)(l) to each group g, the between-group

cross-entropy is

G Q
10(qrp)= ZQg logp—g- (3)
g=l g

The within-group entropy is

. -/Q
1g(q=P)= z —"'—Iogq’——§— (4)

i e Sg Qg Pi / Pg
The total cross-entropy I (q : p) is equal to the between-group cross-entropy Ho plus the

average within-group cross-entropy nggIg,

G
1(q1p)=10(q=p)+ 2 Qg1g(q=p) (5)
g =1
The decomposition of the cross-entropy measure has an informational interpretation in

two stages (Theil and Fiebig, 1984). In the first stage, a message provides the information
that one group of events occurred: the cross-entropy is 10. In the second stage a
subsequent message provides the information that an event falling under this group

occurred: its cross-entropy content is [g Finally, the total information content becomes

the sum ofthe two 10 + Z QgIg.

70

The aggregation properties of information measures allow the comparison between
different distributions or groups of probabilities. Between groups, the higher the value of
the cross-entropy, the higher the information content. The individual between groups
entropy is positive for the higher quality and negative for the lower (Galbraith and
Garcilazo, 2004). The positive contributions are always higher than the negative
contributions, so that the Theil measure is always positive overall (Conceicao and

Ferreira, 2000). To test for statistically significant differences in information measures,
they are related to the log-likelihood ratio test and to Pearson's x2 test (Theil, 1971). For

the purpose of this paper, we consider that the differences are large enough when the
measures differ by more than 50% (Moss, 2005).

The Relationship between Measuring Income inequality and Product
Differentiation
The basic research objective of information measures is to make inferences about a
system from limited partial information about it (Golan, 2002) representing an alternative
to traditional statistical analysis (Soofi, 1994). (Theil, 1967) applied the cross-entropy
measure to the analysis of income inequality. The problem is analogous to the cross-

entropy by using income shares for different population groups instead of probabilities.

Consider a geographic area with n regions. The population share of the l'th group is p,- and

its income share is y,-. The income inequality measure T( y : p) is

. _ " . a
T(y-p)— syllog ,. (6)

i=1 pl

71

The income inequality T measures the expected information of the message that
transforms the population shares into the income shares and can be viewed as a measure
of the inequality among the per capita incomes of the N regions (Theil and F iebig, 1984).
If income distribution was equal, the T measure would be zero. A relatively low number
for T indicates low income inequality and a high number indicates high income
inequality. Most usually, this measure is used to compare groups and not individuals”.
Three types of inequality can be distinguished: 1) total inequality, the overall
comprehensive inequality between individuals, 2) inequality between the groups, the
comparison of the population share with the income share of each group and 3) residual,
the remaining inequality among individuals that is not accounted for by the between-
group inequality (Conceicao and Ferreira, 2000). Our application to product
differentiation by quality rating is similar to the income inequality application. The word
‘inequality’ is comparable to the word ‘differentiation’, and the word ‘product’ to
‘income’. We denominate the cross-entropy measure of product differentiation by D.
When one or more coffees capture higher prices relative to others, the D measure
increases indicating greater differentiation.

Quality Differentiation Measure in Specialty Coffee Procurement

Empirical model
The pi’s, q,~’s, Pi’s, and Q, ’s can be given an interpretation other than probabilities as

long as they qualify as probabilities, i.e. as long as they are nonnegative and add up to 1

(Theil, unknown). Since the share of value and the share of quantity of a coffee in a given

 

2’ The reality of data collection almost always entails some level of aggregation (Conceicao and Ferreira,

2000).

72

total add up to one, we are able to apply the cross-entropy measure to our purpose of
measuring the information contained in prices in different quality rating groups. The
value v of a coffee transaction is given by the product of the price and the quantity,
v = p * q. The unit of prices is dollars per pound ($/lb) and the unit of quantity is pounds

(lb). The value of a transaction relative to the value of all transactions is its share vi:

[’4-
v.=__l_l_ (7)

l n
2px]-
j=1 JJ

The quantity of a transaction relative to the quantity of all transactions is its share q,~*:

q-
qi*=—;i— (8)
Z q j
J' =1
Let v,- be the observed share of value of coffee and let q,- be the corresponding share of

coffee quantity in an individual coffee i. Then the cross-entropy with p,- and q,- of (1)

interpreted as vi’s, q*,~, respectively is:

n Vi
D<v : q) = 2 v.- log—,; (9)
i=1 (It

The problem is therefore analogous to that of the probabilities described above. By
relating the value and the quantity shares as prior and posterior probabilities through the
summation of the logarithm of their ratio, we can measure how much information one
gains when one looks at the value share in relation to the quantity share. D measures the
discrepancy between the distribution of value and the distribution of quantity between

groups of coffees ordered by a sequence of quality rating.

The differentiation measure can be decomposed into differentiation between and within:
D= DB +DW (10)

73

The between-group differentiation is

 

G Vg
DB(v:q)= z Vglog—. (11)
g=1 Qg
k k
Zpiqi - 2611'
where Vg =’_‘:§S_ and qg = renSg
EM; 241'
1:1 i=1

The within—group differentiation is

v- Vi/V
DW(v:q)= Z floquQg.
ieSg g l g

 

(12)

Data
We use the same data set used by Donnet et al (2007). Our data consists of 653 coffees at
the Cup of Excellence and 59 coffees traded at the Q auction (Figure 6). At each auction,
the coffees correspond to different years, different countries and regions within countries.
They are produced from different coffee tree varieties and at different altitudes. All this
information is available to bidders prior to the auction. In addition, potential bidders
receive coffee samples for cupping at their facilities and they would not necessarily rate
the coffees in them in the same way as the CofE juryz". Thus, it is likely that their bidding
behavior and resulting prices do not necessarily reflect the CofE rates. With this note, we

assume that the average of the individual bidders does.

 

2" Thanks to Thomas Oberthur for this comment.

74

 

Pricevs Rating

 

 

 

 

a - ‘ _ .3
-
. I
.
3 r . _ .0
. - _
. -
'8 8 I ' ' r "
C I. I g 8
9- I I . ‘ - VT:
0 I .
U a a - i A A
A
'- I ' A ‘ a a
a
. - “ A A A ‘ ‘ _ (2
2 d In I ‘ ‘ 1‘3“.“ A ‘ a a
g .
a ' f a. A A g A‘ . ”Eff“, ’2“; ‘ AAA "
. ' “ _ .-., _~ Luv-pt ‘ ‘
5A,,“ M'l'fi‘l" "."-.".~,_l4'[ 3 1" :‘Q‘ énl‘ A ‘ ‘
O -l .- '—
I I I I
80 85 90 95
ratrng
‘ CofE I QL

 

 

Figure 6: Quality Rating versus Prices in the CofE and Q Auctions

Decomposition of the Cross-entropy of value and quality rating
Using the aggregation properties of information measures we decompose the total
information contained in prices D Tom; (Figure 7). The first hierarchy is the auction, thus
we have two groups corresponding to the two auctions D CofE and D Q. D CofE represents
the information contained in all prices traded at the CofE and, similarly, D Q is the total
information contained in the Q auction prices. Within each group we sub-group the
sequence ratings in the specialty product category, i.e. 80 to 100, every 2.5 points. This is
done arbitrarily for the purpose of the analysis of this paper. Dividing the 20 points range
by 2.5 yields eight groups. Figure 7 shows the actual groups for which there are
observations in the data set, six groups in the CofE and three groups in the Q. The

number in each group indicates the observations.

75

CofE groups

A

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

19 172 243 120 77 ‘ 22
Q groups
A
15 37 7
30 32.5 85 87.5 90 92.5 95 100

Figure 7: Decomposition of the Differentiation Measure: Groups of Specialty Coffee

Results
Table 7 shows the estimation results of the differentiation measure by quality ratings in
specialty coffees traded at the CofE and Q auctions. The differentiation measures
computes how much is the quality rating connected to price, or how good is the quality
rating to predict price, and therefore constitutes a measure of the degree of differentiation
in a product. Three types of differentiation can be distinguished: 1) total differentiation,
the overall comprehensive differentiation between coffees, 2) inequality between the
groups, represents the contribution of each group to the total differentiation and 3)
residual, the remaining differentiation among coffees that is not accounted for by the

between-group differentiation.

76

 

Table 7: Results of the Quality Differentiation Measure the CofE and Q Specialty

 

Coffee Auctions
E-Auction
Groups CofE Q
D DB DW D DB Dw

Total 0.184424 0.004313

>92.5 0.090580 0.017809

90-925 0.119440 0.023002

87.5-90 0.021130 0.012264

85-875 -0.053280 0.026755 0.003780 0.0002087

82.5-85 -0.074380 0.013280 -0.005860 0.0033178

80-82.5 -0.012500 0.000324 0.002170 0.0006962

 

Interpretation of the Differentiation Measure and Management Implications

To interpret the differentiation measure we go back to questions posed at the beginning of

the paper. The first question is whether there is vertical differentiation across specialty

coffees according to the quality rating. The first thing to note is that differentiation is 40

times higher in the CofE relative to the Q auction (Table 8).

77

Table 8: Degrees of Differentiation by Quality Rating Groups in the Cup of

 

Excellence
Quality Rating Degree of
Group 4 DB differentiation
0.0905 80
>92.5 1
0.1 19440
90-925 1
0.021 130
87.5-90 3
-0.053280
85-875 2
-0.074380
82.5-85 2
-0.012500
80-82.5 3

 

Differentiation means a higher discrepancy between the quantity and value shares of each
group in the total (Figure 8). The upper group accrues 8 percent of the total value with 3
percent of the total quantity. The second group accrues 22 percent of value with 11

percent of quantity. These two groups contribute to the total differentiation the most.

78

 

 

 

 

 

 

 

 

100% __-_. .osl__-- »
0.116
30% ' 0.168
60% — 0.188
40%
20%
0% .02

 

 

 

 

Qg/ Qt Vg/ Vt

 

El 80-82.5 B 82.5-85 El 85-875 Cl 875-90 13 90-92.!) 13 >925

 

 

Figure 8: Product Differentiation in Specialty Coffees: Quantity and Value Shares

of 653 CofE Lots Grouped by Quality Ratings

In the Q auction, the shares of quantity and value are similar in the groups, there is no
discrepancy between the distributions of value and quantity and therefore differentiation

is low relative to the CofE (Figure 9).

79

 

 

 

 

 

 

100%
0.122 0.125
80% I ‘
60%
40%
20% -
0.236 0.238
0% .
Qg/Qt Vg/Vt
:1 80-825 s 82.5-8S l:l 85-875

 

 

 

Figure 9: Product Differentiation in Specialty Coffees: Quantity and Value Shares

of 59 Q Lots Coffees Grouped by Quality Ratings

We can further analyze differentiation degrees across groups in each auction separately
by looking at the between-groups differentiation, DB, and using the 50% rule of
significance. In the CofE, three degrees of differentiation can be distinguished (Table 9).
Differentiation is highest (first degree) in the groups 90 to 92.5 and 92.5 to 95; second
highest in the groups 82.5 to 85 and 85 to 87.5; and third highest in the groups 80 to 82.5
and 87.5 to 90. Buyers distinguish coffees from 90 to 95 more carefully than in any other
group. This group has the highest contribution to the total differentiation measure. Buyers
are less discriminating in the lowest and middle quality groups. A medium degree of
discrimination corresponds to the 82.5 to 87.5 possibly indicating the range for a lower
limit for truly specialty coffee. In the lowest quality group, 80 to 82.5, and 87.5 to 90,
buyers discriminate the least. We may say that buyers are most concern with finding
extraordinary quality. They are concerned secondly with establishing a lower limit for

specialty coffees. And they are least concerned with the coffees in between. With respect

80

to the very lowest group, looking at the data, this is a small group with coffees traded in
the early years of the competition-auctions. It makes sense to not pay much attention to

this group.

Table 9: Degrees of Differentiation by Quality Rating Groups in the Q

 

Quality Rating Degree of
Group D, differentiation
85-87.5 0.003780 2
82.5-85 -0.005860 1
80-82.5 0.002170 2

 

Similarly, we can distinguish degrees of differentiation in the three groups formed by the
quality ratings in the Q auction. Following the 50% significance rule, we find that the
opposite that we find in the CofE. The group with the highest degree of differentiation is
the middle one. This group is below average. It seems that for Q coffees that are either
highest or lowest quality, the quality rating is less related to price than for the middle
group. Since the Q coffees represent a more homogeneous group relative to the CofE (as
our D shows) we may infer that the grading is better calibrated for distinguishing around
average coffees. Another interesting observation from the differentiation-between is that
coffees in the Q auction converge to the average (see (Conceicao and F erreira, 2000).

The answers to the remaining questions regarding 2) how effective is the quality rating as
in index of vertical differentiation, 3) how informative is the price signal to the grower
with respect to the quality rating, and 4) how strong is the quality rating for remunerating
quality, follow from the answer above. The quality rating is a more effective index of
vertical differentiation in the CofE and within this around the highest and lowest quality

coffees. The price signal is 40 times more informative in relation to the quality rating in

81

the CofE versus the Q auction. Finally, the quality rating is a strong index of quality
remuneration in the CofE.

Implications from the results can be determined at all levels in the supply chain.
Increased differentiation means higher value and higher revenue for participants.
Roasters are creating this value and revenue by distinguishing material quality using the
quality rating. Although the rating is the same in different marketing channels, e.g. the
CofE and Q auctions, its effectiveness is not the same. This reflects the more exhaustive
criteria and procedure to evaluate the coffees’ cup attributes that buyers are willing to pay
for and transmit to their specialty customers. This might indicate that buyers regard other
information more importantly in the Q auction, for example, the price of the commodity.
For example, since the Q is a larger volume auction, quality is relaxed relative to the
CofE which is a small volume auction.

Because our analysis is at the procurement level, differentiation means increasing
remuneration to coffee growers relative to quality. Different quality rating groups can be
seen as indicating producers of the value of their actual resources and the returns of
possible investments to upgrade quality to move up (or down) in groups. With this
information, growers can select which chain to participate, and in which category within
a given chain, on the basis of which one remunerates the quality of their coffee more
advantageously.

Conclusion

In this paper, we proposed an index of vertical differentiation in specialty coffees using
cross-entropy analysis in an analogous way to the income inequality index. We illustrated

how the index works and interpreted its values for two e-auctions. The differentiation

82

measure can be interpreted as an index of the information contained in specialty coffee
prices. Higher information chains convey more information so there is more scope for
farmers to make upgrading. Higher information groups reflect a higher effort in quality
definition and more detailed sourcing procedures.

The differentiation measure is a helpful index of the potential rewards from supplying to
buyers in different marketing channels and quality groups (products). Interpreting this
measure in relation to the costs of meeting the quality and participation requirements of
the different segments can help the matching between producers and buyers in a more
efficient way. Sending appropriate price signals (as through e-auctions) and interpreting
them correctly is important for supporting the origin of quality and thus achieving

sustainability of specialty coffee supply chains.

83

CONCLUSION
This dissertation examines coordination at the procurement level in the specialty coffee
supply chain. The research utilizes economics and management tools in the case of e-
auctions.
The first paper shows that prices of specialty coffee are determined by sensory as well as
reputation attributes. Sensory attributes are responsible for the superior sensorial
experience offered by specialty coffees. Reputation attributes denote what is behind this
superior experience. Reputation attributes provide extra information that helps ascertain
complex quality and reduce quality risk. Together, they constitute a signal that enhances
coordination of the supply chain.
The second paper illustrates the role of auctions in the design of the roaster’s
procurement strategy to complement supply relationships and adjust the procurement
strategy during the planning season. The impact of e-auctions is more important in more
uncertain environments and for smaller roasters. These results further suggest that e-
auctions greatly facilitate the discovery process.
The third paper proposes an index of product differentiation of specialty coffees
according to their sensory quality as summarized in the quality rating. The application of
the index to data from two e-auctions indicates that the quality rating is used to vertically
differentiate specialty coffees.
The dissertation research elucidates value creation in the specialty coffee industry. It
shows: product differentiation by sensory and reputation attributes; the role in adjusting
procurement in quantity; and product vertical differentiation by sensory attributes. This

study of e-auctions shows the provision of quality incentives, increased information

84

transmission and risk reduction in specialty coffee supply chains. These are the
coordination elements that allow producing and delivering quality and thus enhancing the
value of the product and the remuneration/satisfaction of the supply chain participants.
Roasters are achieving coordination of supply chains with scope for simultaneously
increasing consumers’ satisfaction and producers’ income. This explicit, high level
coordination of the complex coffee supply chain signifies a profound change of the
management role and the potential benefits to all participants, relative to the commodity.
This increased information in the specialty coffee chains, unlike the commodity coffee
chain, means that growers can now learn about the value of their produce to make
strategic decisions to sustain prices and income. In this context, strategic and operations
management decisions become relevant, for example, entering relationships, organizing
auctions, upgrading quality through improved practices, and erecting barriers to entry.
For producers managing coffee farms and cooperatives that are currently in or soon to
enter into high value supply chains there are more factors to consider given the marketing
opportunities offered through the e-auction supply chain.

The propositions on price determinants at the procurement level in this dissertation can
test be tested and expanded as more data becomes available from e-auctions and other
procurement channels. In addition, research on prices at the retail level can show how the
prices of attributes at the procurement level translate into the prices of attributes for the
consumer, clarifying marketing strategies of masters and tracing the price signal from the
consumer. Moreover, analysis at the production level can clarify how specialty coffee

procurement strategies translate into income to the producers. Finally, the concepts,

85

methods and findings can be incorporated into a framework and theory of specialty

products.

86

APPENDIX

A. The Coffee Supply Chain

A coffee supply chain is the sequence of stages involved in delivering a coffee product to
the final consumer. It includes production (cultivation, harvest and initial processing),
roasting, distribution and consumption. Coffee production takes place in coffee estates25
and small farms in approximately 60 countries in tropical and equatorial regions. This
stage entails the planting, growing and harvest of the coffee cherry and initial processing
to separate the beam from the skin and pulp of the cherry. The result of this stage is a
green bean. Since the majority of coffee is consumed in non-producing countries, coffee
is exported to the consuming places largely in the form of green beans. Several firms may
be involved in the international trade of coffee including exporters, importers and
masters. The trade can also take place directly from grower to roaster. Moreover,
growing and roasting can be vertically integrated in ownership and management”. Green
beans need to be processed in order to display its aromatic and flavor characteristics. This
processing is called coffee roasting and consists of the heating of the coffee that

transforms the physical and chemical properties of the green bean into roasted coffee

 

‘5 Estates are medium and large scale individual farms. Size varies fi'om large estates of over 500 ha down
to only a few coffee trees on small holdings. Medium-size farms, between 5 and 30 ha, mainly occur in
Latin America. In about 80% of the countries that have distinct estate and smallholder subsectors among
coffee producers, the smallholdings predominate (de Graaff, 1986).

2" This happens in the case of coffees from production origins with well-established reputation for
outstanding growing conditions such as Kona and Blue Mountain coffees. For example, the roasting firm
Wataru opted for vertically integrating the production of coffee and own a plantation in Jamaica for their

premier coffee Blue Mountain (Wataru & Co., 2004).

87

products. Finally, coffee is ground and brewed for consumption as a hot or cold beverage.
Brewing methods include the use of a dip brewer, an espresso machine, a French press or
a vacuum pot, among others. Coffee may be further processed to make instant coffee,
which only requires adding water to consume.
Coffee growing and processing

Coffee trees are grown in tropical regions since the species requires even temperatures
between 15 to 24°C year round. In order to fall within this temperature range, the altitude
at which coffee is actually cultivated depends on the latitude (de Graaff, 1986). For
example, in the equatorial regions (e.g. Colombia, Mexico), coffee is grown at altitudes
between 1200 and 2000 meters while in the subtropical regions (e.g. Brazil, Jamaica), at
500 to 1200 meters. The ideal precipitation is 1500-2500 mm over a nine month period
with a three month dry season coinciding with the harvest. Propitious soils for coffee are
humus, porous, and with a neutral to slightly acidic pH. The microclimate is also
important and is provided by taller trees which maintain high humidity, moderate
temperatures, prevent direct exposure to sunlight and heavy precipitation and improve
soil conditions.

There are two main botanical species of coffee in commercial trade, Arabica and
Robusta. Arabica is grown at higher altitudes and its cultivation demands great care. The
cultivation of Arabica varieties is analogous to the cultivation of the finest grapes in the
best vineyards terroirs. Robusta, as the name suggests, is a hardier plant, which displays
greater resistance to climate and weather conditions, diseases and heat. In the cup,
Arabica is distinctly milder and more aromatic with fewer sharp and bitter tastes than

Robusta. This is the reason why most specialty coffees come from this species. The

88

Robusta species displays higher caffeine content, which is why Robustas are used for
espresso. The fruit of the coffee plants are the berries which grow in clusters and
typically carry two coffee beans each. Each tree can produce between one to twelve
pounds of coffee per year depending on the soil, the climate, and other factors. Coffee
berries do not ripen uniformly so the harvest has to be done manually for obtaining the
highest quality of homogeneous ripened berries.

The harvested cherries enter the processing, by which coffee beans are obtained. The first
step is the pulping of the cherries, to separate the pulp from the beans, which are covered
by mucilage. The beans covered in the mucilage can then be sent to patios to dry as
pulped natural coffees or can be sent to fermentation tanks for wet processing; the first
method is the most common in Brazil and the second in the other Latin-American
countries. The wet processing is the fermentation of the pulped beans with water forl6-36
hours depending on the water temperature which is related to the altitude. After this, the
beans are dried to 11-12 percent of moisture content in drying patios. This stage is called
the parchment. What remains is the removal of the outer skin by boiling. The product is
the green beans that are the form in which coffee is exported to consuming countries.

Trade and Procurement

Coffee is traded from producing to consuming countries largely in the form of green
coffee beans packed in 60 or 69 kg bags. The largest trading companies of green coffee
worldwide are the Neumann and Volcafe with 29 percent of the market share. These
companies together with Cargill, Esteve, Aron and Man have 50 percent of the green

coffee market (Ponte, 2001). Prices in the trade of Arabica coffees are set with reference

89

to the New York Board of Trade NYBOT, while the London lntemational Financial
Futures and Options Exchange (LIFFE) is the reference market for Robusta coffees.
Coffee is traded on the basis of its type (Arabica or Robusta), primary processing (wet or
dry), country of origin and grade. Grades are determined by defective bean count, bean
size standards, moisture content limits, confirmation of good uniformity, good color and
freshness (Ponte, 2001 ). There are five grades in use: 1) the highest is specialty coffee; 2)
the second is premium coffee; 3) the third represents the quality of coffees that are
acceptable for trade at the New York Board of Trade (N YBOT); 4) grade four is usually
purchased by institutional roasters through private contracts with producers and brokers;
and 5) grade five is often traded by soluble coffee manufactures and institutional
decaffeinators (Barker, 2003). In addition, lower grades are relegated to internal
consumption within the producing countries (ibid.).

Green coffee prices are recovering from a low of 50 cents per pound level in 2001 and
2002 to almost one dollar per pound in 2005 (Figure 10). Green coffee prices are very
volatile due to the crop sensitivity to changes in temperature, rainfall, and diseases.
Additional volatility factors have been added in the 90s which include: the end of the
stabilization mechanisms of the quota system by the lntemational Coffee Association
(ICA) in 1989 which used to stabilize prices; the increased activity in the coffee futures
market which consequently increased speculative activity and thus price fluctuation
(Ponte, 2001); and the adoption of supplier-managed inventory system by roasters which
transferred the supply and quality risk to trading houses, this system is very sensitive to
supply shortages and is prone to trigger panic-buying situations causing major price hikes

(ibid.). This situation is illustrated by the 19905 Brazilian coffees case.

90

 

 

200

 

150

 

 

100*

    

 

 

50

 

 

 

0 TTjTTIFWTIIITIIIIIIIIIIIITIIT

sssssssssssssss
«asssss
{4 \q \s 9 \s \°l \9 {fl {1" {fl \of’ a“ ”c9 «90 "ff

Year

Coffee Retail Prices (cents per lb:

 

 

 

L—x— Colombia —-—— Braa'lj

 

 

Figure 10: Prices paid to growers in exporting countries: December monthly

average (cents per lb) (Arabica). Source: International Coffee Organization.

Roasting
In order to bring out flavor and aroma coffee must be roasted. During roasting green
coffee beans expand in size, change in color and density and display the oils that give
coffee its characteristic taste. As the bean progressively absorbs heat, the color changes
from green to yellow to light brown and then finally to a dark brownish oily color. The
grades of coffee roasting include, progressively: unroasted, light, cinnamon, medium,
high, city, full city, French, and Italian. Lighter roasts showcase the coffee’s origin
flavor, i.e. the flavors created in the bean by the soil and weather conditions in the
location where it was grown. As the beans darken past a certain point the roast flavors

begin to overtake the origin flavors of the bean (The Real Estate Blend, 2006).

91

Roasters are the leading firms in the traditional coffee supply chains”. The concentration
ratio is high at this level in the supply chain with a CR4 of 63 percent in 1998 (Daviron
and Ponte, 2005). The four largest firms are Altria (ex Philip Morris), Nestle, Sara Lee,
and Procter and Gamble. However, fragmentation of the market started to take place as
coffee became more differentiated and new firms entered with innovative business
strategies to capture the demand of coffee drinkers (Roseberry, 1996; Ponte, 2001).
While in 1969 there were approximately 20 small roasters, by 1979 the number had
doubled, and in 1989 it was 385; since 1990, the number of small roasters opening
annually increased by increments of 100 with an estimation of 1,400 by 1999 (Lingle,
2001b). According to the SCAA, this fragmentation was the most fundamental change in
the coffee industry (ibid.).
Retailing

Coffee retailing can be divided into sales of coffee-per-pound or retail store sales and
coffee-per-cup or food services sales. Supermarkets are the most important retail outlet
for coffee for in-home consumption (Ponte, 2001). Retailers became more interested in
the higher margins obtained from differentiated coffee vs. the commodity and dedicate

more shelf space, advertising and promotion to the product (Duff, 2002). In addition to

 

‘7 Leading firm, in the global value chain framework of political economy, refers to the companies that play
the ‘driving’ or ‘lead’ role in the structuring and management of the supply chain. Lead firms govern the
supply chain by explicitly coordinating “globally dispersed and organizationally fragmented production and
distribution networks” (Gereffi, 2005). The analogous concept in the logistics and distribution of the

business literature is ‘channel captain’, i.e. the firm or economic stage that controls the channel.

92

supermarkets, coffee-per-pound can be purchased in specialty stores and increasingly on-
line directly from roasters.

Foodservice is gaining ground on retail stores for the last decade. Coffee sales at retail
stores in the US accounted for 46 percent in 2005 compared to 55 percent in 1994
(Ganes-Chase, 2006). Starbucks is the leader in selling coffee-per-cup with 5,393
company-operated coffeehouses and 2,952 licensed locations in the US (Starbucks,
2006). Diedrich Coffee is the number two company with about 200 company-owned and
franchised locations in California and other states (Hoover's, 2006). Peet’s and Tully’s
have 120 and 110 stores respectively. Peet’s is considered to be the oldest specialty
coffee company in the US with a history of 40 years while Starbucks is considered to be
its child of age 35. Coffee-per-cup can also be purchased in coffee shops, convenience
stores and supermarkets.

Retail coffee prices follow the pattern of the green coffee prices (Figure 13). Major hikes
at the retail level in the 1990s reflect the supply shortage from Brazil, the major

producing country.

93

 

 

Coffee Retail and Green Prices

 

 

 

 

400 . k 1
300 [AKA AW/ Ike/w

 

 

 

 

 

Coffee Retail Prices (cents

O rrrrrrrrrrrr T rrr fl rr IF‘FT r rrrrr
b % Q ’1, br ‘0 ‘1: Q ’1; b b 95 Q W b
’\ ’\ ‘b °o °o % °o @ ® ®

,9 \°’ @Gé’f‘f’xfilxfiixfilélfils s *1.

Year

 

—a— Retail —0— Green Colombia + Green BraleL

 

 

 

 

Figure 13: Retail prices in the US: December monthly averages (cents per lb). Source:
lntemational Coffee Organization.
Consumption

The coffee market can be judged as being in its maturity phase since consumption is
stagnant or even declining with income on a per capita basis (Ponte, 2001; Beaver, 2006).
Coffee entered a downward spiral of declining consumption and market share 1055 to
other beverages since the early 19605. In the early 20005, consumption was roughly a half
of its highest which occurred in 1962 (Figure 14). Although total consumption is
stagnant, there is a change in consumer preferences from quantity to quality which is
shown in Figure 4 by the appearance of the category ‘specialty coffee’ in 2000.
Consistent with the general trend of food upgrading people are drinking less but a better

quality coffee.

94

 

US Coffee Consurrption

 

 

 

 

 

 

 

 

Cups per capita per day

0.5 i. -. mint—L, . .

—‘-’_J . ‘l ‘

 

 

 

 

0 IIIIIIIIIIrIIIIIIIITIIIIIIIIIITIIIIITITijI

\°l°\°l°\°l°\°l°\°’ sf \9 \°’ \°’ \°’ $3 {939 {95$

 

 

 

L+ Regular -—I— Soluble Decaffeinated Specialty + Total L

 

 

Figure 14: US per Capita Coffee Consumption. Source: Horticultural and Tropical
Products Division, PAS/USDA

Overall, the coffee market is becoming increasingly differentiated and quality is being
distinguished and valued at a higher level. Thus, coffee is considered to be where the
wine industry was 10-15 years ago (Ron Wiegand SCAA presentation cited in Sweet
Marias). Of a total coffee market of $20 billion in the US, differentiated coffees
represented $7.76 and $9.62 billion in 2000 and 2004 respectively, indicating the
increasing market share accruing to non-commodity coffees relative to all coffees (NCA
and SCAA). According to the National Coffee Association (NCA), the share of adult
population in the US drinking differentiated quality coffee increased from 9 percent in

2000 to 15 percent in 2005.

95

B. Definition of Specialty Coffee

Specialty coffee refers to the highest quality green coffee beans roasted to their greatest
flavor potential by true craftspeople and then properly brewed to well-established
standards (SCAA definition)”. Thus, a specialty coffee is defined in the cup. The process
of assessing the quality of a coffee in the cup is called cupping. Coffee cupping is the
analogous to wine tasting and consists of the examination and evaluation of the coffee
attributes in the cup including the olfaction, gustation, and mouthfeel of the coffee
(Lingle, 2001). The SCAA developed a cupping procedure and description for adoption
in the specialty industry to maintain quality in the specialty industry". The evaluation
gives each coffee a quality rating which is a 100 point scale that summarizes the coffee
quality in the cup, also called sensory quality. By standardizing the cupping, the SCAA
intend to prevent the loss of meaning of the term ‘specialty’ from the auditioning of milk,
water, syrups and others (Daviron and Ponte p. 155)”.

In Green Coffee Beans

At the core of the specialty coffee concept is that special geographic microclimates and
cultivation conditions and methods produce beans that will give unique coffee flavor

profiles (Ponte, 2002). The potential for flavor and aroma resides in the precursor

 

‘8 The term ‘specialty coffee’ was first coined by Erna Knutsen, of Knutsen Coffee Ltd., in a speech to the
delegates of an international coffee conference in Montreuil, France, in 1978.

29 The cupping form designed by Howell is the one used in the industry with modifications by different
finns/organizations. For example, the CofE uses a most exhaustive one, but they are all similar and rate the
coffees in a 100 scale according to quality in the cup.

3° The original idea was that the SCAA would certify coffee lots complying with their cupping standards

(Ponte and Daviron 2005 p. 155).

96

compounds in the green coffee beans (Mabbett, 2006). Also the chemicals responsible for
the color and body in the cup are locked away in green coffee (ibid). Coffee is a naturally
differentiated product. The attributes of a cup of coffee result from a particular
combination of chemical compounds which is unique to the conditions and methods of
cultivation, to the coffee origin. These conditions govern the coffee sensory attributes, i.e.
its organoleptic signature, stimulating the taste buds on the tongue and sensory cells in
the nose (Davids, 2002; Arvidson, 2003; Mabbett, 2006).

The Specialty Coffee Association of America (SCAA) continues to define specialty in
this context, as Don Holly remarks (Holly, 2005): "This is the craft of the specialty coffee
industry that had been slowly evolving during the twenty-year period preceding this
speech”. The key in the definition of specialty is that it must taste notably good; the
premium grade (second grade) also has no primary defects, but does not necessarily
evidence a distinctive character in the cup. This is the definition that is increasingly being
accepted in the industry and emphasizes that specialty stands for an outstanding cup
quality trait, and not for process and credence attributes such as organic, fair-trade, and so
on (Oberthur, 2006).

In Roasted Coffee

1n roasted coffee, most experts agree freshness is part of the definition for specialty.
Freshness is critical because if the coffee is not highly aromatic it no longer deserves to
be called specialty (Holly, 2005):”. Supply chain players, however, do not show
consensus about how fresh roasted coffee must be. Retail roasters tend to favor a shorter

optimum life-span (3 to 7 day) for best taste while wholesale roaster contend that the

 

3 ' "While the. tongue may tell, only the nose knows for certain" (Holly 2005).

97

optimum period of time can be greatly expanded into weeks or even months with
appropriate packaging technology. The question remains whether there is such
technology that can retain the aromatic properties of a specialty coffee (ibid.).
In the Coffee Beverage

Finally, specialty coffee requires a correct preparation: the right recipe of coffee to water,
the right grind suited to the method and the coffee's physical characteristics, and the
proper water temperature and contact time (Holly, 2005). A good preparation of a
specialty coffee can be done by different methods. ‘Barista’ is the term used to refer to
experts and professionals of the preparation of espresso-based coffee drinks. The
meaning of a barista is expanded within certain circles, to include what could be called a
coffee sommelier, i.e. a professional who is highly skilled in coffee preparation, with a
comprehensive understanding of coffee, coffee blends, espresso, quality, coffee varieties,
roast degree, espresso equipment, maintenance, latte art, etc.

It follows from this definition that delivering a cupfiof specing coffee entgils new wm

of doirm things at each level of the supply chem

98

C. The Emergence of Specialty Coffee
Changes in Roasting and Retailing

The gradual rise to dominance of industrial capital did not mean that the consumer
market was very price-sensitive and susceptible to new coffee product lines and
advertising. The expansion of large roasting companies with their superior technology,
greater efficiency, more reliable and cheaper product, and marketing sophistication did
not expand coffee consumption as one might have expected. On the contrary, just as
consolidation was beginning to occur, per capita consumption in the United States was
stagnating. It would take forty years for US. consumers to again reach the thirteen pound
per capita level of 1902.

The development of specialty coffee is a response of firms to the decline of coffee quality
and consumption in the 19705 and 19805. This decline is associated with increasing
production of Robusta and higher yielding Arabica varieties and the technologies and
trends in roasting towards using lower quality green beans. Coffee firms were used to
selling large quantities of relatively homogeneous and undifferentiated blends of
mediocre to poor quality (Ponte, 2001), a “bizarre” strategy for a naturally differentiated
product (Roseberry 1996). In Lingle’s words, the coffee industry was “a caffeine delivery
system that the consumer would drink regardless of how it tasted” (Beaver, 2006). The
move toward quality was not initiated by the large coffee firms but by small regional
roasters who developed new sources of supply and new modes and networks of
distribution (Roseberry, 1996). As noted by Ted Lingle, “it was the core belief of a small
group of people that felt they could fulfill the needs of consumers for a good-tasting

product” that led to the growth of the specialty industry (Beaver, 2006). Roasters and

99

retailers started to invest in product innovation and segmentation in order to increase
value added as shown in the explosive growth of specialty coffee retail chains (Ponte,
2001). The idea of quality was not well received by the traditional coffee industry and it
took 30 years for the specialty industry to reach the critical mass to crystallize the
changes (Beaver, 2006).

The changes at the roasting level include the roast of varietals and origin identified
coffees to an appropriate roast to display their aromatic and flavor attributes. The changes
at the retail level are what characterizes the specialty coffee experience, which includes
the: coffees’ nice aromas and tastes; variety in coffee origin and form of preparation; and
the ambience of the coffee shop or the specialty retail or the supermarket aisle. Starbucks
is the company that led the popularization of these characteristics at both roasting and
retail and thus became the specialty coffee icon. As Howard Schultz put it “We created
an industry that did not exist” (Pelley, 2006).

Changes in Consumption

At the same time that roaster and retailers were developing an alternative to poor quality,
consumption preferences were changing from a past characterized by mass consumption
(Roseberry, 1996). These changes in preference were accompanied and guided by the
promotional and educational activities led by the Coffee Development Group (CDG)
(Chwat, 1991; Beaver, 2006). As stated by Ted Lingle, the demand for Starbucks
products might never have happened without the work of the CDG (Beaver 2006). The
CDG got funds from the lntemational Coffee Organization (ICO) and set up 300
coffeehouses on college campuses to introduce the youth to a menu of coffee beverages

in the 19705, i.e. long before the specialty coffee explosion. This promotion work set the

100

stage for Starbucks to successfully introduce its products (such as skinny latte and ice
blended mocha) and turn an industry from bean-based to beverage-based (ibid). Thus, the
development of the early specialty coffee industry was the result of a two-way exchange
of information on taste, roasting Styles and origins of coffee that allowed masters to
experiment with new roasting methods, packaging and procurement (Daviron and Ponte,
2005). Today nearly 90 percent of consumers understand that coffee is a differentiated
product, like wine, cheese or ice cream, and that quality coffee demands premium prices
(Beaver 2006).
Results in Product Differentiation

The specialty changes in the coffee supply chain resulted in a much differentiated market
in which coffee firms distinguish their product from the competition to enhance value and
revenue and move away from the commodity aspect of the product. Changing approaches
from quantity to quality allowed the opportunity to make consumers aware of varietals
and origins, roasts and methods of preparation. The industry is now believed to be
coming of age as a true specialty beverage (Davids, 2006). Davids notes that the seasonal
offerings of small lots of exceptional coffees are a sure sign of this. Specialty coffee
differentiation is increasingly being described using wine-influenced language like
"Special Reserve," "Limited Edition," "Roastmaster's Reserve" and so on. These are the
cream of the specialty coffee and represent coffees from a single crop and single farm or
hillside. Different from mass coffees, these are sold not on the basis of consistency or
brand, but as an opportunity to experience the flavor associated with a unique moment in

time and space and the dedication of a single farmer or group of farmers (ibid.).

101

D. Detailed derivations and proofs of propositions for second paper
Derivation of Equation (4): Substituting the optimal procurement quantity (3) into the

expected profit function (2) and consolidating yields

2 2
E773- 1(p+s—cR)2 D2 +1(cR—v) _Q ——1—(p—v)_D_2
2177 2 (p+s— —v) 2 (p+s—v) 2

+(cR —v)(p+s— cR)DD_ 1 —2 _
(p+s- v) 2 SD} KR.

Replacing the expressions of the maximum and minimum demands,

Eg”_ 2_l_77{;(4(0p+_+:-C:_;:(#2+,72+2w7)+ %(C__(p:SV):)(# 2+,72_2/177)
-—(p-V)(#’+o‘-2#o)

J 22:22:74 .1- ; wen-Kw}

 

Distributing i
2o

En3=lM£+lM ,rlw-s)’
4 (p+S-v) 77 4 (P+S—V) 2 (p+s-v)
+_]_ (CR—V)2 fi:+_l_ (CR-V)2 __]_ (cR—v)2
4(P+S-V) 77 4(p+s—v) 2(p+s-v)

“llp-V)%-l(P-v)o+;(r v)#+ +;(c" v)(p+S-CR)#—2

 

 

4 4 (p+s—v) 77
1(cR-v)(p+s-cR) 1 #2 1

—— —— ——— —— -K
2 (p+s-v) 4S 77 4577 5,11 R

Consolidating,

102

Err” =

[(p+s-cR)-(c, —v)]2 77

(p+s—v)

+
sl—

1
4 (p+s-v)

In this expression the first and the fifth term cancel to yield Equation (4).

Derivation of Equation (9): Expected profits are

D
Err(xR)= jn(xR,D)f(D)dD.
D

Given the uniform distribution of demand with probability density function-'— and profit
in Equation (8),

xR
E7r*(xR)=% flpD+v(xR —D)—cRxR —KR]a'D
D

+ Jj‘[(p-CR)JCR-S(D-JCR)-KRldD
m
D
+ IKPTCRVCR +(p-CAXD-XR)_KR _KAPD'

x

Integrating and consolidating yields (9).

Derivation of Equation (10): The first order conditions for maximizing (9) is

£$Q=_(c,_v)x,+(c,—v)x,p+(c, —cR)D+K, =0.
R

103

Derivation of Equation (11): Substituting (10) into (9) gives,

 

E,._ l (c. era +114. wry + (c. —c.c.x -v)5_12
2—177 2 (c, -v) 2 (c,-v) (c,-—v)

—2 (c v,,)
+— z—(P CA)D -2(P- v)D2 +KA(::::)_

(C(:,- R)— T"
K ___D __ ————K D -K.
i . (c,—v) ’2(c, _v)‘2p+s c, . R

 

Replacing the expressions of the maximum and minimum demands by the distribution

parameters yields

. 1 1(c -c) 1
E :— ___A___R__ 2
fl 277L2 (c,—v) 012+772+ #77)+ 2

+(CA (:RXCIS— V)(}12_ ”2),“; (p_cAXu2+nz+2#n)

--(p- 001 +o -2#o)+K 2-—:—L:_ ——:—L(#- o)

iii—l": “7 2””)

 

 

Distributing i ,
2o

104

1(cA-CR)_+_12(CA-CR) ,7 1(6) —CR)2

4 (CA “‘0 ’7 4 (CA "v) 2(CA -v)
+1121—1») 2_ 1(cR—vr,,_1(cR-v)2

4 (CA -V) ’7 4 (CA —V) 2 (CA "‘0
1(cA-CRXCR*V)F1(CA-CRXCR_V)

+2 (cl-v) 77 2 (2,1 -v)

+-‘-(p-cl)fii+l(p—21>2+—<p—cl)2
(p- )----(p-V)I7+- (2 v2)

.1

4

_1_(CA -cR) Afl+_l_(cR-V)KA_+_1___C__)RKA(CA_C __1_(cR—v) A
2(c -v 2: 2(c1-v) 2c( -v) 2(c1—v)
1 _K___f 1 Kf 1 ,u_ 1

+— +—————K
4—17(—___-cA--v)+ 2p+s— CA 2

E72":

 

 

 

 

 

+

Consolidating and rearranging this gives Equation (14).

Derivation of Equation (12):

 

 

 

 

~ D _EA—‘+xk._l_)
p{A}=p{Dzz}=1_:‘-=1_P”’_€A
D—Q D-l_)
KA +(CA -CR)E+(CR ”V)_Q+KA _
_1_p+s-cA _ (CA-V) '—
D-Q
=1_ KA _(CA —cR)—D-_ (CR-V)D KA +D(CA-v)

 

 

277(P+S- CA)— 277(cA-v) “277(cA-V) 277(CA‘V) 277(CA'V)
l—"‘(c -cR)_ K (p+s— v)

(CA "V) -"277(P+S CAXCA_ V)

=(CA _V)"(CA —cR)— KA(p+S V)

 

 

277(p+s-CA)
=(C _v)_ KA(p+S—V)
R 277(P+S-CA)

105

Detailed Proof of Proposition 3: The relationship procurement quantity in the
benchmark strategy increases with uncertainty if the marginal profit from the relationship

is larger than its marginal cost.:

=—(p+s c023 +___ (CR —v)D _(p+s—cR)_ (CR-v) >0“.

dry (p+s- v) (p+s— v)-Q” (p+s-v) (p+s—v)

 

p+s—cR > cR -v.
The relationship procurement quantity in the optimal strategy using e-auctions increases
with uncertainty if the relationship cost advantage is larger than the relationship marginal

cost:

de. =CA( CR)D +_(___CR— V)D _(c CA _CR)_(CR —V)
—(cA :CV)D +(cA -V)Q” (CA TV)

>0 ich —cR >cR—v.

 

Note: 5,, = land _D_,, = -1 . The result of the proposition follows.

Detailed Proof of Proposition 4: The expected profits in the benchmark strategy

increase with size by half of the relationship marginal profit including the shortage costs:

dEflB=l(p+S-CR)2_1(CR-V)2 +l
d,u 2 (p+s—v) 2(p+s— v)+ 2

 

—(p— s— v): :(p+s— CR): :MPR >0.

The expected profits in the combined strategy increase with size by half of the

relationship marginal profit not including the shortage costs:

 

 

dEIt' __ 1 (CA -cR)2 -(cR -v)2 1 l _1
-§ (CA_V) +§(P-CA)+§(P—V)-§(P—CR)>0-

all!

The result of Proposition 4 follows.

Detailed Proof of Proposition 4: The change in expected profit in the benchmark

strategy is

106

 

dEflN __]_[(p+S—CR)-(CR -V)]2 __ +S-v =_(P+S-CRXCR"V)<
7-4 (p+s—v) 4(p ) (p+s-v) 0.

The change in expected profits in the combined strategy is

9E1:__[(cA-CR)-(cR—v)]2_ 1 2 (P'l'S-V) < .
c177 — (CA—v) 4r72KA [(cA—vXp+s-CA)] O

The change in the difference in expected profits is the difference between the two:

[(c,,-cR)-(CR“’)Iz - l K {(6 _(ff;::-)-C )]

dAEzt _ dErr' _ dEn” _
dn dn dn (CA-V) 4772 "
+(P+S-CRXcR-V)
(p+s-v)

Solving and rearranging, the condition for the derivative to be positive is:

(p+S-CRXCR-v)_[(c/l-CR)-(cR-v)]2 > 1 2 (p+s-v)
(pH—v) (CA—v) 47,2 K2 [(CwapH—cn}

which is the same condition for a positive probability of attending the auction.

107

REFERENCES

Ali, Hela Hadj, Sebastien Lecocq and Michael Visser (2005). The impact of gurus: Parker
grades and en primeur wine prices. INRA Research Unit Working Paper 0507.
Laboratoire d'Economie Appliquee.

Arvidson, Ed. (2003, 01.20.03). Coffee Varietals: A Taste Experience. Bellis_simo Coffee
Newsletter. From http://espresso 1 01 .com/newsletter/3 97.

Ashenfelter, Orley, David Ashmore and Robert Lalonde (1995). Bordeaux Wine Vintage
Quality and the Weather. Chance 8(4): 7-14.

 

Bacon, Christopher (2005). "Confronting the Coffee Crisis: Can Fair Trade, Organic, and
Specialty Coffees Reduce Small-Scale Farmer Vulnerability in Northern
Nicaragua?" World Development 33(3): 497-511.

Barker, Robert (2003). "Green Coffee Grading Part 1." Tea & Coffee Trade
Journal(November).

Barker, Robert (2004). "Green Coffee Grading Part 11." Tea & Coffee Trade Journal
179(5).

Bates, Robert (1997). Institutions and Development. Coffee Institutions and Economic
Development. Diego Pizano and Jose Chalarca. Santa Fe de Bogota,
Commemorative Symposium of the 70th Anniversary of the National Federation
of Coffee Growers of Colombia (1927-1997).

Beaver, Jennifer (2006). "Charting the Coffee Course." The Specialty Coffee Chronicle.
Beaver, Jennifer (2006). Charting the Coffee Course. The Specialty Coffee Chronicle.

Benjamin, Beth A. and Joel M. Podolny (1999). Status, Quality, and Social Order in the
California Wine Industry. Administrative Science Otgrterlv 44(3): 563-589.

Boccaletti, Stefano (1999). Signaling quality of food products with Designations of
Origin: Advantages and limitations. IAMA Agribusiness Forum, Florence, Italy.

Bunduchi, Raluca (2005). Business relationships in Internet based electronic markets: the
role of goodwill trust and transaction costs. Information SLstems Journal 15(4):
321-341.

Bunn, MD. (1993). Taxonomy of buying decision approaches. Journal of Maglgeting 57:
38-56.

California Libraries Association Conference (2005). Getting Started with Opening a
Coffee Cafe, Kiosk or Cart in Your Library. From http://www.cla-

108

net.org/included/docs/handout25.pdf#search=%22getting%205tarted%20with%20
opening%20a%20coffee%20cafe%22

Carriquiry, Miguel and Bruce A. Babcock (2004). Reputation, Quality Observability, and
the Choice of Quality Assurance Systems. Center for AgriculturaLand Rural
Development (CARD) Working Pamr 04-WP 373 (September).

Castle, Timothy J. (2006). Assessing and Reflecting on Estate Coffees. Tea & Coffee
Trade Journal 178(3).

Cattani, Kyle D., Wendell G Gilland and Jayashankar M. Swaminathan (2004).
Coordinating Traditional and Internet Supply Chains. Handbook of Quantitative
Supply Chain Analysis: Modeling in the E-Business Era. D. Wu D. Simchi-Levi,
and M. Shen, Kluwer Academic Publishers. lntemational Series in Operations
Research and Management Science.

Chang, Guangshu, J ichao Xu, Jianfeng Yang, Jinfeng He and Li Zhang (2005).
Comparative analysis on traditional procurement and e-procurement strategy of
supply chain. Journal of Systems Science and Systems Engineering 14(4).

Chwat, John (1991). Coffee promotion and education in the US: The end of an era.
(interview with Mike Levin formerly of the lntemational Coffee Organization's
Coffee Development Group). Tea & Coffee Trade Journal. February.

Clarence-Smith, William Gervase and Steven Topik (2003). The Global Coffee Economy
in Africa, Asia, and Latin America, 1500-1989. New York, Cambridge University
Press.

Combris, Pierre, Sebastien Lecocq and Michael Wsser (I 997). Estimation of a Hedonic
Price Equation for Bordeaux Wine: Does Quality Matter? Economic Journal 107:
390-402.

Combris, Pierre, Sebastien Lecocq, and Michael Visser (2000). Estimation of a hedonic
price equation for Burgundy wine. Applied Economics 32(8): 961 - 967.

Conceicao, Pedro and Pedro Ferreira (2000). The Young Person's Guide to the Theil
Index: Suggesting Intuitive Interpretations and Exploring Analytical Applications.
Massachusetts Institute of Technology (MIT) - UTIP Working Paper No. 14.

Cox, Daniel C. (2005). CO] Presentation to the ICO Executive Board.
Davids, Kenneth (2004). Emerging Origins. The Coffee Review (March).

Davids, Kenneth (2006). "Now or Never: Special Reserve Coffees." Coffee
Review(August).

Davids, Kenneth. (2002, 01.23.02). "The Coffee Review: Building a Culture of
Connoisseurship Around Coffee." from http://espressolOl .com/newsletter/420.

109

Daviron, Benoit and Stefano Ponte (2005). The coffee paradox: Global markets,
commodity trade and the elusive promise of development. London and New York,
Zed Books.

de Figueiredo, John M. (2000). Finding sustainable profitability in electronic commerce.
Sloan Management Review 41(4): 41-52.

de Graaff, J. (1986). The Economics of Coffee. Wageningen, Centre for Agricultural
Publishing and Documentation (Pudoc).

Donnet, Laura, Dave Weatherspoon and John Hoehn (2007). What Adds Value in
Specialty Coffee? Managerial Implications from Hedonic Price Analysis of
Central and South American E-Auctions. lntemational Food and Aggibusiness
Management Review IFAMAR 10(3). .

Duff, Mike (2002). Margins are brewing in the coffee aisle: in-store cafes offer visability,
while grocery shelves increase flavored varieties. DSN Retailing Today.

Elmaghraby, Wedad J. (2000). Supply Contract Competition and Sourcing Policies.
Manufacturing & Service Operations Management 2(4): 350-371.

Fitter, Robert and Raphael Kaplinsky (2001). Who gains from product rents as the coffee
market becomes more differentiated? A value-chain analysis. IDS Bulletin 32(3):
69-82.

Foley, Karen (2004). Deconstructing Relationship Coffee: How This Emerging Business
Model Fits Into the Industry. The Specflltv Coffee Chronicle.

Foss, Nicolai J. (2003). The strategic management and transaction cost nexus: past
debates, central questions, and future research possibilities. Strategic Organization
1(2): 139-69.

Foss, Nicolai J ., Henrik Lando and Steen Thomsen (1999). The Theory of the Firm,
Institute of lntemational Business, Aarhus Business School. Copyright 1999
Nicolai J. Foss, Henrik Lando, and Steen Thomsen

Frank, Robert H. and Philip J. Cook (1995). The Winner-Take-All Society. Penguin
Book, New York.

Galbraith, James and Enrique Garcilazo (2004). The Mathematics and Logic of The Theil
Statistic. Measuring inequality: A practical workshop on theory and technique.
Costa Rica, The University of Texas Inequality Project.

Ganes-Chase, Judith (2006). State of the Industry Report: Global Supply and Demand
Outlook. Charlotte, Specialty Coffee Association of America.

Gereffi, Gary, John Humphrey and Timothy Sturgeon (2005). The governance of global
value chains. Review of lntemational Political Economy 12(1): 78-104.

 

110

Golan, Amos (2002). Information and Entropy Econometrics-Editor's View. Journal of
Econometrics 107(1-2): 1-15.

Grey, W., T. Olvason and D. Shi (2005). The Role of E-marketplaces in The Relationship-
based Supply Chain. IBM Systems Journal 44(1): 109-123.

Griswold, David (2000). Relationship Coffees. fldgett's Coffee eJoumal 1(29).

Halvorson, Robert and Henry Pollakowski (1981). Choice of Functional Form for
Hedonic Price Equations. Journal of Urban Economics 10 (1): 37-49.

Hillside Agricultural Program (2002). Specialty Coffee. Product Bulletin (3).

Holly, Don. (2005). The Definition of Specialty Coffee. From
http://kaffee.netfirms.com/Coffee/SCAASpecCofDef.html.

Hoover's. (2006). Diedrich Coffee, 1nc.F rom http://www.hoovers.com/diedrich-coffee/--
ID_5 1942--/free-co-factsheet.xhtml.

Howell, George. (2005). Cupping Form. Retrieved August 18, 2005, from
http://www.cupofexcellence.org/DesktopDefault.aspx?tabid=1 54.

Institute of Development Studies (2001). The cost of coffee: how consumer education
could help coffee farmers fight the price war. IDS News.

lntelligentsia. (2004). lntelligentsia Direct Trade. From
http://www.intelligentsiacoffee.com/origin/directtrade.

Jeitschko, Thomas D. (1998). Learning in Sequential Auctions. Southern Economic
Journal 65 (1): 98—112.

Jeitschko, Thomas D. (1999). Equilibrium Price Paths in Sequential Auctions with
Stochastic Supply. Economics Letters 64(1): 67—72.

Kaplinsky, R. and R. Fitter (2004). Technology and globalisation: who gains when
commodities are de-commodified?" lntemational Journal of Technology and
Globalisation 1(1): 5-28.

Khouja, M. J. (1999). The single-period (newsvendor) problem: literature review and
suggestions for future research. Omega 27(5): 537-553.

Knox, Kevin (2006). Price, Quality and Value: Questioning Competition Coffees. m
Coffee Review. December.

Knox, Kevin and Julie Sheldon Huffaker (1997). Coffee Basics: A Quick and Easy
Guide, John Wiley & Sons.

Kreps, David and Robert Wilson (1982). Reputation and Imperfect Information. Journal
of Economic Them 27(2): 253-279.

111

Lancaster, K. J. (1966). A New Approach to Consumer Theory. Journal of Political
Economy 74(2): 132-157.

Landon, Stuart and Constance Smith (1997). The Use of Quality and Reputation
Indicators by Consumers: The Case of Bordeaux Wine. Jouraal of Consumer
Policy 20: 289 - 323.

Lecocq, Sebastien and Michael Visser (2003). What Determines Wine Prices: Objective
vs. Sensory Characteristics. The Handbook of Wine Economics. 0. Ashenfelter
and V. Ginsburgh, Princeton University Press.

Lee, Hyon Jung. (2006, 07.20.06). Most Expensive Coffee. F orbes.com.

Lewin, Bryan, Daniele Giovannucci and Panos Varangis (March 2004). Coffee Markets
New Paradigms in Global Supply and Demand. Agriculture and Rural
Development Discussion Paper 3. Agriculture and Rural Development
Department. Washington, DC 20433, The lntemational Bank for Reconstruction
and Development (The World Bank).

Lingle, Ted R. (2001). The Coffee Cuppers Hand Book. Long Beach, SCAA.

Lingle, Ted R. (2001b, 11.15.01). What a Difference a Decade Makes. from
http://espressolOl .com/newsletter/422.

Mabbett, Terry (2006). "Essence of Coffee Roasting." Tea & Coffee(August/September).

Mabbett, Terry (2006). Essence of Coffee Roasting. m & Coffee Trade Journal
(August/ September).

Mathieu, Paula (1999). Economic Citizenship and the Rhetoric of Gourmet Coffee.
Rhetoric Review 18(1): 112-127.

McCoy, Elin (2005). The Emperor of Wine: The Rise of Robert M. Parker, Jr. and the
Reign of American Taste, Harper Collins.

Milgrom, Paul and John Roberts (1992). Economics, Organization & Management,
Prentice Hall, Englewood Cliffs, New Jersey 07632.

Mocha Joe's (2006). Mocha Joe's Roaster Notes. Brattleboro Vermont.

Moss, Charles (2005). Using entropy to measure information content in prices. L.
Donnet. East Lansing.

Moss, Charles B., Troy G Schmitz, Albert Kagan and Andrew Schmitz (2003).
Institutional Economics and the Emergence of E-Commerce in Agribusiness.

Journal of Agribusiness 21(1): 83-101.

112

Moulton, Kirby, Anthony Spawton and Michel Bourqui (2001). Introduction: Consumer
Behavior and Marketing Strategies. Successful Wine Marketing. Kirby Moulton
and by James Lapsley.

Nelson, Phillip (1970). Information and consumer behavior. Journal of Political Economy
(78): 311-329.

Oberhiinsli, Herbert. (2001, 18 May). The cost of coffee - how consumer education could
help coffee farmers fight the price war." IDS News.

Oberthur, Thomas (2006). Comments on L. Donnet Dissertation Proposal. L. D. East
Lansing.

Oczkowski, Edward (1994). A Hedonic Price Function for Australian Premium Table
Wine. AustrzLian JoumaL of Agricultural Economics 38(1): 93-110.

Oczkowski, Edward (2001). Hedonic Wine Price Functions and Measurement Error. T_he
Economic Record 77(239): 374-382.

Peleg, Barchi, Hau Lee and Warren Hausman (2002). Short-Term E-Procurement
Strategies vs. Long-Tenn Contracts. Production and Operations Management
11(4): 458 (22 pages).

Pelley, Scott (2006). Howard Schultz: The Star Of Starbucks. CBS News.

Piana, Valentino. (2003). Product Differentiation. Economics Web Institute. Retrieved
June, 2007, from
http://www.economicswebinstitute.org/glossary/product.htm#vertica1.

Ponte, Stefano (2001). The 'latte revolution"? winners and losers in the restructuring of
the global coffee marketing chain. Copenhagen, Centre for Development
Research (CDR).

Ponte, Stefano (2002). Standards, trade and equity: Lessons from the specialty coffee
industry. CDR Working Paper 02.13. Centre for Development Research,
Copenhagen.

Ponte, Stefano and Peter Gibbon (2005). Quality standards, conventions and the
governance of global value chains. Economyand Society 34(1): 1-31.

Pyke, David F. and M. Eric Johnson (2003). Sourcing Strategy and Supplier
Relationships: Alliances vs. eProcurement. The Practice of Supply Chain
Management. H. Lee C. Billington, J. Neale, T. Harrison, Kluwer Publishers: 77 -
89.

Rivlin, Gary (2006). Wine Ratings Might Not Pass the Sobriety Test. The New York
Times.

113

Rodriguez, German. (2002). Generalized Linear Models: Transforming the Data. From
http ://data.princeton.edu/wwsS09/stata7/c251 0.htm1.

Roseberry, William (1996). The Rise of Yuppie Coffees and the Reimagination of Class
in the United States. American Anthropologist 98(4): 762-775.

Rosen, Sherwin (1974). Hedonic Prices and Implicit Markets: Product Differentiation in
Pure Competition. Journal of Politicflconomy 82(1): 34-55.

Rosen, Sherwin (1981). The Economics of Superstars. The American Economic Review
71(5): 845-858.

Samper, Mario K. (2003). The historical construction of quality and competitiveness: a
preliminary discussion of coffee commodity chains. The Global Coffee Economy
in Africa, Asia, and Latin America, 1500-1989. W. Clarence-Smith and S. Topic
eds. Cambridge University Press.

SCAA, website. (2007, May 29). "What is Specialty Coffee?" from
http://www.scaa.org/what_is_specialty_coffee.asp.

Schamel, Gunter and Kym Anderson (2003). Wine Quality and Varietal, Regional and
Winery Reputations: Hedonic Prices for Australia and New Zealand. Ih_e
Economic Record 79: 357-369.

Schultz, Henry (1938). The Theory and Measurement of Demand, University of Chicago
Press.

Shapiro, Car] (1983). Premiums for High Quality Products as Returns to Reputations.
The (marterlv Jouml of Economics 98(4): 659-680.

Songer, Paul (2004). Sensory Analysis of Coffee. Scales, Scores, Forms and Perception.
The Specialty Coffee Chronicle: 14-15.

Soofi, E. S., and Retzer, J .J . (2002). Information Indices: Unification and Applications.
Journal of Econometrics 107: 17-40.

Soofi, Ehsan (1994). Capturing the intangible concept of information. Jouml of the
American Statistical Association 89(428): 1243-1254.

Starbucks. (2004). C.A.F.E. Practices Overview. From
http://www. sc scertified.com/csrpurchasing/docs/CAFEPracticesOverviewl 12204.
pdf.

Starbucks. (2006). "Company Fact Sheet." from

http://www.starbucks.com/aboutus/Company_Fact_Sheet_Feb06.pdf#search=%2
25tarbucks%20company%20fact%203heet%22.

114

Swaminathan, JM and SR Tayur (2003). Models for Supply Chains in E-Business.
Management Science 40(10): 1387-1406.

Sweet Maria's (2006). Colombia CoE -Hacienda La Virginia.

The Real Estate Blend. (2006). "Dark Roast vs Light Roast." from
http://www.howardawalker.com/blog/?p=1 3.

Theil, Henri (1967). Economics and Information Theory. Amsterdan, North Holland
Publising Company.

Theil, Henri (1971). Principles of Econometrics. New York, Wiley.

Theil, Henri and Denzil Fiebig (1984). Exploiting Continuity, Maximum Entropy
Estimation of Continuous Distributions. Cambridge, Massachusetts, Ballinger
Publishing Company.

Topik, Steven (2003). The Integration of the World Coffee Market. The Global Coffee
Economy in Africa, Asia, and Latin America, 1500-1989. W. a. S. T. Clarence-
Smith. Cambridge, University Press.

Troncoso, Javier L. and Maria J. Aguirre (2006). Short communication. Price
determinants of Chilean wines in the US market: a hedonic approach. Spanish
Journal of Agricultural Reseagg 4(2): 124-129.

United Nations, Conference on Trade and Development (2003). E-commerce and
Development Report 2003. UNCTAD/SDTE/ECB/2003/1. United Nations
Publication, United Nations: 228.

Wataru & Co. (2004). "Company profile." from
http://www.wataru.co.jp/navi/navi__0l_main_e.html.

Waugh, Frederick (1928). Quality Factors Influencing Vegetable Prices. Journal of Farm
Economics 10(2): 185-196.

Williamson, Oliver (1999). Strategy research: governance and competence perspective.
Strategic Management Journal 20(12): 1087-108.

Wolfstetter, Elmar (1996). Auctions: An lntroduction. Journal of Economic Surveys 10:
367-420.

115

   
  

IIIIIIIIIII

1l1111111111111111111111111111111111

1293 02956 1