Available via license: CC BY 4.0
Content may be subject to copyright.
Journal of Agricultural and Applied Economics: page 1 of 20
©
2018 The Author(s). This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original
work is properly cited.
doi:10.1017/aae.2018.5
WHAT EXPLAINS SPECIALTY COFFEE
QUALITY SCORES AND PRICES: A CASE
STUDY FROM THE CUP OF EXCELLENCE
PROGRAM
TOGO M. TRAORE∗
Department of Agricultural Economics and Rural Sociology, Auburn University, Auburn, Alabama
NORBERT L.W. WILSON
Friedman School of Nutrition, Tufts University, Boston, Massachusetts
DEACUE FIELDS III
Department of Agricultural Economics and Rural Sociology, Auburn University, Auburn, Alabama
Abstract. This study investigates the effects of material and symbolic quality
attributes on the Cup of Excellence specialty coffee quality scores and prices. The
estimates from the quality score equations suggest that material attributes are
important determinants, but symbolic attributes have a greater explanatory
power. The hedonic price estimations show that specialty coffee prices are mainly
determined by symbolic attributes and market conditions such as the number of
coffees in the auction. Overall, the study reveals that fruity, oral, sweet, spice,
and sour acid are cuppers’ and buyers’ most favorite coffee avors and aromas.
Keywords. Coffee Taster’s Flavor Wheel, Cup of Excellence, hedonic model,
quality attributes, specialty coffee, truncated regression
JEL Classications. C24, D44, Q13
1. Introduction
Coffee is one of the most popular drinks and valuable agricultural commodities
traded in the world. Coffee provides a livelihood for almost 125 million
people around the world, generating cash returns in subsistence economies and
providing employment to both men and women living in rural areas (Fairtrade
Foundation, 2012). From 1962 to 1989, the coffee market was regulated
by the International Coffee Agreement (ICA). The ICA was a collection of
agreements that set the production quotas and governed the quality standards
for most coffee-producing countries. After the disintegration of the ICA and the
∗Corresponding author’s e-mail: tmt0016@auburn.edu
1
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
2TOGO M. TRAORE ET AL.
liberalization of the coffee market, global coffee production increased leading to
a decline in producer prices and quality of coffees (Ponte, 2002). As a reaction to
the decline in coffee quality, specialty coffee was born. Specialty coffee is dened
as coffee grown in special and ideal climates, with distinctive taste and avor,
and with little to no defects. To be classied as specialty coffee, a coffee needs to
obtain a quality score of 80 or higher on a 100-point scale from the coffee-tasting
process (Specialty Coffee Association of America [SCAA], 2016).
The specialty coffee market is growing rapidly in many countries, and the
United States has the most developed specialty coffee market, followed by Europe
and Asia (SCAA, 2016). In the United States, for example, specialty coffee has
increased its market share from 1% to 25% over the last 20 years, and the
percentage of adults drinking specialty coffee daily has increased from 9% in
1999 to 34% in 2014 (SCAA, 2016). The growing demand for specialty coffee
is attributable to consumers’ awareness regarding issues of quality, taste, health,
environment, equity, and fair wages (Bacon, 2005).
Quality standards and quality control are key aspects of the development
of the specialty coffee market. Specialty coffee companies often distinguish
themselves from mainstream coffee companies because of their strict quality
requirements. Quality control from the green beans to the roasting method
helps bring the best coffee avors and aromas, which are the main sensory
components experienced by coffee drinkers. Coffee quality can be assessed
using three attributes: material, symbolic, and in-person service attributes
(Daviron and Ponte, 2005). Material attributes result from physical, chemical, or
biological processes that create specic characteristics that can be measured using
human senses (e.g., taste, smell, vision, hearing, or touch). Symbolic attributes
are based on reputations, trademarks, geographic origins, and sustainability
practices, whereas in-person service attributes are similar to customer service.
In-person service attributes are the result of the human interaction between
producers/retailers and consumers and involve effective as well as affective work
from the producer/retailer to deliver a good-quality product and gain consumers’
trust; in return, consumers are willing to pay high premium prices (Daviron and
Ponte, 2005).
Much of the economic literature pertaining to specialty coffee has focused
on determining the value buyers place on symbolic attributes. Donnet,
Weatherspoon, and Hoehn (2008), Teuber and Herrmann (2012), Wilson et al.
(2012), and Wilson and Wilson (2014) mentioned quality score, ranking, country
of origin, coffee tree variety, altitude, and farm size as key factors explaining
specialty coffee price formation. Although these previous studies extensively used
the Cup of Excellence (CoE) data set to predict specialty coffee prices, none
assessed how specialty coffee material attributes such as avors, aromas, body
characteristics, mouthfeel, and aftertaste inuence specialty coffee quality scores
and purchasing prices.
Therefore, the main objective of this study is to measure the impact of
material as well as symbolic attributes on specialty coffee quality scores and
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 3
purchasing prices using the CoE data set from 2004 to 2015. Specically, this
study aims at determining which material attributes yield high-quality scores
and premium prices. To the best of our knowledge, this is the rst study
conducted and reported on the effects of material attributes (avors, aromas,
body characteristics, mouthfeel, and aftertaste) on specialty coffee quality scores
and prices. Therefore, this article adds to the existing literature by providing
benchmark information concerning coffee tasters’ and buyers’ preferences for
specialty coffee material attributes.
2. Theories and Hypotheses
2.1. Theory on Coffee Quality Attributes
Specialty coffee roasters and buyers depend on high-quality beans and are
generally willing to pay premium prices for those beans. According to Daviron
and Ponte (2005), coffee quality can be assessed either based on material,
symbolic, or in-person service attributes. Material quality attributes of a product
are attributes embedded in or intrinsic to the product and exist regardless
of the identity of sellers and buyers. They result from physical, chemical, or
biological processes that create specic characteristics. These characteristics
can be measured using human senses (taste, smell, vision, hearing, or touch)
or by using sophisticated devices such as spectrographs (Daviron and Ponte,
2005). In the CoE competition, material quality attributes are measured through
“cupping,” a process through which highly trained experts estimate aroma,
taste, and avor according to the grading system developed by the SCAA. The
grading system consists of evaluating the presence of material attributes such as
aroma, avor, aftertaste, acidity, body, balance, clean cup, defects, mouthfeel, and
sweetness of each coffee before an average quality score is attributed. Through
this system, cuppers use “cupping notes” to record information about material
attributes of a coffee (for the cupping form, see Figure A1 in the Appendix). In
addition to the cupping notes, cuppers provide a quality score for each coffee,
and generally, the higher the quality grade, the higher the price and vice versa.
Symbolic quality attributes are based on trademarks, geographic origins, and
sustainability practices (Daviron and Ponte, 2005). Trademarks and geographic
origin are similar in many ways as they enable consumers or buyers to
differentiate products, reduce asymmetries of information, and create value.
Sustainable practices are practices that allow for ethical production and
consumption (Fairtrade Foundation, 2012). Many organizations such as the
Fairtrade Foundation, the Rainforest Alliance, and UTZ work with specialty
coffee producers to ensure they follow their standards in terms of chemical usage
and other environmental friendly practices. Consequently, these coffees receive
certications that generally translate into premium prices in the market.
In-person service quality attributes are like customer service and are the
result of the interaction between producers/retailers and consumers. They involve
effective and affective work from the producer or retailer to get attention or
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
4TOGO M. TRAORE ET AL.
trust from consumers, who in return are willing to pay premium prices (Daviron
and Ponte, 2005). In the specialty coffee industry, in-person service quality
attributes can take place between producers and green coffee buyers or between
roasters and nal consumers. These attributes include the quality of the product
delivered to the buyer or nal consumer and the human interaction between
consumers and the producer/barista (e.g., a barista calling consumers by their
name or remembering their favorite drink). The CoE program does not collect
information about in-person quality attributes; therefore, the analyses conducted
in this study are based on material and symbolic attributes only.
2.2. The Model
Since Waugh published “Quality Factors Affecting Vegetables Prices” in 1928
and the seminal work of Rosen (1974), hedonic price modeling has been widely
used to relate the price and attributes of various agricultural, food, real estate,
and environmental products. Theoretically, the hedonic price method consists of
measuring the contribution of the attributes of a product to its price. The analysis
involves modeling the price of the product as a function of its attributes to
estimate the marginal money value of each attribute. Consequently, a regression is
used to predict the quality score and price of coffees based on various attributes:
p(z)=(a1,a2,...,an),(1)
where pis the price of product zand (a1,a2,...,an) represents a set of
attributes of product z. The implicit or hedonic price is dened as follows:
∂p
∂a1
=pi(a1,a2,...,an).(2)
In the case of specialty coffee, the hedonic price method enables us to
determine the impact of material as well as symbolic attributes on specialty
coffee quality scores or prices. That is, through hedonic price analysis, one can
determine the price or the value buyers attach to the intrinsic (material) and the
symbolic attributes of each coffee (Wilson and Wilson, 2014).
An industry comparable to the specialty coffee industry where the use of
hedonic technique is well established is the wine industry. The wine industry
uses a tasting system similar to the coffee cupping process to assess material and
symbolic attributes of wines, and at the end of the tasting process, expert tasters
assign a quality grade to each wine. Material attributes and quality grades are
usually displayed on each bottle of wine to reduce asymmetries of information.
Generally, the price of each bottle of wine is correlated to the quality grade (the
higher the quality grade, the higher the price, and vice versa). Several studies have
been conducted using the hedonic method to assess factors explaining differences
in the price of wines from numerous regions/countries of the world. Combris,
Lecocq, and Visser (1997) applied the hedonic price method to Bordeaux wines,
Oczkowski (2001) used the same technique to study the prices of Australian
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 5
premium wines, Schamel and Anderson (2003) estimated hedonic price functions
for premium wines from Australia and New Zealand, and Lecocq and Visser
(2006) applied the hedonic technique to Bordeaux and Burgundy wines, two of
the most important French wine regions.
As stated before, most studies conducted in the wine industry use the hedonic
price model to explain the effects of material and symbolic attributes on wine
quality grades and prices. Therefore, this study adopts the same technique to
determine the effects of material and symbolic attributes on specialty coffee
quality scores and prices.
2.3. Hypotheses
As suggested by the literature, a product quality can be determined by three
attributes (material, symbolic, and in-person service). In the context of the
CoE program, only two quality attributes are measured—namely, material and
symbolic attributes. Therefore, these two attributes are focused on throughout
the rest of the article. Theoretically, specialty coffee quality scores should not be
affected by symbolic attributes because of the blind tasting process. However,
we predict that specialty coffee prices will be explained by both material and
symbolic attributes. Therefore, the following hypotheses are tested:
H1: Specialty coffee quality score is based solely on material attributes, which
means that symbolic attributes have no inuence on quality score.1
H2: Specialty coffee purchasing price is based on both material and symbolic
attributes.
3. Data
Data for the present study include 2,123 observations from the CoE e-auctions
that took place in 11 countries (Bolivia, Brazil, Burundi, Colombia, Costa Rica,
El Salvador, Guatemala, Honduras, Mexico, Nicaragua, and Rwanda) over the
period 2004 to 2015. The CoE hosts at least one competition per year in
most participating countries, and the competition is open to any coffee grower
free of charge. Each participating coffee in the CoE competition is tasted at
least ve times by a local and international jury panel. During the evaluation,
cuppers record a set of material attribute descriptors, such as avor, aroma,
aftertaste, acidity, body shape, balance, cleanness of the cup, defects, mouthfeel,
and sweetness, and they provide an average quality score for each coffee sample.
1 Hypothesis (H1) comes from the fact that the CoE uses a blind tasting, which means that besides
the country where the competition is taking place, cuppers or tasters have no information about the
characteristics of the coffees (tree variety, processing method, etc.). Therefore, we do not expect those
attributes to inuence quality score.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
6TOGO M. TRAORE ET AL.
Only coffees with an average score of 84 or higher2out of 100 receive the CoE
Award and can participate in the national auction. Buyers interested in the CoE
specialty coffees can either buy them directly at the live auction or buy samples
to do their own tasting before buying the coffees at the live auction (ACE/CoE,
2016).
The CoE live auction is an eBay-style auction where buyers bid against each
other for ownership of the winning coffees, and the highest bid wins (Wilson
and Wilson, 2014). The CoE auction prices are on average 4.5 times higher
than the International Coffee Organization (ICO) composite price (Wilson and
Wilson, 2014). Coffee growers receive more than 80% of the nal price, and the
remaining amount is a commission paid to the in-country (the country where
the CoE competition is taking place) organization committee to help run the
program (ACE/CoE, 2016).
The data set includes information on the nal auction price of each coffee
excluding shipping costs, rank of the coffees, number of coffees in the market,
farm data (quantity of coffee bags per lot, certication, coffee tree varietals,
altitude, and processing method), and name and origin of buyers.
Material attributes for each coffee participating at the CoE are recorded in
the cupping notes (for cupping form, see Figure 1A in the Appendix). Cuppers
have used a variety of words to describe the sensory properties of coffees through
the years. A total of 112 different avor/aroma descriptors were used to describe
coffees’ avors and aromas in the data set. To reduce the number of descriptors,
the Coffee Taster’s Flavor Wheel was used (see Figure A2 in the Appendix).
Created by the SCAA using the lexicon developed by World Coffee Research,
the Coffee Taster’s Flavor Wheel is the largest piece of research on coffee avor
(SCAA, 2016). The 112 avor/aroma descriptors are classied into 9 groups per
the SCAA’s Coffee Taster’s Flavor Wheel. The most recurrent coffee body types,
mouthfeel, and aftertaste descriptors are used, and dummy variables are created
to indicate whether the cup is clean and balanced.
4. Estimation Procedures
This study adopts a two-equation estimation procedure to explain how material
and symbolic attributes affect the CoE specialty coffee quality scores and prices.
In the rst step, the method estimates the quality score equation, because price
depends mainly on quality score:
Qi=M
iβ+εi,(3a)
2 Since 2016, the Alliance for Coffee Excellence (ACE) has raised the CoE minimum score to 86 and
temporary scaled back the competition to only ve countries to restructure the program and adopt a new
cupping evaluation.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 7
where Qiis the quality score, Mirepresents the vectors of material attributes
(avors/aromas, cup type, body characteristics, mouthfeel, and aftertaste) that
inuence coffee quality score, and εiis an error term. Equation (3a) is also run
with the incorporation of symbolic attributes in order to assess their impact on
coffee quality score:
Qi=M
iα+R,
iγ+μi,(3b)
where Riis the vector of symbolic variables. In each case, these variables are
dummies taking the value 1 if the coffee has the attribute and 0 otherwise, or
continuous variables for altitude, number of coffees in the market, or lot size
(see Table 1 for the data summary). Variable Miis dened as in equation (3a),
and μiis an error term. Theoretically, the symbolic attributes should not have
any impact on a coffee quality score. Therefore, equation (3b) is estimated to
assess the effects of symbolic attributes such as country of origin, certication,
or number of coffees in the market on specialty coffee quality scores.
To participate in the CoE live auction, a coffee must obtain a quality score
of 84 or higher.3This causes a truncation of quality score, and the estimation
of equations (3a) and (3b) is therefore conducted using a truncated maximum
likelihood regression model with 84 as a truncation point.
Once the quality score is determined, the second step involves estimating the
price equation as follows:
LnPricei=Q
iϕ+Q2
i+R
iη+ϑi,(4a)
where LnPriceiis the logarithm of the price of coffee i(i=1 to 2,123).
All prices are in 2010 prices based on the producer price index. Qiand Riare
dened as discussed previously, and ϑiis the error term. The use of the logarithm
is to normalize price. Equation (4a) is just an updated version of Wilson and
Wilson (2014) and estimated for comparison purposes. To fulll our objective,
equation (4a) is estimated by replacing quality score by its determinants that
are material attributes to assess their direct impact on coffee prices and avoid
multicollinearity:
LnPricei=M
iσ+R
iζ+ψi,(4b)
where LnPricei,Mi,andRiare dened as discussed previously, and ψiis the
residual.
Recall that any coffee participating in the CoE live auction has a score of 84 or
higher; thus, the distribution of price is incidentally truncated, taking the form:
Pi=P∗
iwhen Q ≥84
unobserved when Q <84 .(5)
3 Ibid.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
8TOGO M. TRAORE ET AL.
Tabl e 1. Summary Statistics
Standard
Variable N Mean Deviation Minimum Maximum
Market characteristics
Price (2010 US$/lb.) 2,123 7.4013 5.4311 12 80.2
Quality score (84–100) 2,123 87.0883 2.0797 84 95.76
Lot size (70 kg bags) 2,123 30.8208 14.7662 8 145
Number of coffees 2,123 27.4702 5.9887 10 42
Material attributes
Number of descriptors 2,123 14.3826 4.9197 2 28
Green vegetative avor 2,123 0.1398 0.3469 0 1
Other avor 2,123 0.0276 0.1638 0 1
Roasted avor 2,123 0.2258 0.4182 0 1
Spices avor 2,123 0.2424 0.4286 0 1
Nutty/cocoa avor 2,123 0.5671 0.4956 0 1
Sweet avor 2,123 0.8476 0.3595 0 1
Floral avor 2,123 0.5387 0.4986 0 1
Fruity avor 2,123 0.9574 0.2019 0 1
Sour acid 2,123 0.5780 0.4940 0 1
Clean and clear cup 2,123 0.2128 0.4094 0 1
Balance cup 2,123 0.2396 0.4269 0 1
Transparent cup 2,123 0.0750 0.2634 0 1
Creamy body 2,123 0.3167 0.4650 0 1
Big body 2,123 0.1066 0.3087 0 1
Round body 2,123 0.1763 0.3812 0 1
Buttery mouthfeel 2,123 0.1743 0.3794 0 1
Smooth mouthfeel 2,123 0.2412 0.4279 0 1
Juicy mouthfeel 2,123 0.1759 0.3808 0 1
Lingering aftertaste 2,123 0.0677 0.2513 0 1
Long aftertaste 2,123 0.1459 0.3531 0 1
Symbolic attributes
Wet processing 2,123 0.6559 0.3741 0 1
Dry processing 2,123 0.0624 0.2420 0 1
Other processing 2,123 0.2817 0.4499 0 1
Bourbon variety 2,123 0.2323 0.4224 0 1
Catuai variety 2,123 0.1184 0.3231 0 1
Caturra variety 2,123 0.1463 0.3535 0 1
Typica variety 2,123 0.0182 0.1338 0 1
Pacamara variety 2,123 0.0928 0.2902 0 1
Other variety 2,123 0.0799 0.2711 0 1
Mixed variety 2,123 0.2493 0.4327 0 1
Organic certied 2,123 0.0276 0.1638 0 1
Rainforest Alliance certied 2,123 0.0113 0.1060 0 1
Altitude 2,123 1502.66 237.5356 600 2,300
Brazil 2,123 0.1305 0.3369 0 1
Bolivia 2,123 0.0474 0.2126 0 1
Colombia 2,123 0.0981 0.2975 0 1
Costa Rica 2,123 0.0880 0.2833 0 1
El Salvador 2,123 0.1427 0.3498 0 1
Guatemala 2,123 0.0985 0.2981 0 1
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 9
Tabl e 1. Continued
Standard
Variable N Mean Deviation Minimum Maximum
Honduras 2,123 0.1386 0.3456 0 1
Nicaragua 2,123 0.1269 0.3329 0 1
Mexico 2,123 0.0247 0.1553 0 1
Burundi 2,123 0.0340 0.1814 0 1
Rwanda 2,123 0.0705 0.2561 0 1
Buyer’s origin
Asian buyer 2,123 0.4816 0.4998 0 1
North American buyer 2,123 0.0863 0.341 0 1
European buyer 2,123 0.0576 0.2330 0 1
Nordic buyer 2,123 0.0786 0.2692 0 1
Other buyer 2,123 0.0089 0.0940 0 1
Group of buyers 2,123 0.2870 0.4524 0 1
Source: Based on Cup of Excellence data, 2004–2015 (Cup of Excellence, 2016b).
Therefore, equations (4a) and (4b) are estimated using a truncated maximum
likelihood method with the lowest price for each auction as the truncation point
(Hausman and Wise, 1977; Maddala, 1983; Wilson and Wilson, 2014).
5. Results
5.1. Descriptive Statistics
Descriptive statistics presented in Table 1 show that coffees participating in the
CoE competition have an average quality score of 87, with 95.76 the highest
score, and a mean purchasing price of $7.4 per pound, which is 4.5 times higher
than the ICO composite price of 2015. The most common avors and aromas
are fruity, found in 95.7% of the coffees; sweet, present in almost 85% of the
coffees; and sour acid, present in 58% of the coffees. The wet processing method
is widely used by farmers, and almost 66% of the coffees are processed through
this method. The data show that Bourbon, Caturra, and Catuai are the most
popular coffee tree varieties, planted respectively by 23%, 14%, and 11% of
the farmers. The average farm is located at an altitude of 1,500 m above sea
level, and very few (4%) of the farms have some type of certication (Organic,
Rainforest Alliance, or UTZ). In general, there are 30 bags of 70 kg/bag that are
supplied by each farmer participating in the CoE live auction, and the average
number of coffees per auction is 27. As stated previously, the CoE program
is held in 11 coffee-producing countries worldwide, but our data set indicates
that coffees are mainly from El Salvador (14.3%), Honduras (13.9%), Brazil
(13%), and Nicaragua (12.7%). The CoE live auction attracts buyers from all
around the world, and the data set indicates that almost 50% of buyers are from
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
10 TOGO M. TRAORE ET AL.
Asia and 29% are groups of buyers not necessarily from the same country or
continent.
5.2. Grading Equations
Equation (3a) estimated through truncated regression shows that the relationship
between coffee quality score and material attributes is positive and signicant
except for other avors and roasted avors that have a signicant but
negative impact on quality score (Table 2). Sweet, oral, fruity, and sour acid
avors/aromas have the largest effect on quality score. For example, having
an extra avor/aroma of fruity, oral, and sweet increases quality score by
3.3, 1.3, and 1.01 points, respectively. Having a lot of avors and aromas is
good, but a balanced combination of them is rewarded an extra 0.36 quality
score point. Clean and transparent attributes have a signicantly positive effect
on quality score, conrming the expectation that specialty coffees have little
or no defects. Mouthfeel is another important factor affecting coffee quality
score. Smooth and juicy mouthfeel are all positive and signicant at the 1%
level, while a buttery mouthfeel is positive and signicant at the 5% level.
Aftertaste is another attribute valued by cuppers. Long aftertaste is positive and
signicant at the 5% level, but a lingering aftertaste is not signicant. These
latter results suggest that material attributes signicantly inuence coffee quality
scores.
The introduction of symbolic attributes in equation (3b) tremendously
improves the grading equation. Equation (3a) has an Akaike information
criterion (AIC) of 9,661 and a Bayesian information criterion (BIC) of 9,789,
whereas the additional inclusion of the processing method, coffee tree variety,
certication, number of coffees in the market, country of origin, and year
decreases the AIC to 8,264 and the BIC to 8,576. The Pacamara tree variety with
respect to the base tree variety Bourbon, Rainforest Alliance certication versus
no certication, and altitude have a positive and signicant effect on quality
score, whereas the number of coffees in the market has a negative and signicant
effect on quality score. For example, certication is worth almost an extra 1.5
points in quality score, and an extra meter in altitude is worth 0.3 point in quality
score, but a one-unit increase in the number of coffees in the market decreases the
quality score by 0.028 point. The estimated coefcients for all countries of origin
are signicant but negative, which means that compared with Brazil, coffees from
other countries have relatively lower quality scores. Introduced as correction
variables for the quality score over time, the coefcients for competition year
are negative and signicant. This means that compared with the base year 2004,
quality score has been decreasing over the years. This is indeed true because the
highest mean quality score and the highest single score ever recorded in the CoE
program (95.76) were recorded in year 2004. The improvement of the grading
equation (3a) with the inclusion of symbolic attributes suggests their importance
in the grading system.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 11
Tabl e 2. Quality Score Equations Estimation
Estimate Standard Estimate Standard
Var i abl e Equation (3a) Error Equation (3b) Error
Green vegetative 0.7757∗∗∗ (0.1815) 0.6916∗∗∗ (0.1881)
Other avor −0.9366∗∗ (0.4575) −0.8849∗(0.4657)
Roasted −0.4346∗∗∗ (0.1628) −0.3490∗∗ (0.1702)
Spices 0.4832∗∗∗ (0.1521) 0.5535∗∗∗ (0.1588)
Cocoa/nutty 0.0722 (0.1358) 0.2486∗(0.1436)
Sweet 1.0162∗∗∗ (0.2111) 1.0470∗∗∗ (0.2211)
Floral 1.3401∗∗∗ (0.1419) 1.3430∗∗∗ (0.1457)
Fruity 3.3164∗∗∗ (0.5373) 3.1910∗∗∗ (0.5215)
Sour acid 0.8668∗∗∗ (0.1401) 1.0763∗∗∗ (0.1467)
Clean and clear 0.4175∗∗∗ (0.1594) 0.5894∗∗∗ (0.1603)
Balance cup 0.3563∗∗ (0.1539) 0.3990∗∗ (0.1554)
Transparent cup 0.6656∗∗∗ (0.2352) 0.5695∗∗ (0.2448)
Creamy body 0.1038 (0.1431) 0.1666 (0.1466)
Big body 0.1552 (0.2123) 0.0854 (0.2159)
Round body 0.0409 (0.1742) 0.2247 (0.1790)
Buttery mouthfeel 0.3366∗∗∗ (0.1707) 0.4217∗∗ (0.1741)
Smooth mouthfeel 0.5246∗∗∗ (0.1534) 0.5447∗∗∗ (0.1578)
Juicy mouthfeel 0.7535∗∗∗ (0.1671) 0.7098∗∗∗ (0.1748)
Lingering aftertaste 0.2885 (0.2576) 0.3146 (0.2554)
Long aftertaste 0.4037∗∗∗ (0.1804) 0.3853∗∗ (0.1832)
Dry processing 0.3050 (0.3264)
Other processing 0.2092 (0.2562)
Catuai 0.0931 (0.2904)
Caturra −0.0836 (0.3073)
Typica −0.0127 (0.5480)
Pacamara 1.2808∗∗∗ (0.2906)
Other variety 0.3002 (0.3132)
Mixed variety −0.1400 (0.2466)
Organic certied −0.0876 (0.4046)
Rainforest Alliance certied 1.4938∗∗∗ (0.6340)
Altitude 0.2907∗∗∗ (0.0441)
Number coffees −0.0288∗(0.0148)
Bolivia −1.3564∗∗∗ (0.5109)
Colombia −1.4440∗∗∗ (0.4454)
Costa Rica −1.6886∗∗∗ (0.4311)
El Salvador −1.3276∗∗∗ (0.4011)
Guatemala −1.8498∗∗∗ (0.4440)
Honduras −1.7031∗∗∗ (0.4197)
Nicaragua −0.8001∗∗ (0.3919)
Mexico −1.0937∗(0.5781)
Burundi −1.9906∗∗∗ (0.5781)
Rwanda −1.7353∗∗∗ (0.5022)
Year 2005 −1.6320∗∗∗ (0.4407)
Year 2006 −0.8723∗∗ (0.4398)
Year 2007 −1.1875∗∗∗ (0.4189)
Year 2008 −2.0486∗∗∗ (0.4278)
Year 2009 −2.2413∗∗∗ (0.4253)
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
12 TOGO M. TRAORE ET AL.
Tabl e 2. Continued
Estimate Standard Estimate Standard
Var i abl e Equation (3a) Error Equation (3b) Error
Year 2010 −2.1644*** (0.4407)
Year 2011 −2.5143*** (0.4365)
Year 2012 −1.2506*** (0.4216)
Year 2013 −1.4088*** (0.4244)
Year 2014 −1.8526*** (0.4313)
Year 2015 −0.8217 (0.5031)
Sigma 2.5184∗∗∗ (0.0647) 2.4075*** (0.0638)
Intercept 79.9904∗∗∗ (0.6133) 78.8940*** (0.9368)
N 2,123 2,123
Log likelihood −4,809 −4,077
Akaike information criterion 9,661 8,264
Bayesian information criterion 9,789 8,576
Note: Asterisks (∗,∗∗,∗∗∗ ) indicate signicance at the 10% level, 5% level, and 1% level, respectively.
Source: Based on Cup of Excellence data, 2004–2015 (Cup of Excellence, 2016b).
The conclusion of this analysis is that quality score not only depends on
material attributes as expected, but also on symbolic attributes such as the
country of origin, certication,number of coffees in the market, and altitude. This
result shows that symbolic attributes matter in the grading process and means
that coffee quality improves with altitude, certication, and certain tree varieties,
whereas it decreases with the number of coffees.
5.3. Price Equations
Equation (4a), which is an updated version of Wilson and Wilson (2014),
conrms the nonlinear relationship between quality score and price (Table 3).
The coefcients of quality score and the quality score square are both positive
and highly signicant (99% signicance level). The model (equation 4a) predicts
that an additional quality score point4increases price by 18.68%. However,
after 93.10 the effect of a one-unit increase in quality score becomes negative.
As in Wilson and Wilson (2014), this result is supported by examples in the
data set where higher-scoring coffees garnered lower prices than lower-scoring
coffees in another auction. The number of descriptors used to describe coffees is
a signicant predictor of prices, and any additional descriptor is associated with
a 0.74% increase in price.
Ranking matters, and according to the estimates, obtaining the rst place5
gives the highest premium at 133.08% more than coffees not ranked in the top
4 The derivative of ln(Pi) with respect to quality is ∂ln(Pi)
∂Qualityi×1
Pi=0.3096 +2×
(−0.0153 ×Quality).
5 Because the dependent variable is logged, the percentage impact of a dummy variable iis calculated
as e(βi−0.5∗var(βi)) −1, multiplied by 100% (Kennedy, 1981).
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 13
Tabl e 3. Price Equations Estimation
Estimate Standard Estimate Standard
Var i abl e Equation (4a) Error Equation (4b) Error
Grade 0.3092∗∗∗ (0.0203)
Grade2−0.0153∗∗∗ (0.0019)
Descriptors # 0.0074∗∗ (0.0025)
Green vegetative 0.0294 (0.0305)
Other avor −0.0306 (0.0752)
Roasted −0.0248 (0.0277)
Spices 0.0593∗∗∗ (0.0258)
Cocoa/nutty 0.0359 (0.0233)
Sweet 0.1014∗∗∗ (0.0343)
Floral 0.0914∗∗∗ (0.0235)
Fruity 0.3592∗∗∗ (0.0790)
Sour acid 0.0599∗∗ (0.0234)
Clean and clear 0.0372 (0.0259)
Balance 0.0570∗∗ (0.0254)
Transparent 0.0844∗∗ (0.0403)
Creamy body 0.0762∗∗∗ (0.0237)
Big body 0.0421 (0.0345)
Round body 0.0661∗∗ (0.0289)
Butter mouthfeel 0.0297 (0.0284)
Smooth mouthfeel 0.0572∗∗ (0.0258)
Juicy mouthfeel 0.0423 (0.0287)
Lingering aftertaste 0.0089 (0.0427)
Long aftertaste 0.0121 (0.0303)
Dry processing 0.0158 (0.0411) 0.0351 (0.0515)
Other processing 0.0971∗∗∗ (0.0329) 0.1173∗∗∗ (0.0418)
Catuai 0.0239 (0.0385) 0.0085 (0.0484)
Caturra 0.1057∗∗∗ (0.0392) 0.0906∗(0.0497)
Typica −0.1454∗(0.0757) −0.1328 (0.0961)
Pacamara 0.0744∗∗ (0.0358) 0.1247∗∗∗ (0.0448)
Other variety 0.0512 (0.0411) 0.0330 (0.0519)
Mixed variety −0.0105 (0.0318) −0.0135 (0.0406)
Organic certied 0.1522∗∗∗ (0.0480) 0.1457∗∗ (0.0594)
Rainforest Alliance certied −0.0960 (0.0877) −0.0109 (0.1111)
Altitude 0.0210∗∗∗ (0.0056) 0.0319∗∗∗ (0.0070)
Number coffees −0.0076∗∗∗ (0.0019) −0.0047∗(0.0025)
Ln (lot size) −0.4724∗∗∗ (0.0306) −0.5630∗∗∗ (0.0393)
First place 0.8666∗∗∗ (0.0494) 1.4712∗∗∗ (0.0451)
Second place 0.3860∗∗∗ (0.0425) 0.9166∗∗∗ (0.0468)
Third place 0.2563∗∗∗ (0.0391) 0.7432∗∗∗ (0.0458)
Fourth place 0.1591∗∗∗ (0.0376) 0.6263∗∗∗ (0.0452)
Bolivia 0.1960∗∗∗ (0.0666) 0.0942 (0.0833)
Colombia 0.1779∗∗∗ (0.0587) 0.1220 (0.0744)
Costa Rica 0.0965∗(0.0560) 0.0087 (0.0710)
El Salvador 0.2254∗∗∗ (0.0524) 0.1678∗∗ (0.0660)
Guatemala 0.3650∗∗∗ (0.0588) 0.2522∗∗∗ (0.0748)
Honduras 0.0206 (0.0550) 0.0662 (0.0701)
Nicaragua 0.0761 (0.0525) 0.0152 (0.0669)
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
14 TOGO M. TRAORE ET AL.
Tabl e 3. Continued
Estimate Standard Estimate Standard
Var i abl e Equation (4a) Error Equation (4b) Error
Mexico 0.1881∗∗ (0.0739) 0.1258 (0.0928)
Burundi −0.0440 (0.0764) −0.1679∗(0.0981)
Rwanda −0.0518 (0.0669) −0.1613∗(0.0854)
Year 2005 0.0629 (0.0544) 0.0804 (0.0685)
Year 2006 0.1462∗∗∗ (0.0538) 0.0455 (0.0681)
Year 2007 0.2737∗∗∗ (0.0523) 0.1443∗∗ (0.0670)
Year 2008 0.4970∗∗∗ (0.0500) 0.3108∗∗∗ (0.0632)
Year 2009 0.7165∗∗∗ (0.0529) 0.5837∗∗∗ (0.0667)
Year 2010 0.9739∗∗∗ (0.0608) 0.8402∗∗∗ (0.0780)
Year 2011 0.8772∗∗∗ (0.0572) 0.7068∗∗∗ (0.0725)
Year 2012 0.7393∗∗∗ (0.0589) 0.7012∗∗∗ (0.0749)
Year 2013 0.8863∗∗∗ (0.0562) 0.7996∗∗∗ (0.0717)
Year 2014 0.7297∗∗∗ (0.0556) 0.5928∗∗∗ (0.0720)
Year 2015 0.7059∗∗∗ (0.0684) 0.6739∗∗∗ (0.0876)
Asian buyer −0.0932∗∗∗ (0.0331) −0.0995∗∗∗ (0.0414)
European buyer −0.0637 (0.0491) −0.1298∗∗ (0.0625)
Nordic buyer 0.0374 (0.0417) 0.1217∗∗ (0.0522)
Other buyer −0.1078 (0.1318) −0.4189∗∗ (0.1777)
Group of buyers 0.0302 (0.0337) 0.0728∗(0.0419)
Sigma 0.2847∗∗∗ (0.0066) 0.3393∗∗∗ (0.0088)
Intercept 1.2940∗∗∗ (0.1366) 1.6838∗∗∗ (0.1796)
N 2,123 2,123
Log likelihood 951.321 650.3
Akaike information criterion −1,809 −1,171
Bayesian information criterion −1,543 −802.663
Note: Asterisks (∗,∗∗,∗∗∗ ) indicate signicance at the 10% level, 5% level, and 1% level, respectively.
Source: Based on Cup of Excellence data, 2004–2015 (Cup of Excellence, 2016b).
four coffees; obtaining the second, third, or fourth place gives a premium at
44.02%, 26.71%, and 15.06%, respectively. These results suggest that ranking,
which is based on quality score, is perhaps more important than having the
quality score alone (Donnet, Weatherspoon, and Hoehn, 2008; Wilson and
Wilson, 2014).
Variables such as market size (number of coffees in the auction) and processing
methods have signicant impact on coffee prices. When the number of coffees
in the market increases by one for example, price decreases by 0.76%. Results
also show that the effect of other processing methods (semiwashed, pulped
natural, and honey processing) is positive and highly signicant (99% level of
signicance).
Altitude has a positive and highly signicant effect on prices, which suggests
that coffees grown at high altitude are regarded as superior quality coffee
by buyers. Contrary to Donnet, Weatherspoon, and Hoehn (2008), where
Pacamara was the only tree variety with a signicant effect on price, or Teuber
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 15
and Herrmann (2012), where Bourbon and Pacamara were the tree varieties with
signicant effects on coffee prices, our results show that Caturra and Pacamara
tree varieties have a positive and signicant impact on coffee prices, whereas
Typica has a negative and signicant effect on coffee prices. Furthermore,
this study shows that organic certication is valued by buyers, and compared
with coffees with no certication, organic certication increases a coffee price
by 14.95%.
Another attribute valued by buyers is the size of the lots, which represents
the quantity of coffee supplied by farmers. Buyers prefer small lot sizes, and a
unit increase in lot size decreases price by 0.47%. Country of origin signicantly
affects coffee prices as well. For example, coffees from Bolivia, Colombia, Costa
Rica, El Salvador, Guatemala, and Mexico have higher purchasing prices com-
pared with Brazil. Also, it is worth noticing that coffees from African countries
(Burundi and Rwanda) are not statistically different than Brazilian coffees.
Compared with the base year 2004, other years exhibit positive and highly sig-
nicant coefcients suggesting an increase in the price of CoE coffees because of
either their high quality or their popularity in the coffee industry. Origin of buyers
has little effect on coffee price as only the estimate for Asian buyers is signicant
but negative. This means that compared with North American buyers,Asian buy-
ers value CoE coffees less or buy a wide range of coffees (high and low quality).
The biggest contribution and the most signicant difference between this
study and previous studies is the inclusion of material attributes in the price
equations. In equation (4b), quality score is replaced by its determinants—
namely, material attributes to assess their direct effects on prices. Results show
that all aroma and avor attributes have a positive impact on prices except for
other avors (chemical and papery) and roasted avors, which have a negative
but nonsignicant effect. Spices (3.20%), sweets (3.75%), oral (6.87%), fruity
(40.32%), and sour acid (14.78%) have positive and signicant effects on coffee
prices. Results also indicate that having a balanced coffee is rewarded by a 3.10%
increase in price, while transparent and coffees with a smooth mouthfeel are
valued at a 5.09% and 3.12% price premium, respectively. The effect of aftertaste
(lingering and long aftertaste) was negative but not signicant, which probably
means that buyers do not value those sensory attributes. This result might be
because most buyers do not taste the coffees before buying them and therefore
do not put any value on those attributes. Surprisingly, creamy and round body
types are signicant in the price equation. This result is surprising because body
types were not signicant in the grading equations, suggesting that buyers and
cuppers value specialty coffee attributes differently. Because many buyers of the
CoE are roasters, body type is an important quality characteristic in their process,
which is not the case for coffee cuppers.
As in equation (4a), symbolic attributes have the same effect, with tree variety
having little effect, while altitude, lot size, number of coffees in the market, and
rank were all signicant with the same sign and magnitude. The effect of country
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
16 TOGO M. TRAORE ET AL.
of origin was less pronounced, but still present with four countries being highly
signicant. The variable for year was positive and signicant except for the year
2005. Regression with the inclusion of material variables shows all buyers have
signicant estimates. However, only buyers from Nordic countries and groups of
buyers were paying high prices compared with North American buyers.
6. Discussion and Conclusions
The objectives of this study were three-fold: rst, to add to the existing literature
on specialty coffees by estimating factors affecting coffee quality scores; second,
to determine the effects of material attributes on coffee prices; and, third, to
provide coffee professionals with information that allows them to make accurate
investment decisions.
The rst implication of the study is that although specialty coffee quality
scores should in theory be based on material attributes only, symbolic attributes
such as altitude, certication, market size, and country of origin play a major
role in their determination as well. As an indicator of quality and primary
requirement to enter the market, the relation between quality score and symbolic
variables suggests that one can improve the quality score by having a farm
at high altitude, planting a single tree variety such as Pacamara, and having a
certication (Organic or Rainforest Alliance). The increase in quality score pays
off in two ways: at least 18% increases in price for each additional point and
from 15% to 133% increases in price for being ranked in the top four. Results of
the price equation estimations show that buyers favored small quantities; pulped,
semidry, or honey processing methods; and coffee tree varieties such as Caturra
and Pacamara. These are some factors that farmers can take into account in order
to make sound investments and prot-maximizing decisions.
On the buyer side, the implication of our analysis is that one should refer to the
cupping notes before making buying decisions about CoE specialty coffees. This
is because of the fact that high-quality coffee does not necessarily translate into
high grade especially when comparing coffees across countries. That is, a high-
scoring coffee in one country may have the same or lower quality compared with
a coffee from another country with a low score, so the cupping notes can help
differentiate coffees.
For the coffee industry, our analysis revealed cuppers’, buyers’, and therefore
consumers’ preferences for coffee material attributes such as aromas and avors.
Results show that fruity, oral, sweet, spice, and sour acid aromas and avors are
the favorite avors and aromas. Also,the analysis reveals that buyers (consumers)
prefer balanced coffee. The recognition of these material attributes is important
especially for coffee buyers because it allows them to buy high-quality coffees.
Finally, the analysis shows that buyers place more value on symbolic attributes
than material attributes.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 17
References
Alliance for Coffee Excellence/Cup of Excellence (ACE/CoE). Internet site:
http://www.allianceforcoffeeexcellence.org (Accessed May 2, 2016).
Bacon, C. “Confronting the Coffee Crisis: Can Fair Trade, Organic, and Specialty Coffees
Reduce Small-Scale Farmer Vulnerability in Northern Nicaragua?”World Development
33,3(2005):497–511.
Combris, P., S. Lecocq, and M. Visser. “Estimation of a Hedonic Price Equation for Bordeaux
Wine: Does Quality Matter?” Economic Journal 107,441(1997):390–402.
Cup of Excellence (CoE). Internet site: https://allianceforcoffeeexcellence.org/
cup-of-excellence/ (Accessed May 2, 2016a).
Cup of Excellence (CoE). “Competition/Auction Results.” Internet site: https://
allianceforcoffeeexcellence.org/competition-auction-results/ (Accessed June 1, 2016b).
Daviron, B., and S. Ponte. The Coffee Paradox: Global Markets, Commodity Trade, and the
Elusive Promise of Development. New York: Zed Books, 2005.
Donnet, M.L., D.D. Weatherspoon, and J.P. Hoehn. “Price Determinants in Top-Quality e-
Auctioned Specialty Coffees.” Agricultural Economics 38,3(2008): 267–76.
Fairtrade Foundation. Fairtrade and Coffee: Commodity Brieng. London: Fairtrade
Foundation, May 2012.
Hausman, J.A., and D.A. Wise.“Social Experimentation, Truncated Distributions and Efcient
Estimation.” Econometrica 45,4(1977):919–38.
Kennedy, P.E. “Estimation with Correctly Interpreted Dummy Variables in Semilogarithmic
Equations.” American Economic Review 71,4(1981): 801–13.
Lecocq, S. and M. Visser. “What Determines Wine Prices: Objective vs. Sensory
Characteristics.” Journal of Wine Economics 1,1(2006):42–56.
Maddala, G.S. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge:
Cambridge University Press, 1983.
Oczkowski, E. “Hedonic Wine Price Functions and Measurement Error.” Economic Record
77,239(2001):374–82.
Ponte, S. “The ‘Latte Revolution’? Regulation,Markets and Consumption in the Global Coffee
Chain.” World Development 30,7(2002):1099–122.
Rosen, S.M. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure
Competition.” Journal of Political Economy 82,1(1974):34–55.
Schamel, G., and K. Anderson. “Wine Quality and Varietal, Regional and Wine Reputations:
Hedonic Prices for Australia and New-Zealand.” Economic Record 79,246(2003):357–
69.
Specialty Coffee Association of America (SCAA). “ Specialty Coffee Facts and Figures in the
USA.” Internet site: http://www.scaa.org (Accessed August 4, 2016).
Teuber, R., and R. Herrmann. “Towards a Differentiated Modeling of Origin Effects in
Hedonic Analysis: An Application to Auction Prices of Specialty Coffee.” Food Policy
37,6(2012):732–40.
Waugh, F.V. “Quality Factors Inuencing Vegetable Prices.” Journal of Farm Economics
10,2(1928):185–96.
Wilson, A.P., and N.L.W. Wilson. “The Economics of Quality in the Specialty Coffee Industry:
Insights from the Cup of Excellence Auction Programs.” Agricultural Economics
45,S1(2014):91–105.
Wilson, B.R., J.F. Conley, T.M. Harris, and F. Lafone. “New Terrains of Taste: Spatial Analysis
of Price Premiums for Single Origin Coffees in Central America.” Applied Geography
35,1–2(2012):499–507.
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
18 TOGO M. TRAORE ET AL.
Appendix
Table A1. Flavors/Aromas Used to Describe Coffees from the Cup of Excellence Competitions
(data from 2004 to 2015)
Flavors/Aromas
Vegetative Cocoa Dried fruit Sour
Herb Cocoa Raisin Sour
Tdyme Chocolate Currant Acid
Olive Milk chocolate Date Mascarpone
Fresh Dark chocolate Prune Malic
Ripe Brown sugar Dried fruit Tart
Papery/chemical Molasses Other fruit Alcohol/fermented
Woody Syrup Coconut Wine
Cedar Caramelized Cherries Whiskey
Sandalwood Honey Pomegranate Champagne
Leather Maple Pineapple Rum
Bitter Sugar Mango Cognac
Salty Other Sweet Papaya Grand Marnier
Earthy Vanilla Banana
Tobacco Sugar cane Grape
Tobacco Sweet Apple
Pipe Butterscotch Peach
Burnt Marzipan Plum
Smoky Aromatics Guava
Brown Tea F l avo r Plantain
Roast Tea Pear
Cereal Floral Kiwi
Grain Rose Honeydew
Malt Jasmine Melon
Pepper Hibiscus Apricot
Pepper Lavender Pumpkin
Pungent Sassafras Cucumber
Peppercorn Potpourri Tangerine
Brown spice Patchouli Grenadine
Anise Bouquet Nectarine
Licorice Sandalwood Citrus fruit
Nutmeg Flowery Grapefruit
Cardamom Orchid Orange
Cinnamon Floral Mandarin
Clove Berry Lemon
Nutty Blackberry Lime
Peanut Raspberry Citrus
Hazelnut Blueberry Citric
Almond Strawberry Tomato
Nut
Source: Based on Specialty Coffee Association of America avor wheel and Cup of Excellence data, 2004–
2015 (Cup of Excellence, 2016b).
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
What Explains Specialty Coffee Scores and Prices 19
Figure A1. Cup of Excellence Cupping Form (source: Cup of Excellence, 2016a)
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
20 TOGO M. TRAORE ET AL.
Figure A2. Specialty Coffee Association of America (SCAA) Coffee Taster’s Flavor
Wheel (source: SCAA, 2016)
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/aae.2018.5
Downloaded from https://www.cambridge.org/core. IP address: 38.145.98.88, on 13 May 2018 at 17:43:59, subject to the Cambridge Core terms of use, available
