|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
HOBBYISTS' PREFERENCES FOR MARINE ORNAMENTAL FISH: A DISCRETE
CHOICE ANALYSIS OF SOURCE, PRICE, GUARANTEE AND ECOLABELING
LILIANA A. ALENCASTRO
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
UNIVERSITY OF FLORIDA
Liliana A. Alencastro
To my beloved parents Guillermo and Juanita, and brother Guillermo Jr.
First, I thank God, my family and friends for their continuous love and support
throughout this stage of my personal and professional development, away from home.
I extend my gratitude to my committee chair, Dr. Sherry Larkin, and committee
members, Dr. Robert Degner and Dr. Charles Jacoby, for their valuable guidance during
the development of my thesis research. I am also very thankful to Dr. Ronald Ward for
his generous and important contributions to the analysis presented in this document and
to Dr. Ram6n Espinel for his support and encouragement to start and continue with my
Finally I would like to thank all members and fellow graduate students of the Food
and Resource Economics Department for all their help and for making my academic
experience at the University Florida unforgettable.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S ................................................................................................. iv
LIST OF TABLES ................................................... vii
LIST OF FIGURES ............................... ........... ............................ viii
A B ST R A C T ...................................................................................................... ........ .. x
1 IN TR O D U C T IO N ........ .. ......................................... ..........................................1.
The M arine A quarium Industry .............................................. ............................. 1...
An Ecolabeling Program for Live Marine Ornamentals.........................................3...
Problem Statement....................... .. ........... ...............................9
O objectives and H ypotheses.......................................... ......................... ...............9...
Potential Benefits.......................... .. ........ ............... 11
Contents of the Study..... .................................................................................. 11
2 PREFEREN CE M OD ELIN G ....................................... ...................... ................ 13
A alternative V aluation A approaches ........................................................ ................ 13
C h oice M modeling T h eory ............................................................................................ 16
Econom etric Issues in Choice M odeling ............................................... ............... 19
IIA A ssum ption .......................................................................... .... ............... 19
U observed Preference H eterogeneity ........................................... ................ 20
Inclusion of a B ase A alternative ...................................................... ................ 22
Em pirical Applications of Choice M odeling......................................... ................ 24
Form of Choice M odeling R esults......................................................... ................ 27
3 M E T H O D O L O G Y ................................................... ............................................ 29
E xperim ental D esign ................................................. .............. ................ 29
Sam pling C considerations ........................................... ......................... ................ 33
D ata C o llectio n ........................................................................................................... 3 4
Model Specification......................... ........... .........................35
4 SU R V E Y R E SU L T S .................................................................................................. 3 8
Sample Description .......................... ........... ............................... 38
M aroon C low nfi sh R results ......................................... ........................ ................ 51
M odel Specification .. ................................................................ ............... 51
M odel E stim ates ..................................................... ................... .. .... ........ .... 55
Estimating Effects on Probability of Purchase ..............................................58
A attributes effects ................................................ ....... ............ ................. 58
Influence of attitudes on certification preferences ................................ 62
B lue-Faced A ngelfish R results ...................................... ...................... ................ 64
M odel Specification .. ................................................................ ............... 64
M odel E stim ates ..................................................... ................... .. .... ........ .... 66
Estimating Effects on Probability of Purchase...............................................70
Respondent Com m ents to Survey.................. .................................................... 78
G general Im plications ... .. ........................................... ........................ .... ......... 80
5 SUM M ARY AND CONCLU SION S.................................................... ................ 82
A SURVEY INSTRUMENT AND CODING .................................... ..................... 88
B SUMMARY DESCRIPTIVE STATISTICS ............... .............. ..................... 98
C CATEGORIZATION OF INDIVIDUAL COMMENTS ................ ..................... 99
L IST O F R E F E R E N C E S ................................................................................................. 106
BIOGRAPHICAL SKETCH ........................... ............. 110
LIST OF TABLES
1-1 .M A C certified operators ....................................................................... ...............6...
2-1. Overview of empirical discrete choice modeling ..................................................26
3-1. Definition of experimental fish profiles by species...............................................32
4-1. Respondent familiarity with fish species used in market experiments................... 50
4-2. Distribution of responses by species and choice set.............................................50
4-3. Definition of probit model explanatory variables for maroon clownfish choice
a n a ly sis ................................................................................................................. ... 5 3
4-4. Probit model estimates for maroon clownfish......................................................55
4-5. Definition of probit model explanatory variables for the blue-faced angelfish
ch o ice an aly sis.......................................................................................................... 6 5
4-6. Probit models estimates for the blue-faced angelfish ...........................................67
LIST OF FIGURES
4-1. Current saltwater tank capacity reported by respondents......................................39
4-2. Expected change in saltwater tank capacity within the next 2 years
o f resp o n d en ts .......................................................................................................... 3 9
4-3. The highest price ever paid for a single fish by respondents.................................40
4-4. Incidence of environmental and aquarium interests of respondents........................41
4-5. Level of familiarity with the MAC certification program among respondents..........42
4-6. Distribution of responses regarding the importance of reasons to buy a
tan k -b red fi sh ............................................................................................................ 4 3
4-7. Distribution of responses regarding the importance of reasons to buy a MAC
certified d fi sh .............................................................................................................. 4 4
4-8. Association of coral reef protection with MAC certification by MAC
fam iliarity lev el ........................................................................................................ 4 5
4-9. Association of wild stock sustainability with MAC certification by MAC
fam iliarity lev el ....................................................................................................... 4 5
4-10. The perceived market potential of alternative production techniques that can be
used to produce unique marine ornamental species by respondents.....................46
4-11. Distribution of year of birth of respondents ........................................ ................ 47
4-12. Distribution of the highest level of education of respondents...............................47
4-13. Distribution of annual gross income of respondents ...........................................48
4-14. Geographic distribution of respondents....................................................48
4-15. Distribution of respondents by size of the communities where they reside.............49
4-16. Probability that the base hobbyist would buy a maroon clownfish by source.........59
4-17. Probability that the base hobbyist would purchase a maroon clownfish by
certification status .............. .. .............................. ........ .... ............... 60
4-18. Probability that the base hobbyist would purchase a maroon clownfish by
price, certification status, and source .................................................. ................ 61
4-19. Probability that the base hobbyist would purchase a certified wild-caught
maroon clownfish by respondent familiarity with MAC and beliefs regarding
its effective en ess ........................................................................................................ 6 3
4-20. Probability that the base hobbyist would pay a 15%-17% premium for a
certified wild-caught blue-faced angelfish from Indonesia by various fish and
hobbyist attributes .......... .. .............................. ......... ...... ............... 72
4-21. Probability that the base hobbyist would pay a 15-17% premium for a
wild-caught Indonesian blue-faced angelfish with an extended life warranty by
various hobbyist characteristics .......................................................... ................ 75
4-22. Probability that the base hobbyist would prefer a wild-caught certified
Indonesian blue-faced angelfish to an extended survival guarantee at a constant
p ric e ...................................................................................................... ........ .. 7 6
4-23. Probability that the base hobbyist would pay a 35% premium for a certified and
extended warranted blue-faced angelfish wild-caught in Indonesia.....................77
Abstract of Thesis Presented to the Graduate School
of The University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master Of Science
HOBBYISTS' PREFERENCES FOR MARINE ORNAMENTAL FISH: A DISCRETE
CHOICE ANALYSIS OF SOURCE, PRICE, GUARANTEE AND ECOLABELING
Liliana A. Alencastro
Chair: Sherry L. Larkin
Major Department: Food and Resource Economics
This study presents the analysis of preferences for selected fish attributes (including
the use of an ecolabel) by avid marine aquarium hobbyists for two saltwater ornamental
fish and socio-demographic and attitudinal characteristics. The fish attributes included
price, source (wild or tank bred), length of post-sale survival guarantee, and whether the
fish was ecolabelled as certified by the Marine Aquarium Councill (MAC). The fish
were species commonly known to be harvested from areas with a reputation for
ecosystem damage during collection (i.e., the maroon clownfish from the Philippines and
the blue-faced angelfish from Indonesia). Two discrete choice experiments and the
assessment of aquaria characteristics and respondents' opinions, attitudes and
demographics were implemented through an Internet-based questionnaire. Respondents
were solicited during a week in February 2004 from several Internet bulletin boards and
discussion lists to attract hobbyists that were most likely to be familiar with the MAC.
Respondents represent a very homogeneous group, with above average levels of
education, annual income, and concern about coral reefs and wild stocks protection. Half
of them were not familiar with the MAC environmental certification program and a large
majority preferred tank-cultured over any wild caught organism, including a certified
one, in order to avoid harmful ecosystem effects associated with its capture.
Probit analyses revealed price to be only a secondary factor influencing purchase
behavior. MAC certification was considered mainly as a weak substitute for competing
attributes, such as an extended survival guarantee and tank culture, showing even
negative effects on the likelihood to purchase. Only an increased knowledge of the MAC
program, mainly at an intermediate level, and a strong association of the MAC ecolabel
with effective conservation of reef habitats and wild stocks were likely to positively
influence the likelihood of purchasing a MAC certified fish.
Focusing efforts on improving the credibility of the program, not only of coral reef
protection but also sustainability of fish stocks, and efficient post-harvest activities (i.e.,
improved handling, holding and transportation) could be successful in improving
preferences for a MAC ecolabel. The high level of involvement with large Internet
bulletin boards and discussion lists, and the secondary role of price on purchase decisions
of this hobbyist segment suggest an opportunity for increasing the effectiveness of future
educational and promotional efforts by the MAC at relatively low cost.
The Marine Aquarium Industry
The marine aquarium industry depends upon the supply of marine ornamental
species such as corals, saltwater fish and invertebrates, which are harvested from coral
reefs usually located in developing countries (Wabnitz et al. 2003). In some cases,
collecting aquarium fish represents the main economic activity of low-income coastal
According to the last report on the global trade of marine ornamentals from the
United Nations Environment Program (Wabnitz et al. 2003), the marine aquarium
industry is still of relatively low volume, but it is emerging as a high value sector in terms
of trade. It currently serves a market of 2 million hobbyists worldwide and generates a
trade value of between $200 and $330 million annually. In terms of the volume of annual
production, over 20 million tropical fish and about 9 to 10 million mobile invertebrates
are collected, involving 1,471 and 500 species, respectively. In addition, 12 million stony
corals and 390,000 pieces of soft coral representing 140 and 61 species, respectively, are
also harvested and traded in the market.
The majority of organisms traded in this industry are wild-caught, as tank breeding
and raising accounts for only 2% of the fish and at most 1% of the coral traded in the
marine aquarium sector. This is in contrast to the freshwater aquarium industry where
aquaculture provides 90% of the specimens (Wabnitz et al. 2003).
The primary sources of saltwater ornamental fish are located on the coral reef areas
of Southeast Asia and the Indian and Pacific Ocean Islands, especially the Philippines
and Indonesia. Other sources are located in Australia, Hawaii, Mexico, Florida, the
Caribbean, Brazil, East Africa and the Red Sea (Marine Aquarium Council [MAC]
The main destination for marine aquarium exports is the U.S. market. It accounts
for 80% and 50% of the exports of stony corals and marine fish, respectively. Other
important markets are located in Germany, France, U.K., The Netherlands and, on a
reduced scale, Japan (Wabnitz et al. 2003).
Since most of the species traded in this market are wild-caught, the impact of the
marine aquarium trade on the sustainability of its stocks and the welfare of surrounding
reef ecosystems have become controversial issues. According to Holthus (2001), the
most important concerns are related to
* The environmental and biological consequences of commonly applied destructive
collection methods. One common example is the use of cyanide or other chemicals
to catch aquarium specimens and live reef fish. In addition, collectors often break
corals to get at stunned fish and invertebrates. These methods affect the quality and
expected life span of aquarium fish throughout the commercialization chain, and
they harm corals and other non-target species. Although programs to reduce the
use of these chemicals have been implemented in a few countries, such as the
Philippines, there has not been a reliable system to verify the method of capture for
any given specimen in a way that provides any assurance for suppliers or aquarists.
In addition, destructive collection methods can feed back and cause stocks of target
species to decline due to loss of suitable habitat.
* The unknown potential for overfishing of target species that comes from
uncontrolled and unmeasured effort applied in limited collection areas. The
intensity of harvesting reef animals is seldom assessed, nor is the impact of
harvesting on other organisms that support target species. Therefore, the industry
faces the risk of allowing unsustainable harvest levels and detrimental impacts on
* The need to collect extra specimens due to unnecessarily high mortality rates
arising from inappropriate husbandry, handling, holding and transport practices
along the chain of custody (i.e., collectors, wholesaler, retailers). These procedures
have not been subject to standard guidelines and performance monitoring. It is
important to note that it is impossible to ensure a 100% survival rate for all marine
ornamental species due to their particular requirements and sensitivities, but
improved processes should lower mortality and reduce harvesting pressure.
In the scope of these concerns, the use of collection methods that cause minimal
impacts on coral reef ecosystems and the implementation of controlled practices to
minimize mortality along the supply chain have become appealing topics in the industry.
If adequately controlled and managed, the industry could promote long-term, sustainable
use of coral reefs, which is vital for the economic performance and sustainability of the
industry as a whole and for regions where other options for income generation are limited
(Wabnitz et al. 2003).
One tool that has already been initiated in an attempt to promote sustainability of
the wild-caught aquaria trade is "ecolabeling" or an environmental certification scheme.
The term "ecolabeling" can be generally defined as "the use of labels in order to inform
consumers that a labeled product is environmentally friendlier relative to other products
in the same category"(Goodland 2002, p.2). The label guarantees that the generation of a
particular product does not affect either the sustainability of the resource supporting the
product or the environment surrounding the resource and consumers. Thus, the general
purpose of ecolabeling programs is to promote the sustainability of resources and the
environment through market mechanisms.
An Ecolabeling Program for Live Marine Ornamentals
The Marine Aquarium Council (MAC) was created in 1998 in response to the
aforementioned concerns in the marine aquarium industry. It is an international,
independent, non-profit organization that seeks to ensure the quality and health of marine
ornamentals from "reef to retail." The MAC is based in Honolulu, Hawaii and its
stakeholder network currently includes more than 3,000 aquarists, industry operators,
conservationists and researchers in 60 countries. Although the program is currently
funded by international partnerships, foundations and independent non-commercial
sources, the goal is to support operations through fees paid by industry members for
certification (MAC 2004b).
The MAC has created a third-party certification program to assure compliance with
standards designed to support sustainability. Certified parties can display a label
proclaiming their environmentally sound practices (i.e., collectors, exporters, importers,
retailers) and products (i.e., marine aquarium ornamental fish). This program brings
together collectors, importers, exporters, retailers and aquarists to accomplish a common
goal. This effort became operational in late 2001, and by 2002 some certifications were
already conferred. The initial scope of the program only extends to collection from the
wild, but it is expected to include aquaculture practices in the future.
The ultimate purpose of the MAC ecolabel is to inform consumers (i.e., hobbyists)
about reduced environmental effects caused by certified activities and empower them to
promote sustainability through their purchase decisions. According to information
published by the MAC (MAC 2004c), the most important objectives of the program are:
* To develop core standards to assess marine ornamental practices.
* To create a system to verify the implementation of standards and certify qualified
products and practices.
* To provide a framework that allows the industry to conduct responsible collection,
handling and transporting practices as well as to generate accurate data for the
management of marine ornamental activities; and
* To support responsible management through education and training for industry's
The core (standard) criteria developed by the MAC and used for assessment by the
accredited independent certifiers deal with coral reef conservation, as well as with the
health and sustainability of wild stocks. The standard criteria applied in this program are:
* Ecosystem and fisheries management, which addresses "in-situ" habitat, stock and
species management and conservation in the collection area by verifying that
management is conducted according to principles ensuring marine ecosystem
conservation and stock sustainability.
* Collection, fishing and holding, which focuses on harvesting fish, coral, live rock
and other coral reef organisms and on related activities, (e.g., handling, holding,
packaging and transport prior to export) by verifying that the collection, fishing,
and pre-exporter handling, packaging and transport of marine aquarium organisms
do not harm the health of the collection area, the sustainable use of the marine
aquarium stocks and the optimal health of the harvested organisms.
* Handling, husbandry and transport, which addresses the handling, husbandry,
packing and transport at points along the commercialization chain in an attempt to
ensure the optimal health of organisms during the commercialization process, as
well as the differentiation of labeled products and practices from uncertified ones.
(One important thing to note is that this standard implies that a certified product has
to pass only from one MAC certified industry operator to another).
Currently, the MAC works with four internationally accredited certifiers. They
have awarded the ecolabel to more than 30 agents in the industry including individual
commercial firms and fishermen associations (Table 1-1). However, the number of
actual certifications represents a very small portion of industry operators worldwide. Just
to illustrate the situation at the collection level, in Sri Lanka alone there are reported to be
approximately 50,000 individual collectors participating in the industry, and in the
Philippines only 2 collectors' associations in that area have committed to the MAC
ecolabeling program, representing only 63 members in total out of approximately 7,000
(Bunting and Meyers 2002).1
1 Batasan Tropical Fish Collectors Association (BATFCA) has 31 members and Tangaran Aquarium Fish
Gatherers Association (TAFGA) has 32 members actually certified by MAC. This information was
Table 1-1. MAC certified operators
Collection Areas Fiji
Importers United States
Retailers United States
Source: Marine Aquarium Council (2004d)
Number of certified agents
The implementation of the ecolabeling program for marine aquarium organisms is
in its initial phase; therefore, it is too soon to draw conclusions about the achievements of
the certified parties. However, opportunities for promoting a sustainable marine aquarium
industry have emerged and they include (MAC 2004e):
* Creation and implementation of previously non-existent management plans for reef
conservation in certified marine aquarium collection areas, which has involved the
joint participation of fishermen and other stakeholders in establishing no-take reef
reserves and fish sanctuaries in certified collection areas in the Philippines.
* Reduction in overfishing and elimination of destructive collection practices, such
as when fishers certified by the MAC are trained to use non-harmful fishing
methods, keep records of the species and the amount they harvest to generate a
baseline that will allow management authorities to manage the stock sustainably,
and importantly, only collect what a certified buyer has ordered, with records
provided by fishermen being verified against purchasing information from certified
* Reinforcement of sustainable livelihoods for fishers, who have few economic
options beyond harvesting marine aquarium organisms, with a positive flow-on
effect to the sustainability of their communities, and a potential positive flow-on
effect to other communities if the program performs as expected.
provided by Rezal Kusumaatmadja, Asia Director of the MAC program (personal communication, 6 July
On the other hand, the MAC program faces some challenges related to those
already faced by the Marine Stewardship Council's (MSC) certification program for
seafood, its predecessor.2 One of the challenges relates to overcoming potentially high
levels of skepticism from environmental organizations and consumers concerning the
effective assessment of ecosystem impacts. In addition, issues about costs and criteria of
valuation that could preclude the participation of all fisheries, as in the case of the
seafood certification program, could also be applicable to the MAC program (Gardiner
and Viswanathan 2003). In addition, market surveys about willingness-to-pay for
certified products have shown that consumer preferences depend heavily on consumers'
knowledge about the purpose of ecolabeling, their own environmental and personal
interests, the degree of market development in their geographic region and even the
credibility of the certification agency (Wessells et al. 1999). All of these influences may
make ecolabeling inapplicable or unacceptable in some markets and reduce the scope to
"niche markets" rather than a force for global improvement of fisheries management
(Gardiner and Viswanathan 2003).
One important distinguishing feature of the MAC certification program is the
restricted nature of the chain of custody for marine ornamentals. The system requires
certified operators to work only with other operations that meet the MAC quality
standards. The commercialization of certified products can only involve certified
operators (i.e., collectors, wholesalers, and retailers) until the marine ornamental is sold
to an aquarium hobbyist. In this aspect, the market acceptance throughout the chain is
2 The ecolabeling program for seafood fisheries was developed and implemented by the MSC. This
program constituted the first third-party independent environmental certification initiative in the fisheries
sector. To date, 10 fisheries have been certified and 12 more are undergoing certification assessment.
vital for the support of certified custody systems and the MAC program itself. If the
market does not respond to the label and there are not enough incentives to expand these
chain of custody interactions, the decrease in incentives can create an even more
restricted chain of commercialization than the one envisioned by the MAC. This
challenge faces the MAC and it is one significant difference with the MSC's seafood
certification program where post-harvest, handling and transport issues do not exist.
A few preliminary studies have explored the effects that several fish and buyer
characteristics have on purchase decisions for select ecolabelled aquarium organisms.
Interestingly, preferences for the MAC ecolabel have been shown to differ among agents
in the industry and fish species, at times, there are even negative effects on purchase
decisions. For example, according to Rubinstein (2003), the length of survival guarantee
as a fish characteristic could be preferred and perhaps considered as a more effective tool
in promoting improved fish handling and transport than MAC certification. In addition,
for species that can be cultured, operators may be less willing to pay for MAC
certification. Moreover, the value of the MAC ecolabel can depend on species and the
familiarity the buyer has with the program. In the above-mentioned study, the majority
of respondents were familiar with the MAC certification program, but an increased level
of familiarity reduced the profitability rating assigned by wholesalers and retailers for
some certified species, specially the high value ones. In addition, those more familiar
with the MAC certification (ecolabel) program had a lower maximum price they were
willing to pay for the fish because of anticipated cost increases associated with becoming
certified and purchasing only from certified suppliers. However, a high level of
familiarity with the program was likely to have a positive impact on demand and the
probability of purchasing a certified fish, especially for wild-caught specimens.
As a result of the increasing concerns about the sustainability of the marine
aquarium industry, the MAC ecolabeling program emerged as a voluntary market-based
tool to promote the sustainability of saltwater ornamental fish and coral reef ecosystems.
If demand for certified organisms is generated, the market could drive the industry
toward improved environmentally-friendly production of marine ornamentals; thus, it is
important to examine the influence that an environmental quality assurance attribute, like
the MAC certification ecolabel, has on consumers' purchase decisions.
In addition to certification, other attributes of the product, as well as the socio-
demographics and attitudes of consumers, are likely to influence purchase decisions for
marine aquarium species. Since previous studies have explored fish preferences among
operators within the commercialization chain for this industry, it is necessary to generate
further information about preferences for marine aquarium fish at the final consumer
level and to analyze auxiliary influences on purchase decisions by hobbyists. These
results can be used to estimate the importance of certification and, ultimately, the
potential of the MAC program to promote a sustainable marine aquarium industry.
Objectives and Hypotheses
The main objective of this thesis is to determine the preferences of avid hobbyists
for marine ornamental fish, through the implementation of a market experiment based on
a stated preference technique. To achieve this purpose, the study implements an Internet-
based survey utilizing a choice experiment format to assess the preferences of hobbyists,
who are members of online aquarium discussion boards, for two species of marine
ornamental fish: the blue-faced angelfish (Pomacanthus xanthometapon) and the maroon
clownfish (Premnas biaculeatus).
In this context, the broad objectives of the study are:
* To characterize a sample of highly informed marine aquarium hobbyists, (e.g.,
members of online discussion groups).
* To ascertain opinions regarding tank-bred and ecolabelled specimens and
alternative production techniques.
* To estimate the differences in likelihood to purchase a fish with different attributes,
including whether certified (i.e., ecolabelled), price, source (wild-caught vs. tank-
bred) and length of survival guarantee.
* To estimate the effect that relevant characteristics of hobbyists have on their
preferences for ecolabelled fish.
The specific hypotheses to be tested are listed below:
* Members of this segment are well-informed individuals, with avid participation in
* The majority of respondents have a high level of familiarity with the MAC
* Sustainable harvest practices and conservation of reef ecosystems have significant
influence on preferences for saltwater ornamental fish.
* Price is not the most important attribute affecting the purchase of marine
* Environmental certification is perceived as a positive value-added attribute
influencing purchase decisions for ornamental saltwater fish.
* Certified fish are likely to be preferred to cultured fish and fish with larger post-
purchase survival guarantees.
* Respondents from regions normally exposed to the aquarium industry are likely to
exhibit a positive preference for the MAC ecolabel.
* High levels of education, incomes, and age will positively influence preferences for
a MAC certified fish.
The information obtained from the survey is expected to be of value for several
segments of the industry. One of the main potential users is the Marine Aquarium
Council, since results estimate the importance and potential of the MAC ecolabel in the
hobbyist market segment. These results will complement studies of ecolabeling for this
industry for agents at intermediate levels of the chain of custody. In addition, the
research may identify challenges to the current and future acceptance of the certification
program for the hobbyist segment that can be used to tailor informational materials to
increase the effectiveness and success of the program.
Other potential users are members of the industry (i.e., retailers, wholesalers, and
collectors) planning to adopt certified practices, or simply interested in obtaining
indicators of hobbyists' preferences for the specific attributes analyzed. This study can
provide them with information for developing marketing strategies directed to the
segment of avid hobbyists. Moreover, since one of the species included in the market
experiments (i.e., the maroon clownfish) is available from cultured sources in Florida,
information about observed preferences could also contribute to identification of market
opportunities for Florida's producers of saltwater aquarium fish.
Contents of the Study
The remainder of this document begins by introducing the concepts used to design
and implement the market experiments in this research. Theoretical foundations of
discrete choice modeling, as well as a brief overview of empirical applications in the
literature are given in Chapter 2. Chapter 3 provides a detailed description of the
methodology followed to implement the survey (i.e., design of choice experiments,
sampling technique, data collection, and modeling specifications) and potential
limitations to the scope of the results. Characterization of respondents, and results of the
experiments are presented in Chapter 4. Finally, summary and conclusions are discussed
in Chapter 5.
The value that goods and services have in tradable markets can be estimated from
the direct observation of historical monetary transactions. On the other hand, when the
value cannot be directly obtained from a market transaction since the products have to be
sold in sufficient quantities with the attributes of interest, different tools need to be used.
Such tools allow the researcher to estimate the value of specific attributes by exploring
and systematically analyzing consumers' preferences expressed during hypothetical
Alternative Valuation Approaches
Two main techniques have been developed to obtain non-market values. One of
them, known as the "revealed preference" method, uses observed consumer behavior in
markets. Estimation and analysis of preferences in this context are influenced and limited
to past purchase behavior and price levels.
An alternative approach bases the estimation on "stated preferences". As its name
indicates, this approach focuses on the analysis of preferences that have not been
observed in a market. Instead, consumers indicate what their preferences would be
across different hypothetical scenarios designed by the researcher and implemented
through survey instruments. One characteristic of this valuation approach is its ability to
generate preference information when no actual markets exist. This is, stated preference
techniques allow the analysis of the market potential of a product prior to its introduction
or the viability of new services or policies in the case of environmental valuation
(Bennett and Blarney 2001). One stated preference technique that is being increasingly
used in fields like psychology, marketing, transportation and recently in environmental
economics (Adamowicz et al. 1998; Haaijer 1999) is discrete choice modeling.
With this method, consumers are asked to choose their preferred product/service
alternative from a small set of options. The alternatives are different profiles constructed
by combining attribute levels for the good or service under analysis (i.e., characteristics
and set of values defining them). The total number of possible combinations of attribute
levels is called a complete factorial. In order to not overload respondents, the number of
profiles presented to respondents for evaluation is usually an efficient fraction of the
complete factorial.1 Once selected, the efficient profiles are assigned into small groups
for comparison (i.e., choice sets), according to the most appropriate method of choice set
design.2 The process of selecting the relevant attributes and levels, combining them to
create hypothetical profiles, selecting the efficient profiles and comparing them in a
choice set is known as a choice experiment.
Adamowicz et al. (1998) state that since the alternative profiles are defined in terms
of product characteristics, this technique is especially helpful for studies addressing
multi-attribute effects on preferences and evaluating the attribute trade-offs that
individuals are willing to make. In addition, Orme (2003) considers choice exercises to
be a remarkably accurate representation of the purchase process in real markets,
especially in a competitive context.
1 The term efficient indicates that profiles are selected using a fractional factorial approach. Fractional
factorial designs maintain the orthogonality (i.e., independence) property across attribute levels formed by
complete factorials and allow particular effects to be estimated using a reduced set of profiles (Louviere et
2 A practical description of choice set design techniques is provided by Bunch et al. (1996).
In comparison to other stated preferences techniques, such as the contingent
valuation method (CVM) and conjoint analysis, the literature has identified several
advantages of choice modeling. Whereas CVM focuses on consumers' willingness-to-
pay for a single scenario (profile), choice modeling focuses on the valuation of
preferences for attributes and attribute levels that define the profile, which allows for
analysis of specific attributes and attribute levels. The main trade-off in a single CVM
exercise is limited to monetary terms (e.g., price), while the choice approach allows
evaluation of trade-offs among attributes (including price) and attribute levels
(Adamowicz et al. 1998).
Conjoint analysis and choice experiments both evaluate alternatives derived from
statistically efficient combinations of attributes. Conjoint analysis also elicits the
evaluation of a wider scope of attributes and alternatives than CVM (similar to choice
modeling), but the focus of the preference analysis is different than with choice
modeling. Instead of choosing the preferred alternative, with conjoint analysis each of
the options (profiles) receives an overall score using a rating or ranking system. The
main output in conjoint analysis relates to the estimation of the acceptability of the
attribute levels and the extent respondents prefer one alternative to another (Huber 1997).
In contrast, with choice modeling, respondents evaluate preferences for attributes among
competitive alternatives by selecting a single alternative (profile). Haaijer (1999) notes
that the evaluation task in choice modeling is more realistic than rating or ranking used in
Although the choice technique offers a better possibility for multi-attribute
valuation, in comparison to other stated preferences methods, it is also subject to some of
the concerns of the CVM approach. One of the most salient issues is what is called "Yea-
saying" behavior. This relates to the possibility of observing strategic choices made in
favor of the most acceptable alternatives (e.g., socially or environmentally) instead of real
preferences. Another concern is the potential effect that not explicitly representing
relevant substitutes for the attributes in the choice tasks can have on the estimation of
attribute importance (Bennett and Blarney 2001).
Choice Modeling Theory
Choice modeling is consistent with the concept of random utility theory (RUT).
Therefore, observations obtained in a choice experiment can be analyzed with models
consistent with RUT. In a choice context, RUT defines the utility perceived by
consumers for a given alternative within a choice set as an unobservable construct that
can be partly explained by a deterministic component. There will also always be an
unexplained portion captured by a random component. Thus,
Ujn* = Vjn + Sjn, (2.1)
where j = 1,... J choice alternatives in choice set c = 1,.....C and n =1,..N individuals.
Ujn* is the unobserved utility perceived by individual n for alternative j. Vjn is the
systematic, observable component of the unobserved utility of an alternative defined in
terms of its attributes. Lastly, Sjn is the random component of the unobserved utility
related to alternative j for individual n.
Louviere (2001) developed a comprehensive general model specification of the
systematic component of utility for a choice alternative:
Vjn = Pj +kkPkXkn+ LiiAZin + lkli4kiXknZin+ iYja0ijPjZin, (2.2)
where k =1 ... .K product's attributes or characteristics and I =1.....I respondents'
In this expression, j3j refers to a vector representing the intercept term or a group of
intercepts also known as alternative specific constants (ASC) for the J-1 choice
alternatives.3 Xkn is a vector representing the choice options created by the combination
of the k attribute measures. Zin represents a vector of i characteristics of individual
consumer n. XkZin and PjZin represent vectors of relevant interactions between
individual characteristics and attributes or intercepts, respectively. Interaction terms of
individual characteristics with attributes or intercepts are usually introduced to account
for preference heterogeneity across individuals in repeated choice exercises (Mazzanti
2001; Bennett and Adamowicz 2001).
Under the specification shown in equation (2.2), the researcher can obtain estimates
for the partial utilities derived from each attribute level, which add to a total utility. The
treatment of price as an attribute allows the derivation of monetary trade-offs for the
attribute levels as well as discrete measures of welfare or compensating surplus among
the utilities of the alternatives evaluated in the choice experiment (Irwin et al. 2002;
Bennett and Adamowicz 2001).
Given the presence of a stochastic term (i.e., error), the utility obtained from an
alternative cannot be completely explained by the observable attributes and, therefore,
' Alternative specific constants (ASC) capture any variation in choices that cannot be explained by the
specified covariates. ASC's are usually introduced in the alternatives' utility specification when a base
option (e.g., status quo scenario or no-choice option) is part of the choice set. In such case, the ASC
captures any unobserved effect (positive or negative) that influences the utility of a given alternative in
comparison to the base.
can be specified only in terms of the probability of choice of such alternative. Thus, the
probability of choice of a given alternative can be modeled as
P (j c) = P [(Vjn+Sjn) > (Vin+Sin)], (2.3)
where 1 j in choice set c. This expression states that the probability of consumer n
choosing alternative j from choice set c is equal to the probability that the sum of the
systematic and random components of option j for consumer n in such choice set is
greater than the sum of the systematic and random component of alternative 1 within the
same choice set. Through this probability modeling, the researcher obtains an indication
of the relative utility generated by the alternatives, given that an individual rationally
chooses the option that gives him/her the greatest utility (Bennett and Adamowicz 2001).
The assumptions made about the nature of the error terms determine in part the
type of modeling to be used. The most common probability models used are the ones
from the family of logits and probits. The former applies when it is assumed that errors
are independent and identically distributed (IID) as a Weibull distribution, while the latter
is suitable under the assumption of non-independent, non-identically distributed errors,
which follow a normal distribution (Kennedy 1994).
A choice experiment provides observations about chosen and not chosen
alternatives for each choice set. Therefore, each of the alternatives of a given choice set
can be included in the modeling process. Indirect utility functions, such as the one
presented in equation (2.2), are to be specified for each of the alternatives presented in
the choice set, in such a way that parameter estimates are always constant across
alternatives' utilities while any ASC for the J-1 alternatives can be allowed to vary
(Bennett and Adamowicz 2001). The observed choice for each alternative in a given
choice set is the dependent variable, coded as 1 for the chosen option and 0 for the non-
selected options (alternatives). The explanatory variables in the models are the attribute
levels specified in each alternative in each choice set and any other relevant covariates,
such as the characteristics of respondents.
Econometric Issues in Choice Modeling
Due to its relative computational simplicity, the models most used in empirical
work have been those from the family of multinomial logit models (MNL), especially one
known as the mixed conditional logit (MCL).4 However, the assumption of IID errors
across alternatives, choice sets and individuals implied by the logit models may not hold
in reality. The direct consequence of this assumption is the so-called property of
Independence of Irrelevant alternatives, (IIA), which may not apply in reality either.
The IIA property means that the introduction of additional attributes levels of a
specific alternative in a choice set do not influence the choice among the other
alternatives present. If this property is violated by the data, the MNL will produce
inefficient estimates. As indicated by Fry and Harris (1998), in the presence of IIA
violations, MNL estimates would remain unbiased and consistent, but standard errors
would be inappropriate, ruling out the efficiency property in estimation.
To test for the validity of the IIA assumption in MNL models, Hausmann and
McFadden (1984) developed a procedure that allows for testing the null hypothesis of no
IIA violation. To do so, main effects and all 2-way interactions among attributes must be
4 The conditional logit model is a variation of the classical multinomial logit specification that deals
exclusively with covariates that take different values according to choice categories, which is the case of
product attributes (Liao 1994). A mixed specification refers to the addition of individual characteristics
(which are choice invariant) to the basic conditional model.
introduced in the experimental design, in what is called a "mother logit" specification. If
statistical tests show that estimates of the 2-way interactions in the mother logit (i.e.,
unrestricted) model are not different from zero when compared to the basic main effects
MNL (i.e., restricted model), the IIA property is proved to hold (Haaijer 1999).
If the IIA assumption is not satisfied, several modifications to the MNL approach
have been suggested in the literature. According to Orme (2003), one strategy to
overcome the IIA assumption would be to analyze choice at the individual level instead
of at the aggregate level. He suggests latent class and hierarchical Bayesian models as
means to allow an individual-level analysis. An alternative is the nested logit model.
Under this model specification respondents are assumed to first make a decision between
a base alternative (which can be a constant option across choice sets or the option of not
to choose) and any other alternative. If opting for an alternative different from the base,
then the choice occurs at a second level or "nest".
Other non-IID models, like the heteroskedastic extreme-value logit and probit
models (binary or multinomial), can be used to allow correlation among alternatives
across choice sets. The specification of the error term in these models allows residuals to
be correlated not only among the random utilities of alternatives within choice sets, but
also among repeated choices, avoiding IIA restrictions by definition (Haaijer 1999;
Mazzanti 2001). Due to the relatively higher computational complexity of the
multinomial probit, this approach is not often used.
Unobserved Preference Heterogeneity
Haaijer (1999) distinguishes between 2 types of heterogeneity in the sample
population. One relates to heterogeneity in tastes arising from differences in individuals'
tastes that make utility functions of alternatives differ among individuals. The other
involves the influence of individual characteristics on the variance of the stochastic term
(i.e. error) or on the parameters of the utility function under study. Unobserved effects
influencing choice among individual respondents within the sample could arise from
adaptive behavior in repeated choice tasks since previous choices can influence the
subsequent ones. In the context of repeated choice, Louviere et al. (2000) states that
heterogeneity can also be seen as a special form of serial correlation.
Addressing heterogeneity consists of introducing a specification of the individual
specific effects on the utility function. This can be done in several ways. The most basic
one is by introducing individuals' socio-demographics as explanatory variables of the
utility through interactions with the attributes or with the ASC if such specification is
A more elaborate approach consists of including a unique effect for each
respondent as an additional intercept term, which generates a fixed effects model. The
introduction of individual specific effects in the model avoids the violation of the IIA
property (Mazzanti 2001). In this way, possible correlations arising from repeated choice
tasks and alternatives evaluated by the same individual are allowed to contribute to the
explanation of choice behavior. As observed in some empirical work, after the inclusion
of socio-demographics in MNL, the IIA property has been proved to hold (Morrison et al.
Another approach is to specify individual characteristics as influencing the
structure of the error term, which leads to a random effects model specification. Under
this method, the stochastic component of the unobserved consumer utility (i.e., error
term) consists of the traditional error term associated to each observation in the data set
and an additional term capturing any potential variation across individual respondents
(Louviere et al. 2000; Kennedy 1994).
Individual characteristics can also be specified as influencing one or more
parameters in a systematic way. In this sense, the specification leads to what is known as
a random coefficients model that specifies utility parameters as stochastic terms varying
across individuals, instead of being fixed estimates for the whole sample population
Inclusion of a Base Alternative
A base alternative, also called an opt-out option, is an alternative that is constant
across choice sets in the experiment (i.e., included in each choice set). Besides the option
of holding constant one regular profile across choice tasks, this constant alternative has
been traditionally specified in two main ways. One specification introduces the base
option as the current situation, product or policy (whichever is applicable), labeled as the
"status quo". Alternatively, the constant alternative can be specified as the option of not
to choose. Carson et al. (1994) indicates that the "no-choice" format is more suitable
when the main objective of the research is to measure market share or acceptance of the
alternatives in the choice set. On the other hand, the current situation or status quo
format is much more helpful when the objective is to analyze what attributes and attribute
level combinations makes the hypothetical new alternatives or scenarios preferred over
the current brand, policy or situation.
Banzhaf et al. (2001) and Haaijer (1999) review the main implications of
introducing a constant alternative across choice sets. According to these studies one
advantage of including a base alternative is that it makes the choice set a better
representation of real purchase behavior in markets. This is because consumers may
always decide to delay purchase or stick to their current product or situation (i.e., choose
the base), if their needs are not satisfied by the alternatives available in the marketplace.
Another compelling reason to include an opt-out option is to avoid forced choices in the
experiment arising from exposure only to the hypothetical profiles contained in the
design. The main implication of forced choices is the potential overestimation of the real
preference that respondents have for a given alternative. In other words, an alternative
may be chosen in the restricted choice set when, in reality, respondents would have
preferred an alternative that is not in the set or none of them. Moreover, Carson et al.
(1994) remarks that the use of a constant option can enrich the analysis by allowing the
estimation of market shares, viability of new products and demand.
One disadvantage of including an opt-out option is that it allows respondents to
avoid making difficult choices. If the constant alternative is selected because it was the
easy choice, the observed preference behavior is distorted and the usefulness of the no-
choice probability to estimate market shares would be null. In addition, respondents may
choose the constant option for different reasons in comparison to the remaining
alternatives. In the case of a no-choice option, as Haiijer (1999) discusses, respondents
could choose that option more often because there is not a clearly preferred or
unattractive alternative in the choice set, not because of a real preference for that option.
Again, this situation is very likely to distort real purchase preferences in the experiment.
Another disadvantage of using an opt-out option design is that no information
about preferences for the attributes under analysis is provided when it is chosen, which is
at odds with the main purpose of the choice exercise. In addition, including the opt-out
or no-choice option is essentially adding an additional attribute level to each attribute
analyzed in the experiment. In this sense, additional complexity is created since the
effect of a no-choice option on attribute level combinations needs to be introduced into
the design matrix of the experiment and also into the specification of the model to
analyze choice observations (Haaijer 1999).
Thus, an opt-out option may not be appropriate for each choice experiment and its
inclusion would depend mainly on the focus of the preference analysis. That is, whether
the study is interested mainly in preferences for attributes in an overall likelihood to
purchase context or also specific utility scales among the specified alternatives.
Empirical Applications of Choice Modeling
The choice modeling technique for evaluation of stated preferences was first
applied in the marketing and transportation areas, covering a wide array of specific
problems (Haaijer 1999). However, it has also been increasingly implemented in several
additional fields (Table 2-1). Although this overview represents only a portion of actual
applications, its purpose is to illustrate how issues like violation of assumptions or
introduction of preference heterogeneity into the modeling have been addressed in
practice, and whether they have contributed to the efficiency of results.
Most empirical studies have used unbalanced designs in their experiments, that is,
where the number of levels for each attribute is not the same (Table 2-1). A wide variety
of choice set sizes and number of choice tasks in the experiments have also been
observed (Table 2-1). Opt-out options have been included in the experiments mainly
when there is an interest on measuring preferences or market shares between the specific
alternatives featured in the experiment (Table 2-1). For studies with only two alternatives
per choice set, constant options are not likely to be used (Table 2-1). In those studies, the
alternatives were not representing real competitors, such as brands, but were mainly
different configurations for obtaining the same product or service. This suggests that
when there is no interest in the analysis of market shares for alternatives, paired
alternatives in the choice sets can be appropriate, which also reduces the magnitude of the
choice task faced by respondents.
Models for binary choice as well as for multinomial choices have been used (Table
2-1). The independent variables to be modeled have been either differential utility (for
binary choice experiments) or total utility for each alternative. In the first case, the data
set is usually constructed to include one observation per respondent per choice task, while
for the second case the number of observations per response equals the number of
alternatives in the choice set. Complementarily, choice modeling experiments have been
implemented mainly at the aggregate level and with repeated choice tasks (i.e., several
choice sets) (Bennett and Blamey 2001; Orme 2003). In the case of repeated choices, the
0/1 dependent variable observations, as well as observations of explanatory variables, are
pooled for each individual. For an aggregate analysis, observations from all individuals
are pooled. In addition, the functional form is generally assumed to be a simple linear
For choices between two alternatives, some studies have used a binary probit
specification because it allows the IIA assumption to be relaxed (Table 2-1). For the case
of choices among more than two alternatives, the most common approach seems to be a
conditional logit model accounting for attribute effects only, and then a test for violations
of the IIA property (Table 2-1).
Table 2-1. Overview of empirical discrete choice modeling
Individuals No.Attributes No. Choice
Study Topic (obs.) (Levels) Alternatives a tasks Model IIA b Heterogeneity Use d
Phillips et al. Health care 339 6 2 11 Binary probit N/A RE, D 1, 2
(2002) intervention (3,366) (3-6)
Irwin et al. Neighborhood
Lusk and Fox
Mazzanti (2001) Cultural Heritage
Alberini et al. Environmental
(2003) remediation and
Wielgus and Coral reefs
Riera and Forest
Wessells et al. Ecolabeled seafood
Blarney et al. Water supply options
2 Binary probit
18 Conditional MNL N/A
3,4 Conditional MNL D, HEV
4 Conditional MNL RP
1 Conditional MNL NL
4 Conditional MNL N/A
3 Binomial logit
9 Conditional MNL N/A
Morrison et al. Wetland quality 318 5
(1998) (1,577) (3,4)
a (NC) no choice, (SQ) status quo, (FC) fixed constant.
4 (SQ, NC) 5 Conditional MNL MLA
b (HEV) heteroskedastic extreme value MNL, (NL) nested logit, (MLA) mother logit with attributes interactions.
' (RE) random effects, (D) demographic interactions, (RP) random parameters.
d (1) implicit price of attributes, (2) effect on probability of purchase, (3) compensating surplus for attribute changes, (4) compensating
surplus for different scenarios, (5) effect on utility, (6) market shares.
To correct for violations of the IIA property, if necessary, and to account for
preference heterogeneity, several additional specifications are estimated and compared in
order to find the model with the best performance (Table 2-1). For IIA considerations,
the main types of specifications implemented in these studies accounted for individual
specific effects on observed choice (i.e., preference heterogeneity) or implemented non-
IID models (i.e., models assuming non-independent, non-identically distributed errors).
The introduction of individual characteristics in the basic conditional model, to generate
random effects, random coefficients, mother logit and heteroskedastic extreme-value
specifications, are the most likely approaches used to obtain efficient estimates (Table 2-
1). Interestingly, in several studies, the basic specification (i.e., attributes only) with
demographic-based interaction terms have been enough to avoid violation of the IIA
assumption and to account for heterogeneity in the sample. Therefore, as Mazzanti
(2001) stresses, the theoretically superior fit of more sophisticated specifications than the
additive introduction of demographic interactions may not always apply in empirical
Form of Choice Modeling Results
The results obtained from choice models are the estimates of the effect of the
attribute levels on the overall utility of a given alternative, or in other words, the partial
utilities obtained from each attribute level. However, the coefficient estimates have been
used to expand the scope of the analysis of applied studies (Table 2-1, final column).
Using the assumed cumulative distribution function of the error term, utility estimates of
the alternatives evaluated in the experiment can be transformed into probabilities of
choice. Empirical studies have applied this approach by calculating changes in choice
probability given a range of different values for the explanatory variables (i.e., attributes
and individual characteristics).
Additional results relate to the calculation of the attribute trade-offs that
respondents would be willing to make. Such estimations are obtained by the direct ratio
between the coefficients of the two attribute levels of interest. When the trade-off relates
to a given attribute and the monetary term, the ratio between them reflects the implicit
price of the non-monetary attribute (i.e., how much individuals would be willing to pay
or need to be paid for having an additional unit of the attribute level).
Another application of utility estimates is the calculation of compensating surplus
between hypothetical or actually evaluated alternatives or scenarios. In this context,
compensating surplus is defined as the differential amount of money that would make
individuals indifferent between a pair of scenarios because both of them provide the same
level of utility.
Finally, another useful application of utility coefficients relates to the estimation of
market or participation shares, specially when a constant status quo or no-choice
alternative is introduced into the choice tasks. This type of analysis seems to be more
appropriate for studies addressing preferences for highly competitive alternatives in real
situations rather than for hypothetical options for the same product or service. A few
examples of such competitive choice alternatives could be product or services brands or
different policy scenarios in comparison to the current situation, as could be the case in
some environmental valuation problems.
One of the objectives of this study is to analyze preferences for the MAC
certification status of marine aquarium ornamentals, among other selected fish
characteristics. As the implementation of the MAC program is relatively recent, actual
market information is not available. Therefore, valuation of hobbyists' purchase
behavior with respect to this program needs to be done under the stated preference (SP)
approach. The SP technique selected for this study is choice modeling. This technique is
believed to be the most appropriate because the analysis is desired at the aggregate level
and estimations of attribute level effects on purchase likelihood are needed. In addition,
in comparison to other SP techniques, the choice format is a more realistic representation
of the purchase task in real markets.
The economic foundations and empirical applications of discrete choice modeling
reviewed in the previous chapter provided the basis for the design and analysis of the
market experiments implemented in this research. Descriptions of the experimental
design, sampling approach, questionnaire development and implementation, as well as
specification of the model to analyze the data, are provided in the following sections.
Two choice experiments for two different types of aquarium fish were designed to
evaluate hobbyists' preferences. One of the selected specimens was the blue-faced
angelfish (Pomacanthus xanthometapon), which is a high-value specimen that is
primarily collected from the wild. The second experiment estimated the importance of
fish characteristics on preferences for a more affordable and popular fish, the maroon
clownfish (Premnas biaculeatus), which is both collected from the wild and cultured in
The next important aspect to be addressed was the selection of the fish attributes.
Price, certification status, source and extent of "arrive alive-stay alive" guarantees were
selected based on previously identified fish characteristics influencing preferences for
marine aquarium fish at the wholesale and retail level (Rubinstein 2003; Larkin and de
Bodisco 2003). Price was defined to have three levels for each species in order to capture
potential non-linear effects (Rubinstein 2003). Price levels are based on actual retail
prices obtained from several on-line saltwater fish stores the week prior to the survey.1
Given that fish can be sold at different prices depending on size, the recorded price level
corresponded to the average price for a medium-size specimen of each species, defined in
inches. The alternative price levels were set at approximately 15%-30% above and below
the average recorded prices, which were translated into price premiums as high as $3 and
$15 for the maroon clownfish and blue-faced angelfish respectively. The remaining
attributes each had two discrete levels. In terms of source, a fish was defined as either
collected from the wild or tank-bred. The two levels of survival guarantee were set at 5
days and 14 days, which represents the length of time that fish are guaranteed to remain
alive after purchase (provided handling instructions are followed). If the fish dies in the
hobbyist's aquarium during the guarantee period, the hobbyist gets a replacement fish or
1 Prices were obtained from www.saltwaterfish.com, www.thereefconnection.com, www.aquacon.com,
www.petworldfish.com, www.interocean.us, www.etropicals.com, www.emarinefish.com,
www.liveaquaria.com, www.marinedepotlive.com, www.aquaticretail.com, www.jlaquatics.com,
www.thepetstop.com, www.aquariumstuffers.com and www.themarinecenter.com.
money back. Lastly, fish were characterized as either MAC certified or uncertified (i.e.,
ecolabelled or not).
Each hypothetical product was defined using only three of the four attributes in
order to reduce the choice task of respondents. The objective of the maroon clownfish
experiment was to evaluate the influence of price, fish source and certification status on
stated preferences. The blue-faced angelfish exercise featured certification status, price
and length of survival guarantee as the characteristics that explain stated preferences.
Considering the number of attributes and levels defined for the exercises, a
universe of 31*22 combinations comprised the complete factorial for each experiment.
These factors correspond to one attribute (price) with 3 levels and to two attributes with 2
levels. To minimize the magnitude of the task facing respondents, the smallest number of
orthogonal (i.e., independent) profiles was selected using the 12 (i.e., 3 1*22) possible
profiles for each exercise (Mazzanti 2001). This saturated design (i.e., smallest
orthogonal array) consisted of five profiles and was generated using the SAS statistical
program. Those generic profiles were later populated with the appropriate attributes and
levels for each experiment. It is important to mention, that one of the selected profiles in
each experiment featured a dominant attribute combination (i.e., an alternative that was
highly likely to be chosen since it featured the lowest cost and higher levels of the
remaining attributes) that was excluded from the set of alternatives. Elimination of
dominant profiles is assumed not to affect the orthogonality of the design if few profiles
are eliminated, and eliminations may also contribute to maintaining the efficiency of the
design (Blamey et al. 1999; Irwin et al. 2002). Ultimately, four profiles formed the
combinations of attributes and levels used in the construction of choice sets.
Choice sets were constructed by combining the original four orthogonal profiles in
all their feasible pairs (Bunch et al. 1996). Through this method, six choice tasks of two
alternatives were obtained for each experiment. These six choice tasks for each
experiment were presented to each respondent for analysis (i.e., each respondent faced a
total of 12 choice tasks, 6 for each species). In order to test our original assumption
regarding dominant profiles in regards to the price and source attribute levels for the
maroon clownfish experiments, four choice sets were modified by varying the level of
source during the survey period. This approach generated four new choice sets that were
treated as a second version of the originals. Table 3-1 shows all hypothetical fish bundles
used in this discrete choice analysis. The original versions were shown to a portion of
respondents while the second versions were shown to the remaining respondents. In this
way, each individual faced a total of six choice tasks for this experiment.
Table 3-1. Definition of experimental fish profiles by species
Species Profile Price MAC Source Survival
no. ($/fish) Certified Guarantee
Angelfish F 99.99 No Wild-caught 14
J 84.99 No Wild-caught 5
H 99.99 Yes Wild-caught 5
M 114.99 Yes Wild-caught 14
Clownfish P 24.99 No Wild-caught 5
T 21.99 No Tank-bred 5
Q 24.99 No Tank-bred 5
U 21.99 No Wild-caught 5
N 24.99 Yes Tank-bred 5
G 24.99 Yes Wild-caught 5
R 27.99 Yes Wild-caught 5
S 27.99 Yes Tank-bred 5
Note that both experiments evaluated different options for a same product rather
than estimating market or participation shares. That is, respondents were instructed (and
it was assumed) they would only buy one fish. Therefore, an opt-out option was not
considered applicable and was not included in the design, although respondents were not
prevented from leaving the choice blank.
The sample frame for this study consisted of aquarium hobbyists that were
members of pre-identified online aquarium discussion boards recommended by John
Brandt, member of the 2003-04 board of Directors for the Marine Aquarium Societies of
North America (MASNA) as the Industry/Legislation/Ocean Conservation representative.
Mr. Brandt is also a member of the U.S. Coral Reef Task Force. Specifically, the
targeted respondents were members of the discussion boards of Reef.org, Reef Central,
MASNA, Chicago Marine Aquarium Society (CMAS), Boston Reefers Society and
Saltwater Enthusiasts Association of the Bay Area (SEABay). This convenience
approach for selecting a sample frame is based on cost-effectiveness. Alternative
approaches to select participants were evaluated, but they were not feasible given
resource constraints.2 For empirical purposes, the study examines only individual tank
owners. No aquarium organizations were included in the sample.
Because it was not possible to identify individual sample units (hobbyists), the
sampling technique is most appropriately considered a convenience sample. With this
approach, respondents will be tank owners that visited certain pre-identified websites
2 Randomly sampling approaches would involve screening households for individuals that keep marine
aquaria. Sample frames for the general population based on household information are widely used and
available in survey research (e.g. phone directories, urban /rural mapping). However, this approach would
be very inefficient given the relatively low incidence of marine (saltwater) aquarium ownership in the
general population). The cost in terms of time, effort and money would likely exceed the value and
comprehensiveness of the information obtained. Other options include searching and purchasing existing
databases of aquarium hobbyists maintained by commercial marketing firms; however, one for marine
aquarium hobbyists could not be found. Alternative sources include customer lists from retailers of
saltwater specimens in the U.S. However, due to resource constraints and the difficulty of accessing these
lists for customers located throughout the U.S., these options were not feasible.
during a certain period and voluntarily choose to complete the survey. The main
implication of this approach will be a potential bias toward highly interested individuals,
which means results from the survey will be specific to this market segment. Another
concern is that a balanced or properly stratified sample is not guaranteed, so statistical
inferences about the population of individual tank owners (hobbyists) across the United
States is not possible. However, results can be applied to infer patterns in preferences for
the specific segment of active and informed hobbyists that is sampled. These results can
be of importance to industry members and to related organizations like the MAC.
Due to the nature of the target sample, data collection was implemented through an
Internet-based questionnaire. In comparison to traditional methods of questionnaire
implementation, the use of the Internet provided more rapid response, wider geographic
coverage and lower costs of implementation (Dillman 2000). In addition, it eliminates
the possibility of data entry errors as responses are automatically recorded.
Two weeks before the implementation of the survey and the first contact with
potential respondents, a pre-test was conducted with saltwater aquarium hobbyists known
to the researches, managers of the selected online discussion boards, and a few members
of the food and Resource Economics Department and the Fisheries and Aquatic Sciences
Department at the University of Florida. The test resulted in a change in species used,
additional questions about familiarity with the species used in the experiment, validation
of the prices used and corrections to some wording. Ultimately, the survey instrument
was posted on agsurveys.org and opened to participants between February 18 (8:00 p.m.)
and February 25 (noon), 2004.
The survey instrument contained four sections (Appendix A). The first section
addressed information about characteristics of hobbyists' aquaria, aquarium purchases,
environmental and aquarium interests, and familiarity with the MAC certification
program. The next section included the two choice experiments. Before showing the
actual choice sets, a brief definition of the specimens, attributes and levels, as well as a
brief overview of the choice modeling approach were presented. The blue-faced
angelfish was defined as being wild-caught from the Philippines. The maroon clownfish
was defined as being from Indonesia if wild caught. Both countries are known to have
collection practices that harm marine ecosystems according to the group of pre-testers
and, for example a study by Bunting and Meyers (2002). Next, a question about the level
of familiarity with each specimen was asked and instructions were provided on how to
evaluate the choice tasks. Lastly, the 12 choice sets were presented, one after another on
separate pages. The third section assessed socio-demographics, opinions regarding tank-
bred and reef-safe (i.e., wild-caught and ecolabelled) specimens, and alternative
production techniques for unique specimens. Finally the fourth section solicited
voluntary comments about the survey or subject matter in general.
Regression analysis is used to estimate the effects of attributes and individuals
characteristics on purchase decisions for saltwater ornamental fish in this study.
Following Random Utility Theory (RUT), the utility generated by fish alternative j in
choice set c is modeled nearly according to the following general specification:
Vjn = Pj + kkPkATTjk +LiYiATTjk-MAC DEMOGin+Xma iATTjk=MAC INTmn,
where k=l,...K fish attributes, i=l,......I respondent demographic characteristics,
m=l,...M respondent interest characteristics, and n=l,....N individuals. Vjn is the
observable (i.e., deterministic) component of the overall utility obtained from alternative j
by individual n. ATTjk is the vector representing fish attribute and attribute-level
combinations in alternative j. ATTjk=MAC is the vector representing the level of the MAC
status attribute in alternative j. DEMOGin is the vector of demographic characteristics of
respondent n. INTmn is the vector representing opinions and interests of individual
hobbyist n. To account for preferences heterogeneity across the six choice tasks in each
experiment, the utility model includes an additive specification of the effects of
individual characteristics on MAC purchase preferences, reflected by the interaction
terms ATTjk-MAC DEMOGin and ATTjk=MAC INTmn. The coefficients P3k, yi and um
represent the magnitude and direction of the influence of each vector on the overall utility
generated by fish alternative j and J3j is the intercept term unique to each fish alternative j.
Fish alternative j would be chosen if it provides a higher utility, this is, if Vj > Vi,
where 1 refers to another alternative in choice set c. Assuming a normal distribution for
the error term, and given a paired choice set approach, the probability of fish j to be
chosen over fish 1 can be modeled by a binary probit model. The binary probit model is
selected in this analysis in order to avoid restricted assumptions about uncorrelated
alternatives arising from IID models (e.g., logit models) and to overrule IIA assumptions.
In addition, probit computation is relatively simple in binary choice cases.
The resulting probability of choice of fish alternative j in choice set c is expressed
Prob (Vj > Vi)= 4 [Vj], or
Prob (Vj >Vi)= Pfj + k3kPkATTjk +LiYiATTjk-MAC DEMOGin+Lma iATTjk=MAC
where 4 represents the normal cumulative distribution function (Liao 1994). The
dependent variable in equation (3.2) corresponds to the observed choice for fish
alternative j by respondent n, which is coded as 1 for the selected alternative and 0
otherwise. The independent variables correspond to the variables defined in the utility
specification in equation (3.1). This probit model is estimated using maximum likelihood
The statistically significant coefficient estimates obtained from estimation of
equation (3.2) for each species will be used to evaluate the effect of attribute levels on
probability of purchase for both species.
Results account only for respondents that completed sections one to three of the
entire survey instrument, that is, with the exception of the final comments page. From a
total of 666 respondents who started answering the first "page" of the first section of the
questionnaire, only 615 individuals continued to the end of the third section. After
excluding incomplete questionnaires, replications (as identified through identical IP
addresses) and submissions from respondents less than 18 years old, 546 completed
surveys remained for analysis.
The presentation of results begins with characteristics of the respondents and their
level of knowledge of the industry and opinions regarding wild-caught and tank-bred
species. Following the characterization of respondents, the discussion of the findings
obtained from the two choice experiments is introduced. For both species, estimates of
the importance of fish attributes and statistically significant individual characteristics on
overall utility are discussed first. Those estimates are then used in model simulations in
order to analyze the probability of purchase related to different levels of attributes and the
effect that individual characteristics have on preferences for the MAC certification
The majority of respondents are owners of large saltwater tanks (Figure 4-1).
About 50% of respondents keep saltwater tanks with a capacity of at least 100 gallons.
Around 45% of respondents maintain 25 to 100 gallon tanks, while only 5% keep tanks
of 25 gallons or less. In addition, respondents were asked about the size of any
freshwater tanks they keep. Responses indicated that 34% also have a freshwater tank.
y 30% 25% 25%
o 20% 16% 17%
& Less 10-24 25-49 50-74 75-99 100- More
than 180 than
10 Num ber of gallons 180
Figure 4-1. Current saltwater tank capacity reported by respondents (n = 539)
The majority of respondents (64%) indicated plans to expand their current saltwater
tank capacity within the next two years (Figure 4-2). Only 2% were considering reducing
Figure 4-2. Expected change in saltwater tank capacity within the next 2 years of
respondents (n = 546)
Most respondents can be classified as "intermediate length" hobbyists. Results
showed that 82% have been involved in the hobby for 10 years or less, but only 17%
have kept tanks for less than a year.
Respondents were asked where they expended the most money on saltwater
aquarium organisms. About 65% prefer local aquarium stores and 22% spent the most on
purchases through the Internet. Local general pet stores were preferred sources for 7% of
Findings indicate that respondents are willing and able to pay high prices for
aquarium organisms (Figure 4-3). About 57% have paid between $25 and $75 for a single
saltwater finfish, and 16% have paid between $75 and $100. Moreover, 18% have paid
more than $100 for single specimen. Respondents were also asked about the highest
price ever paid for a single invertebrate. About 46% of respondents have paid between
$25 and $75 and 19% have spent more than $100 for a single live invertebrate organism.
These preliminary observations indicate that price should not be a major constraint on
purchase decisions for this group of hobbyists.
S 30% 27%
0 10% 7%
CL No Lessthan $10-less $25-less $50-less $75-less Morethan
purchase $10 than$25 than$50 than$75 than$100 $100
Figure 4-3. The highest price ever paid for a single fish by respondents (n = 544)
The majority of respondents reported high levels of involvement in the aquarium
hobby (Figure 4-4). Around 76% of them consider keeping marine aquaria as their
primary hobby and 59% are members of an aquarium society. About 19% of respondents
are members of environmental organizations. In addition, 73% and 88% of participants
have conducted their own breeding and research on live marine organisms, respectively.
This indicates the group of respondents as a whole has a relatively high level of
involvement and interest in the hobby and also a high level of specialized knowledge.
Ninety-five percent of respondents preferred purchasing a tank-bred fish to a wild-caught
one. In addition to keeping a reef tank (91%), 63% of respondents have visited a coral
reef. These observations imply that these respondents are most appropriately considered
as avid hobbyists with a concern for the environmental effects associated with collection
Buy tank-raised fish 95'
Keep reef tank 91%
Own species research 88%
Primary hobby 76%
Own tank breeding 73%
Have visited coral reef 63%
Member aquarium society 59%
Keep "fish only" tank 123%
Member environmental group # 19%
0% 20% 40% 60% 80% 100%
Figure 4-4. Incidence of environmental and aquarium interests of respondents (n > 541),
percent answering "yes"
The extent of the respondents' knowledge of the MAC certification program for
marine ornamental fish was also assessed (Figure 4-5). Although respondents seem to be
particularly informed about issues related to their hobby, almost half (49%)of them were
not familiar with the MAC ecolabeling program. Although the remaining 51% indicated
some level of familiarity, only 11% were very familiar with the program.
Somew hat familiar
Figure 4-5. Level of familiarity with the MAC certification program among respondents
(n = 545)
The importance of several issues related to reef ecosystems and specimen condition
were analyzed in the context of two different sources: tank-bred and MAC-certified
(ecolabelled) fish (Figures 4-6 and 4-7, respectively). Respondents were asked to rate five
issues on a scale from not compelling at all to very compelling.1 Tank adaptability of
specimens was the most important reason to buy a cultured fish according to the
respondents (i.e., 95% considered this as a compelling reason). Next, in order of
importance, 94% and 86% of respondents indicated that buying a cultured fish prevented
damage to coral reefs and sustained wild stocks, respectively. These findings suggest
that the majority of respondents perceive wild-caught fish as less robust than cultured
ones, probably due to an association with harmful collection methods. It is also
suggested that respondents may actually buy cultured marine ornamentals in order to
protect marine ecosystems. A potentially greater availability of cultured marine fish and
the presence of regulations to protect the environment surrounding aquaculture facilities
1 Each issue was assigned scores ranging between -2 and +2 with 0 representing no effect on purchase
appear to be less compelling reasons to buy tank-bred fish (i.e., they accounted for 60%
and 59% of respondents, respectively).2
0% 20% 40% 60% 80% 100%
Reef damage prevention
Wild stock sustainability
Not compelling Indifferent oE Highly Compelling
Figure 4-6. Distribution of responses regarding the importance of reasons to buy a tank-
bred fish (n > 539)
The main reasons to buy MAC-certified saltwater fish were protection of coral
reefs and sustainability of wild stocks according to 86% and 85% of respondents,
respectively (Figure 4-7). These two reasons were followed in importance (on average)
by the possibilities that certified fish are healthier and that certified commercialization
paths prevent unnecessary mortality during collection and transport. In contrast, the
environmental success of the Marine Stewardship Council (MSC) seafood certification
program, the predecessor of the MAC program, appears to have less influence on
purchases of certified fish (i.e., only 58% of respondents considered it as compelling).
These observations also suggest a strong level of concern about reefs and wild stocks and
a strong perception that wild-caught fish are harmed during collection and subsequent
2 These percentages are the ratio between the number of respondents that rated the issues as +1 or +2 and
the total number of respondents assessing each issue.
handling. Overall, the protection of marine ecosystems by the MAC certification
program seemed very important to respondents.
0% 20% 40% 60% 80% 100%
Reef damage prevention
Wild stock sustainability
Seafood label success
^ ~58% ~
Not compelling 0 Indifferent [3 Highly compelling
Figure 4-7. Distribution of responses regarding the importance of reasons to buy a MAC
certified fish (n > 539)
Other interesting findings include the relationships between associating coral reef
protection and wild stock sustainability with a MAC certified fish and the level of
familiarity with the MAC program (Figures 4-8 and 4-9, respectively). Participants that
were only somewhat familiar with the MAC program were most likely to associate coral
reef protection and sustainability of wild stocks with MAC certification. These
observations indicate that, although most participants associate protection of reefs and
wild stocks with the MAC ecolabel, respondents who are most familiar with the MAC
certification program have a reduced perception that the program is effective in
addressing these two considerations.
60% 85% 91%
Not familiar at all Somewhat familiar Very familiar
n Not compelling 0 No influence [ Highly Compelling
Figure 4-8. Association of coral reef protection with MAC certification by MAC
familiarity level, on average across respondents (n = 541)
60% 85% 88%
Not familiar at all Somew hat familiar Very familiar
SNot compelling ENo influence Highly Compelling
Figure 4-9. Association of wild stock sustainability with MAC certification by MAC
familiarity level, on average across respondents (n = 541)
Participants were asked to rate the market potential of several techniques for
producing unique marine ornamentals (Figure 4-10). Although data indicate an above
average level of sophistication for respondents, these results indicate preferences for
traditional production methods. Around 70% of respondents viewed improved diets as
having the highest potential to produce marketable marine ornamentals. Around the
same share of respondents ranked selective breeding as having a positive potential. In
contrast, innovative techniques, such as the use of hormones and genetic modification,
were considered to have a positive market potential by only 9% and 13% of respondents,
Although the majority of respondents (67%) viewed genetically modified
specimens as having little market potential, 50% of them said they would be somewhat
likely to purchase such specimens. However, the question about likelihood of purchase
used the word "transgenic", so it is possible that respondents do not really know what this
expression means. About 27% of respondents were not likely to purchase transgenic fish,
18% were very likely to do so and around 5% claimed they were undecided.
0% 20% 40% 60% 80% 100%
High potential Moderate potential Low potential
Figure 4-10. The perceived market potential of alternative production techniques that can
be used to produce unique marine ornamental species by respondents (n >
Respondents have homogeneous demographics especially in regards to sex, age,
education and income levels. The majority of respondents were males (i.e., 90%). Most
can be characterized as young or middle-aged individuals as 72% were born between
1960 and 1979. Overall, 39% of respondents were between 24 and 34 years old, 33%
were between 34 and 44, 19% were older than 44. Only 10% were younger than 24
1950-1959 1960-1969 1970-1979 1980- Feb 1986
Figure 4-11. Distribution of year of birth of respondents (n = 546)
About 66% of respondents had some college or a college degree and an additional
24% had some graduate school training (Figure 4-12). Respondents also tended to have
high-income levels. About 43% of participants had annual household incomes (before
taxes) between $50,001 and $100,000, and 35% of respondents had household incomes
of more than $100,000 per year (Figure 4-13).
Less than High school
high school graduate
Some College Some Graduate
college graduate graduate degree
Figure 4-12. Distribution of the highest level of education of respondents (n = 538)
S25% 22% 21%
10% 5% 7%
Less than $25,001 $50,001 $75,001 $100,001 At least
$25,000 $50,000 $75,000 $100,000 $199,000 $200,000
Figure 4-13. Distribution of annual gross income of respondents (n = 537)
In terms of geographic distribution, 90% of the surveys came from hobbyists
currently living in the United States (Figure 4-14A). Around 7.5% of the remaining
participants reside in Canada and 2.5% were from Mexico and other countries. Within
the U.S., 54% of participants reside in the Northeast and the Great Lakes regions, 34%
came from the Southeast (including Puerto Rico) and the Pacific (including Hawaii), and
12% came from Alaska, the Southwest and the Mountain-Prairie regions (Figure 4-14B).
Canada Mexico Cther 34%
8% 0% 2% 35%/o
Northeast Great Lakes- Southeast & Pacifc & Suttwest IVburtain Alaska
Bg Rvers Ro Rco Hawaii Rarne
Figure 4-14. Geographic distribution of respondents by region (A) and within the United
Respondents live in large small communities (Figure 4-15). About 40% live in
communities with more than 500,000 inhabitants and 39% reside in areas with less than
100,000 inhabitants. In total, 21% of respondents reside in communities of between
100,001 and 500,000 people.
15% 13% 13% 13%
Less than 25,001 50,001 100,001 250,001 500,001 More than
25,000 50,000 100,000 250,000 500,000 1,000,000 1,000,000
Figure 4-15. Distribution of respondents by size of the communities where they reside
According to the responses, the hobbyists surveyed in this research represent a
homogeneous segment of the market. They are relatively young, and highly interested in
the hobby. They are aware of, and give considerable importance to, protection of coral
reefs and wild stocks. They have limited familiarity with the MAC certification program
and above average levels of education and income. All these characteristics play an
important role in the framework for analyzing purchase behavior
Before respondents were presented with the choices, they were asked to rate their
level of familiarity with both species on a scale ranging from not at all familiar to very
familiar. The majority of respondents had limited familiarity with the blue-faced
angelfish, with 49% being somewhat familiar and 33% being not at all familiar (Table 4-
1). On the other hand, the majority of respondents (i.e., 65%) were very familiar with the
maroon clownfish and less than 2% knew nothing about it (Table 4-1).
Table 4-1. Respondent familiarity with fish species used in market experiments
Fish familiarity level Blue-faced angelfish Maroon clownfish
Not at all 33.03% 1.29%
Somewhat 49.26% 33.70%
Very 17.71% 65.01%
n = 542 for blue-faced angelfish and 543 for maroon clownfish.
The rate of non-response for any choice set in the two exercises was never higher
than 2% (Table 4-2). In terms of measuring indirect utility, this result shows that only a
small percentage of respondents did not derive any utility from the attributes and attribute
levels presented. Therefore, the lack of a no-choice option probably did not distort
estimates of preferences for this group of hobbyists.
Table 4-2. Distribution of responses by species and choice set
Species Set N Alternative 1 Alternative 2 No-choice
Angelfish 1 546 F 68.68% J 30.59% 0.73%
Angelfish 2 546 J 30.95% H 68.13% 0.91%
Angelfish 3 546 H 63.37% F 36.08% 0.55%
Angelfish 4 546 M 55.86% J 43.41% 0.73%
Angelfish 5 546 H 57.33% M 41.94% 0.73%
Angelfish 6 546 M 60.26% F 39.01% 0.73%
Clownfish la 339 P 2.36% T 96.76% 0.88%
Clownfish lb 207 Q 95.17% U 4.35% 0.48%
Clownfish 2 546 T 50.73% N 48.17% 1.10%
Clownfish 3a 339 N 95.58% P 3.83% 0.59%
Clownfish 3b 207 G 16.43% Q 83.09% 0.48%
Clownfish 4a 339 R 9.73% T 89.97% 0.29%
Clownfish 4b 207 S 88.89% U 9.18% 1.93%
Clownfish 5a 339 N 93.81% R 5.01% 1.18%
Clownfish 5b 207 G 8.20% S 91.30% 0.50%
Clownfish 6 546 R 76.00% P 23.00% 1.00%
One salient fact is, in comparison to the blue-faced angelfish experiment, the
distribution of responses for the hypothetical clownfish show a strong bias towards
alternatives featuring a tank-bred alternative. Since this one attribute dominated choices
for 80% to 90% of respondents in this experiment, the effects of the remaining attributes
could be confounded with the effect of the tank attribute. Therefore, the modeling and
analysis for this experiment will need to take this situation into account.
Choice observations for both experiments were modeled through binary probit
specifications, using a maximum likelihood approach. Data were analyzed using TSP
(Time Series Processor) statistical software version 4.4.
Maroon Clownfish Results
Two choice observations coded as 1 and 0 from each of the 546 respondents to the
six-paired choice sets were obtained. Therefore, a total of 6,552 observations were
computed and analyzed for this experiment.
First, a basic pooled probit model representing only the main effects and
interactions of the 3 attributes used in this experiment (i.e., certification status, source and
price) was defined. Interactions among attributes were introduced to capture the
conditional effects of price and certification status given the presence or absence of the
After accounting for effects of attributes exclusively, selected characteristics of the
respondents were introduced into the basic model in the form of interactions with the
certification attribute in order to capture potential preference heterogeneity with respect
to certification (Bennett and Blamey 2001). Such interactions would indicate how these
characteristics modify the estimated effect of MAC certification on preferences for
marine ornamental fish. Dummy variables indicating attitudes, such as if the respondent
was at least somewhat familiar with the MAC certification program, at least somewhat
familiar with the featured species or considered prevention of coral reef damage and
overfishing as important reasons to buy certified fish, were also introduced into the
model. Additional dummy variables were used for income, age, geographic regions and
Since the total data set consists of a series of observations from each respondent
and to further account for heterogeneity, a test of random effects usually implemented in
panel data with dichotomous dependent variables was conducted. With this procedure,
the stochastic component (i.e., error term) of the indirect consumer utility consists of the
traditional error term associated with each observation in the data set and an additional
term capturing any potential variation across individual respondents (Kennedy 1994).
Since the correlation between individuals and the error term was not statistically
significant, this document reports the results of the pooled probit model, where any
preference heterogeneity is captured by the demographic-based interaction variables
According to economic theory, the effect of price on purchase decisions was
expected to be negative. However, due to the luxury nature of the product examined (i.e.,
saltwater aquarium keeping) that negative effect was not expected to be large. It was also
hypothesized that MAC environmental certification would have a positive effect on the
utility consumers obtain from buying a saltwater ornamental fish, because of the widely
known concerns about damage to coral reefs and overfishing associated with collection in
the wild. Due to these concerns, it was expected that collection from the wild as a source
would have a negative effect on purchase decisions and tank-raising would exhibit a
positive influence on choices.
3 Alternative specifications expanding the basic model by the inclusion of just respondents' attitudes
interacting with certification and the further inclusion of respondents' characteristics without interactions
were also considered. Likelihood ratio tests indicated that the chosen model had the best goodness of fit.
Table 4-3. Definition of probit model explanatory variables for maroon clownfish choice
PRICE $21.99; $24.99; $27.99
CERT 0 if the fish is not MAC certified,1 if the fish is MAC certified
TANK 0 if the fish is wild-caught, 1 if the fish is tank-raised
PRCR Interaction between price and certification levels
PRTK Interaction between price and type of fish source
CRTK Interaction between certification levels and type of fish source
CRTKPR Interaction between price, certification levels and type of fish source
MCFAMI 1 if respondent was familiar with clownfish and the fish offered was
certified, 0 otherwise
MACFAMI 1 if respondent was familiar with the MAC certification program and the
fish offered was certified, 0 otherwise
RFCOMPI 1 if respondent considered that coral reef damage prevention was an
important reason to buy a certified fish and the fish offered was certified,
FHCOMPI 1 if respondent considered that wild stock protection was an important
reason to buy a certified fish and the fish offered was certified, 0
AGE2I 1 if respondent was 24-34 years old and the fish offered was certified, 0
AGE3I 1 if respondent was 34-44 years old and the fish offered was certified, 0
AGE4I 1 if respondent was over 44 years old and the fish offered was certified, 0
PACIFI 1 if respondent was from the U.S. Pacific region and the fish offered was
certified, 0 otherwise
SEI 1 if respondent was from the U.S. Southeast region and the fish offered
was certified, 0 otherwise
LAKENEI 1 if respondent was from the U.S. Northeast or Great Lakes region and the
fish offered was certified, 0 otherwise
SWMTALI 1 if respondent was from the U.S. Southwest or Mountain Prairie or
Alaska region and the fish offered was certified, 0 otherwise
DEDUCI 1 if respondent was at least a high school graduate and the fish offered
was certified, 0 otherwise
INCOM2I 1 if the annual income level of respondent was between $25,000 and
$50,000 and the fish offered was certified, 0 otherwise
INCOM3I 1 if the annual income level of respondent was between $50,001 and
$75,000 and the fish offered was certified, 0 otherwise
INCOM4I 1 if the annual income level of respondent was higher than $75,000 and
the fish offered was certified, 0 otherwise
a See Appendix B for summary of descriptive statistics of the variables.
However, the negative effect of wild-caught as a source was expected to be
influenced by certification status. Since most of the unique, colorful fish sought by
hobbyists are currently collected from the wild, it was hypothesized that the presence of
certification could reduce the negative effect. This expectation is supported by the
premise that this hobby is a luxury activity, and therefore, the more unique and diverse
the good, the more utility consumers obtain from it.
It was also expected that respondents indicating at least some level of familiarity
with the MAC program and those associating marine ecosystem protection and wild stock
sustainability with the MAC ecolabel would display a positive preference for
certification. Since about 95% of respondents were familiar with the fish and it is
increasingly available from aquaculture, a high level of fish familiarity could be expected
to limit the positive influence of certification. The hypothesized limited effect of
certification in this case is based on the fact that to date the MAC certification program
does not apply to cultured specimens and respondents could not see a relationship
between the two attributes as a feasible option.
It was expected that respondents living in U.S. regions affected by the commercial
collection industry (i.e., revenues support local communities), such as in the Pacific and
Southeast, would show a positive preference for certification. In addition, higher age and
education levels were hypothesized to lead to a positive effect on preferences for
certification since higher sensitivity for environmental protection is likely to exist among
increasingly educated and beyond middle-age individuals. Finally, respondents with high
incomes were expected to exhibit a positive preference for certified fish because of their
ability to afford a premium for MAC ecolabelled specimens.
Both the basic and expanded models explain the relationship between dependent
and independent variables better than the model with only an intercept, as the LR
statistics for zero slopes indicate (Table 4-4).
Table 4-4. Probit model estimates for maroon clownfish
Variable Estimate t- value
C -3.86 *** -4.51
PRICE 0.10 ** 3.07
CERT -4.36 *** -3.72
TANK 0.26 0.24
PRCR 0.17 *** 3.84
PRTK 0.08 1.84
CRTK 3.22 ** 2.05
CRTKPR -0.15 ** -2.44
LR (intercept only) 2691.94
Correct predictions 78.49%0
*** Statistically significant at the 1% confidence level.
** Statistically significant at the 5% confidence level.
* Statistically significant at the 10% confidence level.
For the basic model (model 1) where only the effects of the attributes are assumed
to influence consumers' utility from choice decisions, 7 of the 8 coefficients are
statistically significant (i.e., different from zero) at the 10% confidence level. The effect
of the TANK variable (the only one not statistically significant) is; however, captured
with the 2-way and 3-way interaction terms in the model, which were statistically
significant.4 Therefore, looking only at the coefficient of tank will lead to an incorrect
interpretation of the effect of the tank-bred attribute. In general, given this model
specification, coefficients of the interacted variables complement each other and need to
be considered in that way in order to obtain a comprehensive interpretation of attributes'
influence on consumer's utility.
As suggested by the distribution of choices across choice sets, estimates show that
tank-bred has a positive impact on the utility derived by the hobbyist from the purchase
of such fish. Moreover, if the fish offered to the respondent is certified (i.e., ecolabelled),
tank-bred increases utility even more (i.e., the coefficient of CRTK is positive and
Contrary to expectations, MAC certification as a characteristic of wild-caught fish
appears to influence purchase decisions in a significantly negative way. The positive
coefficient of the interaction between source and certification shows that if the fish is
tank-bred the effect of ecolabeling on respondents' utility and purchase decisions would
remain negative, but at a much lower level (ceteris paribus). This result suggests that this
specific group of hobbyists has a negative perception of the MAC certification program
and that an ecolabel is not a compelling factor in purchase decisions. Open comments
4 Likelihood ratio tests were used to examine whether the interaction terms improved the explanatory fit of
from respondents collected at the end of the survey instrument provide some support for
this interpretation. About 6% respondents strongly stated their lack of faith in the MAC
program and a preference for tank-bred and raised fish as a means to avoid damage
caused by capturing wild fish (Appendix C).
One interesting and unexpected finding for these avid hobbyists is that higher
prices appear to increase the utility derived from buying a fish. This result is indicated by
the positive coefficient of the main effect (PRICE). The positive estimates from the two-
way interaction terms show that consumers would continue to increase their utility
regardless of price if the fish is wild-caught and certified or only tank-bred. However, if
a fish is certified and also tank-bred, incremental changes in price would negatively affect
consumers' purchase decisions (i.e., reduce the probability that the fish would be
purchased), as the negative coefficient of the three-way interaction term reflects.
According to these results, price did not appear to be an important factor when deciding
to choose a fish, but its effects were conditional on the source and certification status of a
fish. This particular situation could be related to the profile of this specific segment of
respondents. Descriptive statistics showed that the majority of them have above average
annual incomes and already paid more than $75 for a single fish.
After introducing interaction terms to capture the effects of demographics on
preferences for certified (ecolabelled) fish, the estimated attribute effects remained the
same. Results from model 2 show that only a few interactions had statistically significant
influences on certification preferences. The level of familiarity with the MAC
ecolabeling program (MACFAMI) and whether the respondent believed the program
could prevent reef damage and overfishing (RFCOMPI and FHCOMPI, respectively)
were the only three factors that were found to affect certification-related preferences.
The positive coefficient of the variable MACFAMI indicated that respondents who
were at least somewhat familiar with the MAC program had higher preferences for a
certified fish than those who were not at all familiar with the MAC ecolabel. Likewise,
the positive signs of RFCOMPI and FHCOMPI suggest that hobbyists who related reef
ecosystem and stock protection to the MAC ecolabel had higher preferences for a
certified fish. Moreover, estimates also indicated that the positive impacts of protecting
both coral reefs and wild stocks on preferences for the MAC ecolabel are of similar
As t-tests of statistical significance showed, and contrary to initial hypotheses,
higher incomes and education levels, along with being from a geographic region that
directly benefits from a collection industry were not found to influence preferences for
the MAC ecolabel for this segment of hobbyists.
Estimating Effects on Probability of Purchase
The relationships described so far correspond to the effects of attributes on the
latent indirect utility function of the homogeneous group of avid saltwater aquarium
hobbyists. To illustrate the effects of the attributes on the probability of purchasing a
fish, the estimated probit model was used in simulations where only the attribute levels
were systematically changed to isolate effects. To this end, a constant "base" product
was defined. The base price level was set at its mean, (i.e., $24.99), while certification
and source were set at zero to reflect an uncertified, wild-caught fish. The dummy
variables for familiarity with the fish and attitudes toward protection of reefs and wild
stocks were set equal to one, reflecting the average profile of respondents (i.e., the
majority were at least somewhat familiar with the maroon clownfish and considered reef
and wild stock protection as compelling reasons for buying a MAC-certified fish). The
remaining variables were set equal to zero implying that the "base" respondent did not
reside in the U.S., was not familiar with the MAC ecolabeling program, had only a high
school (secondary) education, earned less than $25,000 annually, and was from 18-24
years old. Using this base, product and hobbyists simulations identify the change in the
probability that a hobbyist would purchase the fish for each level of the statistically
significant variables, while keeping the remaining variables at the base level.
The probability of purchasing a maroon clownfish would overwhelmingly increase
if fish were tank-bred, regardless of certification status (Figure 16). The probability of
purchase increases from 12% to 89% for an uncertified fish and from 11% to 75% for
certified (ecolabelled) fish.
100% 100% -
bili 60% bili 60%
20% 11% 20% 12%
WId -caught Tank-bred WId-caught Tank-bred
Figure 4-16. Probability that the base hobbyist would buy a maroon clownfish by source.
A) Certified fish. B) Uncertified fish.
Contrary to expectations, results indicate that the ecolabel reduces the probability
of purchase for both levels of fish source, when price and other factors are kept constant
(Figure 4-17). However, the magnitude of the negative effect is much lower if the fish is
wild-caught (from 12% to 11% versus from 89% to 75% for a tank-bred maroon
clownfish). Overall, the highest probability of purchase was associated with an
uncertified tank-bred fish.
These results suggest that the MAC certification ecolabel is believed to improve the
environmental consequences of wild-caught collection activities by this particular group
of hobbyists. Moreover, these hobbyists would be more likely to buy an uncertified
maroon clownfish at the same price. Such an unexpected outcome is explained by
respondent comments. According to these comments (which are summarized in
Appendix C) and the simulations, it can be inferred that respondents have more
confidence in using culturing techniques as a replacement for wild-caught sources to
minimize coral reef damage and overfishing concerns associated with this hobby. In
terms of wild-caught fish, the estimated reduction in the probability of purchase is very
small, but the absence of a significant positive effect of certification strongly suggests a
lack of confidence in the effectiveness of the MAC ecolabel to ecosystems.
S- 80% 75%
60% 60% -
Figure 4-17. Probability that the base hobbyist would purchase a maroon clownfish by
certification status. A) Wild-caught fish. B) Tank-bred fish.
The positive relationship between price and the latent indirect utility function
indicates that the probability of purchasing a maroon clownfish increases with prices, but
varies depending on the certification status and source (Figure 4-18). Since the
experimental design only offered certified fish at higher prices, this result could be
reflecting an underlying preference for ecolabelled maroon clownfish.
For a wild-caught maroon clownfish, changes in the probability that the "base"
hobbyist would purchase the fish increased with price but differed considerably based on
certification status. Respondents were more likely to purchase a certified clownfish at
every price level but the probability increased at higher prices; the likelihood of purchase
reached 38% for a certified (ecolabelled) maroon clownfish at $3 above the average (i.e.,
24.99), which only represents a 12% price increase. For a tank-bred maroon clownfish,
the probability of purchase increases with price but at a diminishing rate. Over the range
of prices considered, the probability of purchase increases as much as 16 percentage
points for a certified maroon clownfish (i.e., from 75% to 91%).
10% ---... -...- ------ ...----..
0 0.50 1.00 1.50 2.00 2.50 3.00
Wild caught Certified
Tank-bred .......................... Uncertified
Figure 4-18. Probability that the base hobbyist would purchase a maroon clownfish by
price, certification status, and source
The results suggest that respondents would be likely to pay a premium for a wild-
caught ecolabelled maroon clownfish in order to obtain an indication of ecosystem
protection. Respondents also seem more willing to buy a tank-bred maroon clownfish at
any reasonable price because of the implied relative environmental benefits. However,
results suggest respondents are not willing to pay a premium for an ecolabel (i.e.,
environmental protection) if the maroon clownfish is tank-bred.
Influence of attitudes on certification preferences
The influence of hobbyist characteristics on preferences for certification was
measured through simulations of the statistically significant characteristics with respect to
base conditions as was done previously. As the MAC certification program currently
applies only to wild-caught fish, the analysis does not consider effects for tank-bred fish.
Results show that increasing the level of knowledge respondents have about the MAC
certification program increases the probability that the base hobbyist will buy a certified
maroon clownfish from 11% to 13% (Figure 4-19).5
If respondents indicated that coral reef damage prevention was not a compelling
reason to buy a certified fish, their likelihood of purchasing a certified maroon clownfish
fell from 11% to 8% (Figure 4-19). Likewise, if respondents indicated that protection of
wild stocks was not a compelling reason to purchase a MAC-certified fish, their
probability of buying an ecolabelled maroon clownfish fell from 11% to 7% (Figure 4-
19). Thus, prevention of overfish stocks had an equal effect as protection of coral reef
habitats to those buying a certified fish. This observation reinforces the ecosystem-
5 Although this analysis does not separate respondents who were somewhat familiar from those who were
very familiar, the estimated positive effect of "familiarity" on preferences for certified fish depended on
those hobbyists that were only somewhat familiar with the MAC program. Further analysis differentiating
the effects of these levels of familiarity would be useful to clarify the observed differences in preferences
for a MAC ecolabelled maroon clownfish.
friendly preference of the hobbyists surveyed in this study. However, these effects are
relatively small given the overall probability of purchase (i.e., an 11% base). Recall that
these respondents are much more likely to buy tank-bred species to ensure ecosystem
protection than to rely on the effectiveness of the MAC certification program. Currently,
the main focus of the MAC has been protection of reef ecosystems, but an effort to
increase the reliability and effectiveness of the MAC program's ability to efficiently
provide not only protection for coral reefs protection but also protection for fish stocks
could yield benefits in terms of increasing the probability of purchase.
Certified Familiar with MAC program Unconvinced that certification Unconvinced that certification
w ill protect coral reefs w ill protect w ild stocks
Figure 4-19. Probability that the base hobbyist would purchase a certified wild-caught
maroon clownfish by respondent familiarity with MAC and beliefs regarding
Considering all these results, it is concluded that education efforts focusing on
increased public knowledge of the program's purpose, scope of action and proof of
effectiveness in preventing damage to marine ecosystems and fish stocks would likely
improve preferences for ecolabelled maroon clownfish to this segment of hobbyists.
Blue-Faced Angelfish Results
The blue-faced angelfish experiments examined the effects of certification status,
length of survival guarantee, and price characteristics of the fish, in addition to the
respondent characteristics and beliefs examined for the maroon clownfish.
Four probit models were specified to analyze the choice behavior observed from
Model Imeasured the likelihood that a hobbyist would pay a 15%-17% premium
for a certified fish and how that likelihood would change with an extended survival
guarantee and significant individual characteristics. Model 2 measured the likelihood
that a hobbyist would pay a 15%-17% premium for an extended survival guarantee and
how that likelihood would change with MAC certification status. Model 3 measured the
likelihood of purchasing a wild-caught blue-faced angelfish from Indonesia and how that
likelihood would change with an extended survival guarantee, assuming constant prices.
Model 4 measured the likelihood that a hobbyist would pay for both certification and an
extended survival guarantee.
Models 1 and 2 included applicable fish attributes and respondents' characteristics
as explanatory variables. Models 3 and 4 included the respondent characteristics as
explanatory variables since all the fish attributes were used to define the dependent
variable. This analysis is different than the one presented in the previous section since
the variation among attribute levels designed for this experiment was limited, in
comparison to the maroon clownfish experiment. Two observations from each
respondent were used for models 1 and 2 (i.e., 1,092 in each model), while only one
observation was used for models 3 and 4 (i.e., 546 in each model).
The explanatory variables used in models 1-4 are summarized in Table 4-5.6 The
same hypotheses described in the maroon clownfish analysis were applied to the blue-
faced angelfish results.
Table 4-5. Definition of probit model explanatory variables for the blue-faced angelfish
Parameter Description a
CERT 0 if the fish is not MAC certified, 1 if the fish is MAC certified
GUAR 0 if the fish has a 5-day survival guarantee, 1 if the fish has a 14-day survival
BAFAMI 1 if respondent was familiar with the Blue-faced angelfish and the fish offered was
certified, 0 otherwise
MACFAMI 1 if respondent was familiar with the MAC certification program and the fish
offered was certified, 0 otherwise
RFCOMPI 1 if respondent considered that coral reef damage prevention was an important
reason to buy a certified fish and the fish offered was certified, 0 otherwise
FHCOMPI 1 if respondent considered that wild stock protection was an important reason to
buy a certified fish and the fish offered was certified, 0 otherwise
AGE2I 1 if respondent was 24-34 years old and the fish offered was certified, 0 otherwise
AGE3I 1 if respondent was 34- 44 years old and the fish offered was certified, 0 otherwise
AGE4I 1 if respondent was over 44 years old and the fish offered was certified, 0
PACIFI 1 if respondent was from the U.S. Pacific region and the fish offered was certified,
SEI 1 if respondent was from the U.S. Southeast region and the fish offered was
certified, 0 otherwise
LAKENEI 1 if respondent was from the U.S. Northeast or Great Lakes region and the fish
offered was certified, 0 otherwise
SWMTALI 1 if respondent was from the U.S. Southwest or Mountain Prairie or Alaska region
and the fish offered was certified, 0 otherwise
DEDUCI 1 if respondent was at least a high school graduate the fish offered was certified, 0
INCOM2I 1 if the annual income level of respondent was between $25,000 and $50,000 and
the fish offered was certified, 0 otherwise
INCOM3I 1 if the annual income level of respondent was between $50,001 and $75,000 and
the fish offered was certified, 0 otherwise
INCOM4I 1 if the annual income level of respondent was higher than $75,000 and the fish
offered was certified, 0 otherwise
a See Appendix B for summary of descriptive statistics of the variables.
6As in the previous experiment, alternative specifications expanded the basic model by including
respondents' attitudes in an interaction with certification. A likelihood ratio test supported the selection of
this specification. In addition, for models 1, 2 and 4, the interactions among respondents' characteristics
and certification are the same as including the characteristics as a main effect because the fish was assumed
to be certified.
All four models of consumers' indirect utility adequately explain the relationship
between dependent and independent variables better than the model with only an
intercept, as the LR statistics for zero slopes indicate (Table 4-6).7 Overall, familiarity
with the species was never statistically significant, indicating that it has no effect on the
probability that a blue-faced angelfish will be purchased. This suggests that hobbyists do
not seek at specific species to buy.
The variable representing the length of survival guarantee in model 1 is associated
with a negative coefficient that is significantly different from zero at the 1% confidence
level (Table 4-6). Contrary to expectations, this indicates that the extent of a survival
guarantee can have a negative influence on the preference for a certified fish. In other
words, if a certified fish were offered with an additional 9 days of survival guarantee (i.e.,
14 days), consumers would reduce their willingness to pay a 15%-17% premium. This
result suggests an association between the physical condition of a blue-faced angelfish
and certification. Respondents appear to perceive the MAC ecolabel as already providing
a healthier fish, as such the ecolabel may be considered a substitute for a longer survival
Results also show that respondent characteristics, such as familiarity with the MAC
program, confidence that the MAC ecolobel can provide ecosystem protection, location
of residence, income and age also affect consumers' choices (Table 4-6).
7 Attribute-only specifications were first estimated for models 1 and 2. As in the maroon clownfish
analysis, those basic models were rejected via Likelihood ratio tests in lieu of the "expanded" models
presented in Table 4-6.
Table 4-6. Probit model estimates for the blue-faced angelfish
Model 1 Model 2
C -1.36*** -4.52 0.49 ***
CERT -0.73 *
GUAR -0.25 *** -3.00
BAFAMI 0.05 0.55 -0.04
MACFAMI 0.65*** 7.41 0.33 ***
RFCOMPI 0.82 *** 4.76 0.22
FHCOMPI 0.50*** 2.97 0.32
AGE2I 0.12 0.79 -0.02
AGE3I 0.13 0.87 -0.25
AGE4I 0.38** 2.22 0.08
PACIFI -0.42 ** -2.33 -0.04
SEI -0.38 ** -2.14 -0.24
LAKENEI -0.38 ** -2.40 -0.33 *
SWMTALI -0.42 ** -2.19 -0.56 **
DEDUCI 0.23 1.65 0.06
INCOM2I 0.48 ** 2.16 -0.13
INCOM3I 0.49 ** 2.19 -0.48
INCOM4I 0.34 1.60 -0.30
N 1040 1066
LogL -576.94 -673.89
LR (intercept only) 203.60 116.90
Scaled R2 0.19 0.10
Correct predictions 71.82% 65.75%
*** Statistically significant at the 1% confidence level.
** Statistically significant at the 5% confidence level.
* Statistically significant at the 10% confidence level.
-1.07 *** -2.63
As hypothesized, results show that those respondents who were at least somewhat
familiar with the MAC program and/or that associated this program with effective coral
reef and wild stock protection were more likely to buy a certified fish, as indicated by the
statistically significant positive coefficients.
According to expectations, respondents over 44 years old with an annual income of
between $25,000 and $75,000 were more likely to choose a certified blue-faced angelfish
and pay 15%-17% premium (Table 4-6). Statistically significant coefficients for income
indicate a similar effect on purchase behavior between the levels of $25,000-$50,000 and
$50,001-$75,000, having the latest interval a slightly larger effect. Those who have at
least a college education would also be more willing to pay for certification. Other
categories for age and income levels were not statistically significant, and therefore, they
are not interpreted.
Like the maroon clownfish experiment and contrary to what was expected, U.S.
regions with a collection industry did not show preferences for certification (Table 4-6).
Moreover, all geographic locations (which reflects U.S. regions) have negative and
statistically significant impacts on the decision to pay a 15%-17% premium for a MAC
certified fish. This implies that, in comparison to international hobbyists, respondents
from the U.S. were less willing to pay a price premium for an ecolabelled blue-faced
angelfish from Indonesia, even though the certification may indicate the fish should be
healthier. In addition, estimates for each region show that the negative impact on
certification preference is very similar among location categories.
Results for model 2 identify the factors influencing willingness to pay a 15%-17%
premium for nine additional days of guaranteed survival. Contrary to the hypothesis, the
statistically significant negative sign of the coefficient for certification indicates that
respondents were less likely to pay. This result further supports the hypothesis that an
ecolabel and a survival guarantee are perceived as close substitutes in terms of the
benefits to hobbyists.
Only few individual characteristics were observed to have statistically significant
influences on the probability that an avid hobbyist would pay a premium for a blue-faced
angelfish with an extended survival guarantee. As expected, the level of familiarity with
the MAC certification program has a positive influence, as suggested by the reduction of
the negative influence of certification status on preferences for a survival guarantee.
Unexpectedly, beliefs with respect to the effectiveness of the ecolabel to protect reef and
wild stocks did not significantly modify the effect of certification on preferences in this
scenario. In addition, results show that the influence of certification status on preferences
for an extended life warranty was even more negative for respondents from Alaska,
central and northeastern U.S. regions. Contrary to the hypothesis, respondents from
regions known to be directly involved with the marine ornamental industry did not show
a significant preference for certification. Income, age and education did not influence the
effect of certification status on purchasing decisions in this scenario.
Results for model 3 indicate that characteristics of respondents influenced
preferences for certification more than did an extended a survival guarantee, at the same
price. Again, the level of familiarity with the MAC program and association of the
effective prevention of coral reef damage with the program have a positive impact on
preferences for certification. This estimate reinforces the expected higher and positive
influence of the perceived effective reef protection on preferences for MAC certification.
As expected, income has a positive effect on the preference for certification over a
survival guarantee if price is kept constant. In addition, the coefficients associated with
the income levels of $25,000-$50,000 and $50,001-$75,000 show a positive influence of
those levels on purchase decisions for a certified fish. In this scenario, considering wild
stock protection as a compelling reason to buy a certified wild-caught blue-faced
angelfish was not a significant factor influencing consumer choices. Findings also show
that the expected influences of age and geographic location on preferences for
certification were not observed for these respondents in this scenario.
Model 4 reports the estimates of respondents' characteristics influencing the
willingness to pay a 35% premium for both MAC certification and an extended survival
guarantee for a wild-caught Indonesian blue-faced angelfish. According to expectations,
variables indicating the level of familiarity with the MAC program and the association of
effective coral reef protection with the MAC ecolabel show a positive influence on
purchase decisions. However, the association of wild stock protection with the MAC
ecolabel did not show any significant influence on preferences. Estimates for geographic
location indicate that hobbyists from 3 U.S. regions, except those from the southeast,
were less likely to pay a 35% premium for a certified blue-faced angelfish with an
extended survival guarantee. The estimate for the southeast region was not statistically
significant, and it indicates the absence of an observable influence on preferences with
respect to this scenario. Age, education and income level did not have any significant
effect on preferences for certification.
Estimating Effects on Probability of Purchase
As in the previous experiment, estimates from the four probit models can be used in
simulations to analyze the effect that changes in statistically significant explanatory
variables have on the probability of purchase in each scenario. To do so, independent
variables were set at "base" values against which changes were measured. For this
experiment, the base fixed all variables (i.e., dummies) at the zero level, except for
variables indicating if respondents considered reef and wild stock protection as
compelling reasons to buy a certified fish. Those two variables were set at a value of 1.
The same base was used across models. In summary, the "base" hobbyist did not reside
in the U.S., was not familiar with the MAC ecolabeling program, had only a high school
(secondary) education, earned less than $25,000 annually, and was from 18-24 years old.
Simulations for model 1 represent purchase probability levels for buying a certified
wild-caught blue-faced angelfish from Indonesia at a 15%-17% premium. The
probability of buying such fish at the base condition is 49% (Figure 4-20). In this case,
the probability relates to purchasing a certified fish with only a five-day survival. An
increase of nine additional days of warranty reduces the probability that respondents will
pay the price premium for certification by over 20% (i.e., from 49% to 39%; Figure 4-
20). This observation supports the idea that respondents perceive certification and
survival guarantees as substitutes, especially for a high-value, wild-caught specimen like
the blue-faced angelfish.
Familiarity with the MAC program has a considerable impact on probability of
purchase. Respondents who were at least somewhat familiar with the certification
program increased their likelihood to purchase the certified fish at the specified price
premium by 25 percentage points (from 49% to 74%) or 51% (Figure 4-20).
Respondents who did not considered reef habitat and wild stock protection as
compelling factors to buy a certified (ecolabelled) organism were less likely to pay a
price premium for certification, reducing the probability of purchase by 59% and 39% (to
20% and 30%) respectively, when compared to the base (Figure 4-20). These results also
show that although both variables are important for respondents, reef protection was
perceived as a more influential condition relative to certification than wild stock
protection. This perception probably was supported by the fact that the fish was defined
to be collected from the wild in Indonesia.
Unconvinced Location of residence
80% 74% that MAC
ensures these 64% 67% 68%
40% 39% 30% 33% 34% 34% 33%
Figure 4-20. Probability that the base hobbyist would pay a 15%-17% premium for a
certified wild-caught blue-faced angelfish from Indonesia by various fish and hobbyist
In terms of the effects of socio-demographics, respondents from the U.S. were 30%
less likely to pay the premiums for a certified fish that also had some level of life
warranty (Figure 4-20). As indicated by estimates, this reduction is very similar for all
U.S. regions, indicating no significant differences in purchase behavior across regions.
Moreover, this outcome suggests that the value of MAC certification differs for
American and international hobbyists. One explanation could be the existence of more
competition at the retail level in the U.S. Further research can explore and validate this
or other underlying reasons for such differences.
Results indicate that in terms of age, only respondents older than 44 years showed a
difference in purchase behavior (Figure 4-20). Such respondents increased the
probability of purchase by 30%, ceteris paribus (i.e., from 49% to 64%). In addition,
those who have attended college increased their likelihood to purchase by over 18%
when compared to the base (i.e., from 49% to 58%).
With respect to income effects, only those respondents with annual income levels
of between $25,000 and $75,000 affected probability of purchase (Figure 4-20). Those
individuals with incomes between $25,000 and $50,000 increased their likelihood to
purchase the specified fish by almost 37% (i.e., from 49% to 67%) and those with
incomes ranging from $50,001 to $75,000 increased their likelihood to purchase by 39%
(i.e., from 49% to 68%). This small difference suggests similar preferences across the
two income levels.
Simulations for model 2 show the probability levels associated with the willingness
to pay a 15%-17% premium for an extended life warranty for a fish with varying
certification status (Figure 4-21). At the base, the probability of purchase for such a fish
is 69%. This level corresponds to an uncertified fish. The presence of the MAC ecolabel
reduces the likelihood to buy the specified fish in this scenario by over 10%, from 69% to
62% (Figure 4-21). This observation suggests that respondents perceive MAC
certification, within the context of the experiment and the definition of the angelfish, as
directly related to increasing the chance of survival and quality of the purchased fish.
If respondents were at least somewhat familiar with the MAC program, the
presence of an ecolabel increased the probability of purchase by about 7% (5 percentage
points) in comparison to the base (i.e., from 69% to 74%; Figure 4-21). This suggests
that for these respondents, certification is perceived more as a complement to survival
guarantee than as a substitute. The result also indicates an association between
certification and issues beyond improving fish quality and chance of survival.
The negative impact of U.S. geographic location with respect to certification
observed in the previously discussed model also appeared with model 2, although just for
two regions (Figure 4-21). For this scenario, respondents from the Great Lakes and Big
Rivers region would have a 29% lower likelihood of purchasing a certified blue-faced
angelfish with extended survival guarantee (i.e., from 69% to 49%). Respondents from
the Southwest, Mountain Prairie and Alaska regions showed an even larger reduction in
probability to purchase. They would be about 42% less likely to pay for the extended life
warranty in this situation. This observation suggests that these respondents focused on
the implications of certification for fish quality more strongly than other individuals,
which diminishes the role of the warranty attribute.
Note that given the same price increases for scenarios analyzed in models 1 and 2,
the base probability of buying a fish offering only an extended survival guarantee was
59% higher than the base probability of purchasing a certified fish that offered a limited
life warranty (figure 4-20 vs. figure 4-21). Such a situation suggests a somewhat stronger
preference for life guarantees over certification as related to ensuring the quality of a
high-value fish, whose supply may have harmful implications for the fish stocks and reef
80% 69% 62%74%
Base (uncertified) MAC certified MAC familiar Great Lakes-NE SW-Mountain-
Figure 4-21. Likelihood that the base hobbyist would pay a 15-17% premium for a wild-
caught Indonesian blue-faced angelfish with an extended life warranty by
various hobbyist characteristics
Simulations for model 3 illustrate the probability of buying a certified fish over a
fish with an extended life warranty (i.e., 14 days) at the same price (Figure 4-22). The
base probability is 50%, which is the probability that the hobbyist with the base
characteristics would prefer certification over an additional 9 days warranty.
Again, not considering coral reef protection as a compelling factor to buy a
certified organism appears to reduce the probability of purchase by about 60% (i.e., from
50% to 20%). Familiarity with the MAC program has a positive effect on likelihood to
purchase (Figure 4-22). If consumers have at least some level of familiarity with the
MAC program, their likelihood to prefer certification to an extended life warranty
increases by 26% when compared to the base (i.e., from 50% to 63%).
Income level also shows positive impacts (Figure 4-22). Respondents with income
levels between $25,000 and $75,000 increase their probability of purchasing the certified
fish by about 40%-42%, keeping other variables at the base. Again, the two income
categories show similar effects on preferences.
Note that in comparison to model 1 where there was a 15% premium, respondents
were only 2% more likely to buy a certified fish at the same base levels although there
was not any associated price increase. This observation suggests that price is not an
important issue influencing purchase decisions. Moreover, this observation suggests that
MAC certification, by itself, is not as environmentally compelling as expected for this
segment of respondents, especially when a substitute attribute is offered.
80% 70% 71%
Base MAC farniliar Unconvinced Income $25,000- Income $50,001-
MAC ensures $50,000 $75,000
Figure 4-22. Probability that the base hobbyist would prefer a wild-caught certified
Indonesian blue-faced angelfish to an extended survival guarantee at a
The simulated effects on the probability of purchase for a certified and 14-day
guaranteed fish at a 35% price premium had a base level of 75% (Figure 4-23). For this
scenario, the only factor that would increase the likelihood to purchase a fish as specified
is the familiarity with the MAC program. Again, hobbyists that were at least somewhat
familiar with the program increased their likelihood to purchase by over 17% (i.e., from
75% to 88%). As observed in previous scenarios, respondents who did not consider
preventing coral reef damage as justifying buying a certified fish decrease their likelihood
of purchase. They had a 41% lower probability (i.e., reduction from 75% to 44%) of
paying a premium for a certified fish when it also has an extended life warranty (Figure
The negative effect of U.S. geographic location was also observed in this scenario
(Figure 4-23). U.S. hobbyists have at least a 24% lower probability of purchasing a
certified fish as specified, in comparison to the base. The effect of the Southeast region
was not statistically different from zero and it was not considered in this analysis.
At the same base levels, the base probability for willingness to pay a higher
premium for a fish featuring certification and extended survival guarantee was higher
than the base probabilities showed in the previous scenarios (i.e., 75%). This observation
indicates two main things. First, that the specific segment of participants in this survey
do not weight price considerations as heavily as other fish attributes, especially those
related to environmental issues. Second, considering the preferences for extended
survival guarantees observed in model 2, it is concluded that the increased likelihood to
purchase in model 3, was also related to an extended life warranty for the fish.
57% 56% 57%
Base MAC familiar Unconvinced Pacific Great Lakes SW-
MAC and NE Mountain-
ensures reef Alaska
Figure 4-23. Probability that the base hobbyist would pay a 35% premium for a certified
and extended warranted blue-faced angelfish wild-caught in Indonesia
Considering the effects that significant variables had on probabilities of purchase
for each of the scenarios analyzed, the main implications are summarized below:
* In general, results suggest that "arrive alive-stay alive" life warranty and
certification (ecolabels) are perceived as substitutes by avid hobbyists surveyed in
* This unique group of respondents seemed to have stronger preferences for survival
guarantees than for environmental certification.
* Considering prevention of coral reef and wild stock damage as compelling reasons
to buy a MAC ecolabelled organism influences positively preferences for
* Increasing the familiarity with the MAC program also has high positive
contributions to the likelihood of purchase a MAC ecolabelled organism.
* Members of this particular segment of the consumer market who reside in the U.S.,
especially in the Pacific, Central and Northeastern regions, show a reduced
likelihood to purchase certified specimens. Further research investigating the
existence of factors influencing the opinions of hobbyists from the U.S. market
would be useful to validate these results.
Respondent Comments to Survey
Several respondents provided comments with respect to their perceptions about
MAC certification program and wild-caught vs. tan-bred sources (Appendix C). They
mainly stated their preferences for tank-bred fish over any similar wild-caught specimen
(including certified ones) in order to promote safer reefs and to more directly ensure more
tank-adaptable organisms, particularly in regards to specimens collected from zones were
harmful harvesting methods are known to exist (e.g., the Philippines). Comments also
included reasons for the lack of preference for ecolabelled fish. In summary many
questioned the credibility of the MAC ecolabeling program.
One of the main reasons underlying the reduced credibility of MAC certification, as
perceived by the survey respondents, was the absence of an effective cyanide-testing
program. Some respondents do not believe that the MAC program is effectively
deterring the use of this harmful collection method in certified collection areas.
Therefore, they believe no real conservation of reef ecosystems and wild stocks is being
achieved. In addition, some respondents indicated their lack of trust on the program with
respect to the effective monitoring of practices throughout the chain of
commercialization. They are concerned about certified operators selling marine
aquarium organisms as certified, although they do not come from certified collection
areas. They are also concerned about the lack of tests to ensure standards for improving
fish health and chance of survival are being met. Lastly, some respondents indicated
doubts about the real objectives of the Marine Aquarium Council regarding money use
and about the applicability of the program for small-scale operators.
Some respondents indicated a potential support for the MAC program for wild-
caught specimens. They stated that their aversion to certified organisms could be reduced
only if they are absolutely sure about the effectiveness of the program to ensure proper
collection and handling of organisms and the sustainability of reef ecosystems. In this
context, the effective implementation of a cyanide-testing program is a crucial issue to
increase credibility. Finally, some comments indicated the necessity of informing
consumers about the objectives and actual environmental achievements of the program.
In the context of the reported concerns, taking advantage of the information
generated by the system of custody, beyond the label, could be useful for MAC.
Providing buyers specific information on collectors and regions (e.g., from paper trail)
where certified specimens come from could help reduce the concern of fraud at the retail
level (i.e., placing the MAC label in a tank which contains specimens from areas actually
not certified by the program) and also could increase the perception and utility obtained
from certified species, given the luxury nature of the hobby and characteristics of the
As it was preliminary hypothesized, members of this segment can be characterized
as well informed hobbyists, with avid participation in the aquarium hobby and high
interest in preservation of marine ecosystems. Interestingly, only the minority of
respondents showed a high level of familiarity with the MAC certification program for
wild-caught fish, and the majority actually prefer tank- bred specimens as a mean to
avoid harming the ecosystem.
For both experiments, it was observed that, according to expectations, price is not
the most important attribute considered when purchasing marine ornamental fish. In
addition, marine ecosystem implications have a high influence on observed preferences.
On the other hand and contrary to hypotheses, MAC certification did not positively
influence consumers' purchase decisions. Moreover, a MAC certified fish is likely to be
less preferred in comparison to cultured and/or guaranteed fish. With respect to observed
attitudes towards MAC certification, the following observations can be made:
* Although respondents show high levels of interest in environmental issues, most of
them do not consider the MAC environmental certification program as an effective
means for promoting sustainability of reef ecosystems and fish quality.
* Although respondents have shown that price is not an issue when deciding to buy a
fish given their past purchase behavior (i.e., highest prices paid) their willingness
to pay for an ecolabelled fish limited.
* Positive effects of familiarity with the MAC ecolabeling program on preferences
for certified fish depended on hobbyists that were only somewhat familiar with the
* Higher preferences for the MAC ecolabel among this segment of respondents could
be achieved by focusing the program on providing support that it prevents coral
reef damage and promoted sustainability of wild stocks.
In comparison to a previous study of preferences for marine aquarium ornamentals
at different levels of the marketing channel (Rubinstein 2003) fish attributes such as
tank-bred and survival guarantees are also likely to be preferred to certification by
hobbyists. Likewise, an increased familiarity with the program showed a positive
significant impact on preferences for MAC certified fish. However, most of the
participants in the present study did not have a strong familiarity with the MAC program
and the reduction of the negative effect of MAC certification was due to a limited level of
Contrary to expectations, respondents from regions affected by the aquarium
industry are not likely to have a positive influence on preferences for the MAC ecolabel.
Finally, only for the case of a higher value specimen restricted to collection from the wild
(i.e., blue-faced angelfish), some demographic characteristics like education, income, and
age supported the hypothesis of positive influences on MAC preferences.
SUMMARY AND CONCLUSIONS
The Marine Aquarium Council (MAC) ecolabeling program is a means to promote
the sustainability of saltwater ornamental fish and coral reef ecosystems through market
mechanisms. Information about potential influences on consumers' purchase decisions,
including preferences for the MAC ecolabel, is needed to assess the potential
effectiveness and success of the program.
The main purpose of this research is to provide information about the influence of
selected fish attributes, including ecolabels, as well as the impact of individuals'
characteristics on preferences for marine aquarium fish at the final consumer level. To
achieve the purpose, an Internet-based survey was implemented following a discrete
choice experiment technique. Two discrete choice experiments were conducted to
analyze the importance of attributes in purchase decisions. One experiment presented a
high-value specimen, the blue-faced angelfish, and the other introduced a very affordable
and popular fish, the maroon clownfish. The former was used to analyze the effects of
price, MAC certification ecolabel status and the length of a post-purchase survival
guarantee on hobbyist preferences. The latter investigated the effects of price,
certification status and fish source (wild-caught or tank-bred). The angelfish and
clownfish were assumed from Indonesia and the Philippines, respectively.
The sample of respondents addressed in this research represented a specific group
of individual hobbyists, members of on-line discussion boards who were recruited
following a convenience sampling approach. Considering that this sample is unlikely to
represent the entire population, the validity of the reported results is limited to this
Descriptive analysis showed that this segment of respondents represents a very
homogeneous group in general. Most of them are males between 24-44 years old, with
above average levels of education and annual income. They are aware of, and give
considerable importance to, protection of coral reefs and wild stocks.
As hypothesized, they show a particularly high level of involvement in and
knowledge about their hobby. Around 80% consider keeping marine aquaria as their
primary hobby, 59% are members of an aquarium society, 88% research the specimens
they keep, and over 60% have been paid more than $50 for a single fish. Contrary to
expectations, about 50% are not at all familiar with the MAC certification ecolabeling
Observations from the choice experiments were analyzed through binary probit
models using maximum likelihood procedures. The estimated models were used in
simulations to illustrate probabilities of change in consumer behavior given changes in
the values of statistically significant explanatory variables.
Results from both exercises showed interesting and unexpected findings. As
hypothesized, price was observed to be a secondary factor influencing purchase behavior,
however price was positively related to increases in the likelihood of purchase. Contrary
to expectations, the role of MAC certification was perceived as a close substitute for
other fish attributes (i.e., extended life warranties and tank-bred specimens), with less
strong or even negative effects on purchase behavior. An important observation is that
respondents' comments revealed a strong lack of credibility for the MAC program and a
higher confidence in alternatives such as tank culture as a means to avoid harmful
consequences related to collection from the wild.
For the maroon clownfish experiment, an extreme preference for tank-bred fish was
observed. When compared to a wild-caught fish, with price kept constant, tank culture as
a source increased the probability of purchase remarkably. This considerable increase
was observed for both levels of certification status, although the effect was larger for an
uncertified specimen. Simulations also showed that regardless of price variations,
respondents seem increasingly willing to buy tank-bred fish, although at a diminishing
rate with incremental increases in price.
On the other hand, the effect of MAC certification at a constant price was negative.
When compared to an uncertified specimen, the likelihood that an avid hobbyists would
purchase a certified fish decreased for both wild-caught and tank-bred fish, although the
decrease was larger for a tank-bred fish. However, simulations with price increases
showed that respondents' willingness to pay for certification increased at an increasing
rate if the fish was wild-caught and at a decreasing rate if the fish was cultured. Such
observations suggested that avid hobbyists would be increasingly likely to pay price
premiums associated with MAC certification if a fish is caught in the wild. Respondents
seemed to be concerned with ecosystem protection. However, they would not pay price
premiums for certification if the fish was tank-bred. This result suggests a low market
potential for an extension of the MAC program to cultured specimens, at least among this
segment of hobbyists. Furthermore, since several clownfish species are available from
culture, this observation could be applicable to other clownfish species.
In the case of the blue-faced angelfish, respondents' preferences were analyzed in
four different scenarios. Results revealed that an extended survival guarantee (by 9 days)
and ecolabel were perceived as close substitutes in terms of ensuring better quality fish
collected from the wild. However, the positive influence of extended life warranties on
purchase decisions was higher than the effect of the MAC ecolabel. In addition, it was
again observed that this specific segment of hobbyists did not weigh price considerations
as heavily as other attributes, especially those related to environmental issues.
The effects of respondents' characteristics on preferences for MAC certification
were also analyzed in both experiments. Increasing the level of familiarity with the MAC
program and the association of effective prevention of coral reef and wild stock damage
with the MAC ecolabel showed highly significant positive influences on preferences for
certification by avid hobbyists. This confirms the initial hypothesis that marine
ecosystem implications have a high influence on preferences for saltwater ornamental
fish for this group of hobbyists.
Demographic variables like age, income level, education and geographic
distribution also showed significant influences on preferences for certification, but only
in some of the blue-faced angelfish scenarios. Results showed that those respondents
older than 44 years, who have at least a college education and an annual income between
$25,000 and $75,000 were more likely to choose a certified fish at the specified price
On the other hand, and contrary to initial expectations, regions exposed to the
aquarium industry did not show a positive reaction to certification. Moreover, in
comparison to international hobbyists, results indicate that respondents from the U.S.
were less likely to purchase a certified fish. Such a result suggests a stronger perception
of survival guarantees as a substitute for certification (ecolabeling) when considering fish
quality. Further research to confirm this finding is needed.
Considering the observed negative perception for the MAC ecolabel, the market
potential of the program for this group of hobbyists looks limited. However, since only
50% of respondents have some level of familiarity with MAC, efforts to improve the
level of knowledge and perceived credibility of the program are recommended. Focusing
efforts on an effective diffusion of the program's scope of action and promulgating plans
the Marine Aquarium Council has to address not only coral reef protection but also
sustainability of fish stocks and efficient post-harvest activities (i.e., handling, holding
and transportation). Such information could be very useful and successful in improving
preferences for a MAC ecolabel.
It is important to note that although results from this research do not apply to the
entire population of marine aquaria owners, an understanding of the preferences of this
specialized group could be very useful for creating market diffusion strategies. Due to
their level of involvement in the hobby and high exposure to information, these
respondents should be easier, faster and cheaper to reach with educational and
promotional efforts. In addition, due to the secondary role that price plays in influencing
purchase behavior and the increased capacity to afford price premiums, this group would
be very likely to react positively to price increases and to contribute to support of the
program, if their preferences for the MAC ecolabel are improved. Lastly, a survey of the
general population and those covering additional species, would be useful in obtaining a
better estimate of overall demand. A study of expected costs associated with becoming
and remaining certified could also help estimate expected premiums associated with the
SURVEY INSTRUMENT AND CODING
Section 1 of 3: Background and Opinions
Q What is the total size of tanks that you currently maintain?
0 None (I do not have an active tank)
1 1 to 9 gallons (1 to36 liters)
2 10 to 24 gallons (37 to 93 liters)
3 25 to 49 gallons (94 to 188 liters)
4 50 to 74 gallons (189 to 283 liters)
5 75 to 99 gallons (284 to 377 liters)
6 100 to 180 gallons (378 to 683 liters)
7 More than 180 gallons (More than 683 liters)
1 1 to 9 gallons (1 to36 liters)
2 10 to 24 gallons (37 to 93 liters)
3 25 to 49 gallons (94 to 188 liters)
4 50 to 74 gallons (189 to 283 liters)
5 75 to 99 gallons (284 to 377 liters)
6 100 to 180 gallons (378 to 683 liters)
7 More than 180 gallons (More than 683 liters)
Q2 Please indicate whether you plan to change your saltwater tank capacity within the
next two years:
0 No change
Q3 Regarding your saltwater aquarium hobby, how long have you maintained a tank?
0 I do not have a tank
1 Less than a year
2 1 to 10 years
3 11 to 20 years
4 More than 20 years
Q4 Please identify the range that contains the highest price you have ever paid (excluding
tax and shipping, if applicable) for a single aquarium finfish and invertebrate (shrimp,
giant clam, etc.):
Q4.A Saltwater finfish
0 U.S.$0 (I have not purchased before)
1 Less than $10 U.S.
2 $10 to less than $25 U.S.
3 $25 to less than $50 U.S.
4 $50 to less than $75 U.S.
5 $75 to less than $100 U.S.
6 $100 U.S. or more
Q4.B Saltwater invertebrate
0 U.S.$0 (I have not purchased before)
1 Less than $10 U.S.
2 $10 to less than $25 U.S.
3 $25 to less than $50 U.S.
4 $50 to less than $75 U.S.
5 $75 to less than $100 U.S.
6 $100 U.S. or more
Q5 Where do you spend the most money purchasing live saltwater specimens?
1 Local general pet store
2 Local aquarium store
5 Not applicable (I do not have a saltwater tank)
Q6 The following yes/no questions are about your interests.
Yes (1) No (0)
Q6.A Have you ever visited a coral reef?
Q6.B Is keeping an aquarium your primary hobby? ___
Q6.C Have any species reproduced in your tank? C [
Q6.D Have you ever done research on the specimens you keep
in your tank?
Q6.E Have you ever bought a fish or invertebrate that was bred
in a tank instead of caught in the wild?
Q6.F Are you a member of any environmental organization? E
Q6.G Are you a member of any aquarium society? _
Q6.H Do you have a "fish only" tank?
Q6.I Do you have a "reef' tank? C C
Q7 How familiar are you, if at all, with the Marine Aquarium Council (MAC)
0 Not at all familiar
1 Somewhat familiar
2 Very familiar
Section 2 of 3: Choice Exercises
You will be presented with a series of choices between two fish. The fish will differ
either by one or more of the following characteristics: