SEGMENTING ANGLERS BY
LIFESTYLES, LAKE TYPES, AND MANAGEMENT PREFERENCES
By
Jody Christopher Simoes

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Fisheries and Wildlife - Doctor of Philosophy
2014

ABSTRACT
SEGMENTING ANGLERS BY LIFESTYLES,
LAKE TYPES, AND MANAGEMENT PREFERENCES
By
Jody Christopher Simoes
Information on angler behaviors and preferences is vital for management efforts that
enhance fishery resources and provide public benefits. This dissertation adds to the angler
human dimensions literature on angler segments and preferences by using two surveys of
Michigan’s recreational anglers in support of three dissertation chapters to: (1) test the
effectiveness of a widely-used marketing segmentation methodology for profiling angler
behavior, (2) enhance existing lake classifications with data on user (i.e. angler) characteristics
and fishing behaviors and, (3) model anglers’ preferences over management outcomes in the
Great Lakes.
In the first chapter, although the commercially-available classification approach
examined is easily implemented, the evaluation suggests that while adept at producing market
segments that are accessible and stable over time, statistical relationships between our segment
variable and the fishing behaviors and purchase behaviors were generally weak. The results
highlight limitations of using this generic approach to develop market segments that exhibit
similar within-group fishing and license purchase behaviors.
Chapter two examines the relationship between a recent inland lake landscape-based
classification effort and angler-reported trip characteristics from an ongoing, statewide angler
survey. The research goal was to develop managerially-relevant inland lake angler descriptions.
Results from a multinomial probit model largely supported the descriptive analysis. Marginal
effects for several of the angler trip characteristics and species targets were large and statistically

significant. While not originally intended as an angler behavior classification strategy, the result
demonstrate that the lake characteristics and species assemblages used to classify Michigan’s
inland lake resources may also serve as an integrated (social and biological) classification system
for inland lake fisheries resources.
In support of Great Lakes management plans, chapter three reports on a stated preference
survey of Michigan anglers’ preferred outcomes from management activities in four Great Lakes
and Lake St. Clair. Discrete choice trade-off questions were analyzed using conditional logit
models. Results showed anglers generally expressed stronger preferences for management
outcomes related to ecosystem health attributes and recreational opportunity attributes. To
further demonstrate the managerial applications of the trade-offs quantified by the research,
predicted preferences were calculated to illustrate the importance of species abundance relative
to increasing average fish size. The model was also used to calculate choice probabilities for
three hypothetical Great Lakes management strategies which differed in their emphasis on
Pacific salmon, prey base, risks of ecosystem collapse and average fish size. Specifically,
predicted probabilities were calculated for management outcomes that emphasized; (A) an
fishery with more Pacific salmon and less native sportfish, (B) less Pacific salmon and more
native sportfish, and (C) the status quo or same as today management option. In general, choice
probabilities for the average angler, which can be interpreted as the average amount of support
for management outcomes, were greater for management outcomes favoring a native species
emphasis.

ACKNOWLEDGMENTS

I would like to thank the individuals and organizations that have contributed to my
research. First, I would especially like to thank my committee members, Dr. Roger Calantone,
Dr. Daniel Hayes, Dr. Tammy Newcomb and, in particular, my research advisor Dr. Frank Lupi.
I am grateful for the support of Michigan State University’s Department of Fisheries and
Wildlife, College of Agriculture and Natural Resources, Department of AgBioResearch,
Graduate School, as well as the Michigan Department of Natural Resources Fisheries Division,
and the Great Lakes Fisheries Trust. This project was also supported by the Sport Fish
Restoration Program under the Sport Fish Restoration Act. Several individuals within the
Michigan Department of Natural Resources provided moral support and encouragement- if not
directly, then by their sincere interest in the results of these project. In particular, I would like to
acknowledge: Jim Breck, Jim Dexter, Tracy Kolb, Roger Parsons, Tami Pattison, Kristin
Phillips, Theresa Pohl, Christine Schwerin, Kevin Wehrly and Gary Whelan. Special thanks go
to Tracy Kolb for assistance with all things Microsoft Access.
I am also grateful to my data entry crew: Geralyn Gerwitz, Emily Pankey, Patricia
Thompson, and Shannon Walker. Drew Rayner deserves special recognition for his tireless
work ethic. Cheers to the “Lupi Lab”, past, present and by association: Min Chen, Jessica Klatt,
Scott Knoche, Max Melstrom, Scott Weicksel, and Kwame Yeboah. In general, there was too
much shop talk and not enough beer. Regrets.
Thanks also to my friends, who provided me with ample opportunities to witness angling
behavior first-hand and for continually reminding me that there is indeed life outside of this
cubicle.
iv

I would also like to thank my family. From them I have learned about sacrifice,
boldness, selflessness, patience and love. Without their vision and continued support, nothing
I’ve done would have been possible. Finally, I would like to thank my wife Katie, for believing
in me and inspiring me to be more. We’ve somehow managed to start our own family during
this time. Katie, this is for you and our boys.

v

TABLE OF CONTENTS

LIST OF TABLES

viii

LIST OF FIGURES

x

INTRODUCTION

1

CHAPTER 1
Evaluating the effectiveness of a commercial geodemographic market segmentation
methodology for classifying recreational angler behavior
Motivation
Background
A Marketing Approach to Natural Resource Management
Recreation Participation Research
Big Data and a Commercial Geodemographic Market Segmentation Approach
Methods
Data Collection
Analysis
Results
Michigan Angler Segment Classifications
Esri TapestryTM LifeMode and Behavioral Variable Associations
Discussion

5

CHAPTER 2
Describing anglers using an inland lake classifications system
Motivation
Classification of inland lake anglers
Methods
Data Collection
Model
Marginal effects
Results and Discussion
Descriptive Statistics
Multinomial Probit Model
Demographics
Trip Characteristics
Species Targeted
Summary and Conclusions
Lake Class 1
Lake Class 2
Lake Class 3
Lake Class 4
Lake Class 5

33
33
33
34
36
36
37
39
40
40
45
48
49
50
52
53
53
54
54
55

vi

5
5
7
7
9
10
14
14
16
18
18
22
29

Lake Class 6

55

CHAPTER 3
Management preferences of Michigan’s Great Lakes Anglers
Motivation
Choice Experiment
Random Utility Theory
Marginal Rate Substitution (MRS)
Data
Survey Development: Pretests
Survey Sample, Survey Strategy, Response Rate
Experimental Design
Results and Discussion
Management preferences of Michigan’s Anglers
Lake Erie
Lake Huron
Lake Michigan
Lake St. Clair
Lake Superior
Comparing Preferences of Great Lakes and Non Great Lakes Anglers
Conclusions and Management Implications

57
57
57
58
62
66
68
68
68
69
70
70
75
76
77
78
78
79
80

APPENDICES
APPENDIX A
APPENDIX B
APPENDIX C
APPENDIX D
APPENDIX E
APPENDIX F
APPENDIX G
APPENDIX H
APPENDIX I
APPENDIX J

82
83
93
94
101
105
106
108
110
114
122

REFERENCES

123

vii

LIST OF TABLES

Table 1 Esri Tapestry LifeMode Classifications

21

Table 2 Comparison of Segmentation Approaches

24

Table 3 Comparison of Segmentation Approaches

25

Table 4 Comparison of LifeMode fishing behavior characteristics

27

Table 5 Comparison of LifeMode license purchasing characteristics

28

Table 6 Predictor variables for the multinomial probit model

39

Table 7 Lake characteristics, classification and angler characteristics

45

Table 8 Marginal effects, standard errors and significance levels from probit estimates

46

Table 9 Attributes and attribute levels for states preference choice model

65

Table 10 Parameter estimates for conditional logit model

73

Table 11 Predicted probabilities

75

Table B 1 Michigan Recreational Angler Survey Response Rate

93

Table C 1 MDNR sample, respondents and nonrespondent characteristics

95

Table C 2 Post-stratification weights: Age and License Distributions (male residents)

96

Table C 3 Post-stratification weights: Age and License Distributions (female residents)

97

Table C 4 Post-stratification weights: Age and License Distributions (male nonresidents) 98
Table C 5 Post-stratification weights: Age and License Distributions
(female nonresidents)

99

Table D 1 Comparison of LifeMode general fishing behavior characteristics

100

Table D 2 Comparison of LifeMode Michigan-specific fishing behavior characteristics

101

Table D 3 Comparison of LifeMode and species fished for in last 12 months

102

Table D 4 Comparison of LifeMode and bait types used in the last 12 months

103

viii

Table D 5 Comparison of LifeMode and license purchasing behavior

104

Table E 1 Inland Lake Classification

105

Table F 1 Great Lakes Angler survey response

107

Table G 1 MDNR sample, respondents and nonrespondent characteristics

109

Table G 2 Post-stratification weights: Age and License type distributions

109

Table H 1 Predicted Probabilities: calculations for Lake Huron

111

Table H 2 Predicted Probabilities: calculations for Lake Michigan

112

Table H 3 Predicted Probabilities: calculations for Superior

113

Table I 1 Comparison of Great Lakes and non-Great Lakes anglers

117

Table I 2 Comparison of Great Lakes and non-Great Lakes anglers

118

Table I 3 Log likelihood ratio tests: Great Lakes and non-Great Lakes anglers

119

Table I 4 Lake Huron model comparison

120

Table I 5 Lake Michigan model comparison

121

Table J 1 Marginal Rate Substitution calculations

122

ix

LIST OF FIGURES

Figure 1: Image of example Great Lakes Angler Survey Choice Experiment

61

Figure A.1: Michigan Recreational Angler Survey

87

Figure A.2: Michigan Recreational Angler Survey Contact Letters

91

x

INTRODUCTION

Fisheries management represents a complex socio-ecological system (Martin & Pope
2010) and licensed anglers represent one of the most important constituent groups for state
fishery managers (Connelly, Keeler, & Knuth. 2013). Therefore, information on angler
behaviors and preferences is vital for management efforts that enhance fishery resources and
provide public benefits.
Michigan’s vast freshwater resources include four Great Lakes, Lake St. Clair and
thousands of inland lakes. Faced with the challenge of managing heterogeneous systems across
large landscapes, many state agencies utilize resource classification systems to describe and
inventory resources and plan for and prioritize their management activities. In addition nearly all
state fisheries management agencies have implemented angler data collection strategies to
inform fisheries management plans and to better understand and communicate with anglers (e.g.
creel surveys, license sales analyses, angler surveys, marketing and outreach activities) (Simoes
2009).
The objectives of this dissertation are to (1) evaluate a segmentation method for
developing actionable angler market segments for marketing and outreach efforts, (2) develop a
managerially-relevant inland lake angler classification system and, (3) measure and forecast
angler’s preferences for Great Lakes management. These objectives are addressed in turn in
each of the three chapters of this dissertation.
Two surveys of Michigan’s recreational anglers support these efforts. The Michigan
Recreational Angler Survey (MRAS) was designed as part of a long-term, statewide survey
1

strategy to describe the status and distribution of angling effort across all of Michigan’s fisheries
(Simoes 2009). The MRAS database provides a rich source of self-reported demographic
information as well as fishing behaviors. As an outgrowth of that effort, the multi-mode Great
Lakes Angler Survey was developed to inform Great Lakes management plans. In addition to
demographics and general fishing behaviors, the Great Lakes Angler Survey included questions
related to program awareness, management preferences, management opinions and a stated
preference choice experiment section.
Researchers have used a broad range of individual-level variables and broader
socialization or social-structural variables to describe recreational subgroups and their behaviors.
Commercially-available segmentation systems linked to zip codes (e.g. Esri Tapestry™
Segmentation) are widely available and offer a convenient method for developing informationrich and accessible consumer profile descriptions. The approach has also been advocated for
informing angler recruitment and retention strategies because some segments exhibit different
propensities to purchase licenses Southwick Associates (2006). Chapter one builds upon this
earlier research by using self-reported fishing behavior data from the MRAS and license
purchase behavior from the Michigan Retail Sales System license database to further evaluate
the effectiveness of this segmentation strategy for developing behaviorally-distinct angler market
segments. Weak associations between the behavior variables and the market segments we tested
suggest that the exclusive use of LifeMode classifications to reach anglers with specific
behaviors is unlikely to be effective. However future examination may be warranted in light of
the availability of less aggregated Esri Tapestry™ segments and the anticipated improvements in
these segments with continued growth and diversity of consumer information resources.

2

In lake-rich states such as Michigan, the multitude of inland lakes, their diverse
morphological, chemical and biological characteristics and differing fisheries management
strategies result in wide array of angling experiences available to anglers. To characterize
Michigan’s inland lake fishery resources and inform management decisions, Wehrly et al. (2012)
identified six distinct inland lake classes and fish assemblages utilizing biological data and
spatially extensive variables. In order to develop managerially-relevant inland angler profiles,
Chapter two employed a multinomial probit model to examine the relationship between the
outcome variable (lake class) with information from the MRAS database including
demographics and detailed information on fishing trips (the nature of the trip, methods used and
species targeted) as our predictor variables. The results demonstrate that some trip
characteristics and trips targeting particular species significantly increased the likelihood of
fishing in particular inland lake classes. While the influence of lake characteristics on angler
behavior may seem intuitive (i.e. the effect of lake size, lake remoteness, and species profiles)
our results show that the relationship is not completely reliable. For example, some inland lake
classes attracted anglers targeting species that were not relatively abundant, and therefore not
prominent in the original classification strategy. While not originally intended as an angler
behavior classification strategy, the result demonstrate that the lake characteristics and species
assemblages used to classify Michigan’s inland lake resources may also serve in an integrated
(social and biological) classification system for inland lake fisheries resources.
Great Lakes management strategies are characterized by conflicting and often polarizing
management approaches related to Pacific salmonines and native species rehabilitation.
Additional system-wide management challenges stem from invasive species impacts and broader
environmental changes occurring in the Great Lakes. Policy decisions and management

3

strategies are further complicated by a lack of Great Lakes angler preference information. The
Great Lakes Angler Survey for Chapter three included a stated-preference choice experiment to
examine Michigan anglers’ preferred outcomes from management activities in four Great Lakes
and Lake St. Clair. Broadly, respondents were more supportive of management outcomes
favoring walleye, yellow perch and salmon species, and were indifferent in terms of lake trout
abundance. However, preferred lake management outcomes differed by lake. To further
illustrate managerial implications of the results, we used the model to predict support for Great
Lakes management strategies differentiated by their emphasis on Pacific salmon, prey base, risks
of ecosystem collapse and average fish size. In general, choice probabilities, interpreted as
predicted support, were greater for outcomes with a native species emphasis. In contrast to
single-item opinion measurement questions, the stated-preference choice model provides Great
Lakes managers with an understanding of the relationship and relative importance of multiple
management attributes as they contribute to angler preferences for management outcomes, and
the tradeoffs anglers might make when considering both ecosystem health attributes and
recreational opportunity attributes jointly.

4

CHAPTER 1

Evaluating the effectiveness of a commercial geodemographic market segmentation
methodology for classifying recreational angler behavior

Motivation
Despite continued population growth, U.S. fishing license sales peaked in 1988 and have
continued on a gradual downward trend, with national sales in 2013 down 11% (USFWS 2014).
For most states, license fees combined with federal excise taxes represent the vast majority of
state expenditures for fisheries management. Therefore, general declines in license sales and
angling participation have a direct impact on the ability of fisheries agencies to manage fisheries,
protect and enhance aquatic resources and meet angler needs. Beyond the possible financial
implications, a declining recreational angler constituency may erode public and political support
for fisheries management and aquatic resource conservation (e.g. Organ et al. 2012; Sutton, Dew
& Higgs 2009). Angling participation is also of interest to the fishing industries and businesses
that support angling activities (Murdock, Backman, Ditton, Hoque & Ellis 1992; Sutton, Dew &
Higgs 2009; Aas & Arlinghaus 2009; Kuehn, Luzadi & Brincka 2013). Declines in U.S. license
sales and fishing participation have been a catalyst for state, federal, academic and private sector
participation research and recruitment and retention initiatives. Beginning the mid 90’s, surveys
of agency personnel revealed that many state fisheries agencies had implemented public
awareness and education strategies (Ross & Loomis 1999), angler human dimensions surveys
(Wilde, Ditton, Grimes, & Riechers 1994; Fisher & Burroughs 2003), and recruitment and
retention efforts (Mather, Parrish, Stein, & Muth 1995). To specifically address declines in
angling and hunting participation, natural resource agencies have instated advisory boards

5

(Illinois Department of Natural Resources 2009) developed recruitment and retention action
plans (Oregon Department of Fish and Wildlife 2012), developed agency-stakeholder
recruitment and retention councils (Minnesota Department of Natural Resources 2013) and
incorporated management metrics for ongoing evaluation of participation rates (Michigan
Department of Natural Resources 2013). Researchers have examined the relationship between
angling participation and demographic characteristics, motivations and cognitions (e.g. Fedler &
Ditton 2001; Bissel, Duda, & Young 1998; Floyd & Lee 2002; Heberlin, Serup, & Ericsson
2008; Schroeder, Nemeth, Sigurdson, & Walsh 2008; Arlinghuas 2006; Hunt & Ditton 2002;
Arlinghaus, Tilner & Bork 2014; Schroeder, Fulton, Nemeth, Sigurdson, & Walsh, 2008)).
Consulting firms have developed marketing plans (e.g. Responsive Management 2009) and
conduct ongoing participation and market trend analyses (e.g. Southwick Associates 2012) to aid
agency decision making. Finally, revenues generated as part of the Sport Fish Restoration Act
provide grant support for outreach, education and participation research and fund the
Recreational Boating & Fishing Foundation “Take Me Fishing ™” national campaign aimed at
increasing boating and fishing participation through advertising, direct mail marketing,
educational programs and commissioned research. A series of recent projects, funded by Sport
Fish Restoration multi-state grants, have classified anglers using a commercially available
geodemographic classification system (i.e. Esri Tapestry™ Segmentation) with the goal of
identifying market segments to inform recruitment and retention campaigns (see American
Sportfishing Association 2007; Southwick Associates 2007; Southwick Associates 2006).
Geodemographic segmentation is a neighbored classification approach which combines
information from public and private resources to characterize and group households. The
method is widely applied by businesses, and the above projects have shown that commercially

6

available geodemographic segments can be used to identify addresses in neighborhoods that are
more likely to purchase licenses. This paper further explores theses segmentation approaches
with a specific focus on their relationship to angler behaviors.
Commercially available market segmentation systems are widely applied in the private
sector for informing business decisions. Data compilers, marketing firms, and the businesses
utilizing this information benefit from increased accessibility to a range of information on
consumers and consumer behavior from both public and private sources as well as improvements
in technology and computational power. Our primary objective was evaluate the utility of the
Esri Tapestry™ Segmentation approach as an angler market segmentation methodology,
specifically in developing distinct angler subgroups in terms of their fishing and license purchase
behaviors. While there is a rich body of work conducted by human dimensions researchers to
support the concept of angler segments exhibiting significant differences in motivations,
preferences, behaviors and participation, the human dimensions literature lacks an evaluation of
these commercially available, generic segmentation approaches. Knowledge of current
limitations and future research needs is critical as managers, researchers, stakeholders and
businesses consider these programs for profiling angler subgroups, predicting participation and
developing targeted marketing and communication efforts for education, promotion and related
recruitment and retention purposes.

Background
A Marketing Approach to Natural Resource Management
Operating within diverse and often competitive consumer markets, market segmentation
techniques have been widely applied in the private sector to drive all aspects of marketing tactics
7

(e.g. distribution, product development, site location, pricing, media choice) and to define overall
organizational strategic planning and mission development (Tonks 2009). The opposite or “mass
marketing” approach for products and services finds fewer applications in businesses and in
market research (e.g. Vogt 2011, Bruwer & Li 2007).
In the public sector, movement toward a marketing-oriented approach has mirrored
development within the private sector, as agencies revise standardized service approaches in
favor of targeted strategies aimed at satisfying the identified needs of particular client groups
(Crompton & Lamb1986). Angler human dimensions research, so often motivated by agency
management issues related to user group conflicts, equity, regulation formation, and participation
declines is grounded in the premise that the “average angler” does not exist, originally inspired
by Shafer’s (1969) research. More than three decades of human dimensions research has
produced a rich body of literature to support the concept of angler populations as a collection of
heterogeneous subgroups. These subgroups often exhibit significant differences in preferences,
motivations and behaviors (e.g., Bryan 1977; Chipman & Helfrich 1988; Ditton 1996; Fisher
1997; Romberg 1999; Connelly, Knuth, & Brown 2001; Finn & Loomis 2001; Hunt & Ditton
2002; Sutton 2003; Kyle, Norman, Jodice, Graefe, & Marsinko 2007; Anderson, Ditton, & Hunt
2007; Hutt & Bettoli 2007).
The user profile research conducted to support natural resource agency decision making
connotes a “marketing perspective” (Ditton 1996). Many human dimensions researchers have
noted the marketing-related applications of this research, including: promoting more active and
targeted engagement within a broader competitive leisure market (Aas & Arlinghaus 2009);
designing promotional strategies (Floyd & Lee 2002; Schroeder et al. 2008); informing outreach
and communication programs (e.g. Romberg 1999; Brown 1987; Connelly, Brown, & Knuth

8

2000; Ditton 2004); developing marketing strategies and identifying market products (Scott,
Ditton, Stoll, & Eubanks. 2005; Connelly et al. 2000); targeting angler groups for public
involvement and education efforts (Romberg 1999); providing information to transform attitudes
and beliefs of citizens (Chase, Lauber, & Decker, 2001) and , increasing customer awareness and
satisfaction and addressing user group conflicts (Ditton 1996).
The cognitive and motivational theoretical approaches commonly used by market
researcher (as well as human dimensions researchers) to predict and explain human behavior
have been adapted from the psychology and social psychology fields. Cognitive hierarchies
model the impact of vales, attitudes and norms on behavioral intentions and behavior (Fishbein
& Ajzen 1975). Subsequent reconceptualization (e.g. Warhsaw 1980a 1980b) emphasized the
importance of condition-specific cognitive measures, consumer needs (i.e. motivations) and
constraints (e.g. capability, affordability, accessibility). Contemporary consumer behavior
models illustrate a wide array of internal (e.g. attitudes, motivations, perception) and external
(e.g. family, demographics, reference groups, culture) domains that may influence individual
behaviors (Hawkins & Mothersbaugh 2010). These factors interact with each other and together
drive the development of self-concept and it’s manifestation: “lifestyle”- or the way consumers
live, including the products purchased and how consumers use them, think about them and feel
about them (Hawkins & Mothersbaugh 2010, O’Shaughnessy 2013)
Recreation Participation Research
Leisure and human dimensions research related to recreation participation (including
fishing and hunting) has explored a broad range of variables at the individual-level (i.e.
cognitions, demographic characteristics, motivations) as well as broader socialization or socialstructural influences (i.e. family, society, environment). Measures of cognitions can contribute
to the understanding recreation behavior including participation patterns (Hunt & Grado 2010;
9

Vaske & Manfredo 2012). A number of demographic and socio-demographic variables have
also been found to relate to fishing participation including gender, race and age (e.g. Fedler &
Ditton 2001; Floyd & Lee 2002; Floyd, Nicholas, Lee, Lee, & Scott 2006; Hunt & Ditton 2002.)
The relatively limited research examining the relationship between motivations and fishing
participation has found associations between affiliation (Fedler 2000) as well as appreciation and
health motivation factors (Schroeder et al. 2008).
Recent research related to hunter recruitment and retention and global fishing
participation rates has advocated for expanding upon the use of individual-level variables to
include higher-order influences (see Larson, Stedman, Decker, Siemer, & Baumer 2014 and
Arlinghaus, Tillner, & Bork 2014). These researchers argue that individual decision making and
behaviors are nested within and influenced by social-structural variables. For hunting, Larson et
al. (2014) posit the concept of “social habitat”- to identify complex social and environmental
factors that impact individual actions. These variables include family, friends or mentors (i.e.
micro); communities and organizations (i.e. meso); and urbanization, media, culture, and state
and federal policies (i.e. macro) variables.
Big Data and a Commercial Geodemographic Market Segmentation Approach
Projects funded by a Sport Fish Restoration multi-state conservation grant awarded
jointly to the Association of Fish and Wildlife Agencies and the American Sportfishing
Association, included appending Esri Tapestry™ Segmentation geodemographic descriptions to
available state fishing license sales data for several states, including Michigan (see American
Sportfishing Association 2007; Southwick Associates 2007; Southwick Associates 2006). Esri’s
multivariate geodemographic segmentation strategy utilizes data reduction and clustering
techniques to profile the aggregate preferences and lifestyles of consumers residing in a
geographic area. A combination of geographic, demographic, general behavior and
10

psychographic measures totaling more than 60 attributes originate from a combination of sources
including Census data and databases containing information on consumer behavior
characteristics (e.g. Survey of the American Consumer TM from GfK MRI). Based on a variety
of socioeconomic and demographic factors, the methodology classifies US neighborhoods into
65 market segments as well as aggregated summary groups based on lifestyle and lifestage (i.e.
LifeMode Groups) and geographic and physical features (i.e. Urbanization Groups). Esri
recommends appending geodemographic market segments with available client data (e.g.
customer data) to develop more robust costumer or “lifestyle” profiles (Esri TapestryTM
Segmentation Reference Guide 2014). The combination of consumer data and “neighborhood
geography” permits profiling geographic areas to inform development and distribution of new
products and to target marketing efforts (Esri TapestryTM Segmentation Reference Guide 2014).
Geodemographics is characterized as a neighborhood classification approach, and is used
broadly in market research, market analysis and advertising (Harris, Sleight and Webber 2005).
In both the United States and Great Britain the initial availability of electronic/machine readable
small area data from the Census (1970 and 1971) spurred the fields of geodemographics and
growing interest in applications in the public and private sectors for service delivery and direct
marketing purposes (Birkin, Clarke, Clarke, & Wilson 1996). Geodemographic analysis is based
largely on the premise that lifestyle, and thus consumption is largely driven by demographic
factors (Mitchell 1995). Market research companies and data providers such as Nielsen and Esri
have also incorporated consumer data (e.g. product consumption, media usage, consumer
surveys) into their geodemographic approach.
The term “lifestyle” is used frequently within Esri TapestryTM reference manuals and
throughout the fishing license analyses (i.e. Southwick Associates 2006). Understanding the

11

complex nature of purchase behaviors prompted the development of the lifestyle concept as a
multi-measure construct incorporating important behavioral predictors including measures of
activities (i.e. behavior), interests and opinions (i.e. attitudes) - together termed, lifestyle (see
Gonzalez and Bello 2002 for a review). The lifestyle construct has been applied widely in
market research to classify consumer segments and subsequently inform marketing strategies
(e.g. Thach & Olsen 2004; Bruwer & Li 2007; Gonzalez & Bello 2002; Zhu, Wang, Yan, & Wu.
2009; Todd & Lawson 2001). These studies typically involve segmenting consumers based on
respondent’s answers to multi-measure lifestyle questions, and describing consumer preferences
and usage rates. A review of marketing service and consumer research company webpages listed
as Esri’s sources for private and public household-level information (e.g. GfK MRI and
Acxiom), suggests that widely recognized attitude, interest and opinion or “lifestyle” variables
are indeed collected through consumer surveys.
The approach used by Southwick Associates (2006) provided a mechanism for further
describing the diversity of license purchasers, in terms of culture, lifestage, household structure,
employment, media preferences, purchase behaviors, hobbies, interests and related demographic
and behavioral characteristics. Broadly, the reports identified potential sales growth areas with
the objective of increasing fishing license sales to and revenues from particular segments. The
Michigan report (Southwick Associates 2006) examined segment size, rates of decline, churn
rate, and sales to population to identify segments deserving “extra consideration” for recruitment
and retention strategies. Several Michigan segments demonstrated a higher propensity to
purchase fishing licenses. The researchers provided inferences based on dominant segment
characteristics such as income and lifestage. The preliminary nature of the analysis was

12

underscored, with recommendations for additional research to understand angler perceptions and
constraints and controlled/experimental promotional mailings to segments.
Recent research examining internet search behaviors in order to support angler
recruitment programs underscores the increased usage of internet and electronic media and the
potential of using new platforms for collecting and monitoring information spatially and
temporally (Martin, Pracheil, DeBoer, Wilde, & Pope 2012). As a result of increased
connectivity and technological advancements the amount of data generated, stored and consumed
is expected to continue to grow exponentially (McKinsey Global Institute 2011). Large-volume
data sets (i.e. “big data”) are increasingly being applied in customer analysis and behavioral
prediction across a multitude of sectors (George, Haas, & Pentland 2014). Businesses and
market research companies utilizing and developing large-volume data accumulate fine-grained,
individual-level data from multiple public and private sources originating from governmental
organizations, private business transactions, consumer purchases, information searches, social
media usage, and consumer reviews to name a few (George et al. 2014) . In addition to growing
the volume of information available for analysis, greater connectivity has altered consumer
expectations- necessitating more personalized, tailored and data-driven messaging to attract
customers and grow a customer base (Arthur 2013).
Commercial market segmentation systems built upon “big data” sources offer a
convenient method for identifying large, stable and accessible groups of customers exhibiting
similar within-group behaviors (i.e. market segments) driving the activities of an organization’s
marketing mix- or the product, place, price and promotion and communication activities
(content, timing and medium). Information on the size, growth, location and behaviors of angler
subgroups is critical for building customer relationship management information systems (Vogt

13

2010; Dann, Alvarado, Palmer, Schroeder, & Stephens 2008; George et al. 2014); informing
resource management decisions, and developing strategic marketing efforts that maximize
marketing efficiency (i.e. return on investment). With increasing access to a wide array of
variables, commercially available market segmentation systems also represent a method for
developing location-based market segments that incorporate many of the individual-level, social
psychological and social-structural variables shown to affect recreation participation rates. The
richness of statewide angler survey efforts coupled with dynamic lifestyle characterizations from
commercially available consumer segmentation systems may also provide opportunities for
further monitoring and evaluation of expanding and new angling markets. Motivated by the
growing availability of commercial segmentation software and the potential for developing
angler subgroups suitable for these efforts, our research objective is to evaluate the commercially
available, generic segmentation approach (i.e. Esri Tapestry™ Segmentation) as an angler
market segmentation methodology, specifically in developing distinct angler subgroups in terms
of their fishing and license purchase behaviors.

Methods
Data Collection
To evaluate the utility of the commercially available geodemographic classification
system (Esri Tapestry™), we utilized angler data generated from the Michigan Recreational
Angler Survey (MRAS). The MRAS is a long-term statewide angler survey that tracks angling
effort across all of Michigan’s fisheries (Simoes 2009). Following a modified Tailored Design
Method (Dillman 2007), four contacts were made, which included (1) an initial survey, cover
letter and business reply envelope, (2) a reminder/ thank you postcard, (3) a replacement survey,
14

cover letter with a more urgent appeal and business reply envelope, and (4) a final postcard
(Appendix A). The MRAS uses simple random samples of fishing licenses drawn monthly from
the Michigan Department of Natural Resources Retail Sales System (RSS) database. The MRAS
average monthly response rate over this period was 48%.
The MRAS database provides self-reported demographic information including,
household income, education, household structure and race and ethnicity as well as general
fishing characteristics and fishing behaviors specific to Michigan. Additional angler
characteristics, specifically age, gender, address and license purchasing behavior were obtained
from the RSS. For this research, we utilized data on anglers from the monthly samples from the
RSS during the 2009 – 2012 fishing seasons, N = 45,504, along with the MRAS survey
responses for these samples. All analyses of the survey data use post-stratification weights to
ensure the respondent sample matches the sample joint distribution of age and license type.
As mentioned above, in addition to describing 65 distinct Tapestry Profile Segments, Esri
Tapestry™ produces 12 aggregated summary LifeMode groups characterized by “lifestyle and
lifestage composition”. LifeMode groups encompass as few as two Tapestry Segments (High
Hopes LifeMode group) to as many as nine (Senior Styles LifeMode group). We evaluated
Esri’s 12 category LifeMode classification based on Esri’s recommendation for analyzing a
smaller number of markets and where lifestyle- or lifestage-related behavior are important.
Additionally our choice of the 12 category LifeMode classification is based on previous angler
segmentation efforts which developed a similar number of distinct angler segments (e.g. Fisher
1997, Connelly et al. 2001, Romberg 1999, Hutt & Bettoli 2007). Both respondent and
nonrespondent angler records (addresses) were geocoded using Esri Business Analyst 10.1. The
resulting Esri TapestryTM classifications were used to conduct our evaluations.

15

Analysis
As previously stated, our objective was to assess whether angler classifications based on
their geodemographic segment accurately discriminate among known fishing behaviors and
purchase behaviors of actual anglers. Many authors have offered criteria and statistical methods
to aid in the evaluation of market segments (e.g. Wilkie 1994; Kotler 1984; Smith 1989; Tonks
2009; Everitt, Landau, Leese and Stahl 2011; Milligan 1996). Depending on the context, some
combination of researcher judgment and statistical tests support segment evaluation, which
usually include: (a) accessibility (an organization must be able to easily reach or access the
identified segments), (b) durability or stability (segments need to remain relevant over an
extended period of time), (c) segment size or substantiality (segments should be large enough to
warrant separate attention), (d) identifiability or homogeneity (i.e. maximized between-segment
variation and minimized within-segment variation (e) responsiveness or defensibility (segments
differ on characteristics relevant to the organizations services or products; i.e., segments require
different marketing approaches or marketing mixes).
Analysis of bivariate correlation was used to compare our recent classification results to
previous results (Southwick Associates 2006). Chi-square tests of independence were used to
examine relationships between fishing behavior variables and LifeMode group. Given our large
sample size and in order to provide comparative descriptive measures of the degree of
association (Gilner, Vaske and Morgan, 2001) we report Cramer’s V as a measure of the strength
of the relationship between variables based on the chi-square statistic. We also report Goodman
and Kruskal Tau denoted by , to quantify the reduction in misclassification error (i.e.
proportional peduction in error). Whereas Cramer’s V is based on the value of the X2 statistic,

16

differences in values may be interpreted across studies (Agresti 1979) with larger , signifying
stronger association between variables.

is defined as:

where E1 is the number of errors expected based on the marginal distribution of the dependent
variable and E2 is the summation of errors over the categories of the independent variable based
on the conditional distribution of the dependent variable (Agresti 1979). We used One-Way
ANOVA to test between-group differences of LifeMode groups. For variables with unequal
variance, we report alternative versions of the F statistic (i.e. Welch’s adjusted F ratio). Again,
given our large sample size, small differences in group means may be statistically significant.
Therefore, eta squared (η²) and omega squared (ω²) are reported as two measures of strength of
association. Eta squared is an effect size measurement (strength of association) for the ANOVA
and is interpreted as the proportion of the variance in the continuous, dependent variable
explained by the independent variable (Tabachnick & Fidell 2001), calculated as the ratio of the
sum of squares between groups to the total sum of squares:

Although Eta squared is commonly reported, it describes only the proportion of systematic
(between groups) variance in the sample. Omega squared estimates the strength of association
between the independent variable and dependent variable in the population, accounting for the
number of groups and for error variance, or the error that cannot be explained by systematic

17

differences between groups. Therefore, Omega squared is considered a less biased measure of
magnitude of effect than is Eta squared (Tabachnick & Fidell 2001):
(

)(

)

The interpretation of relative effect sizes are discussed by Cohen (1992) and Cohen (1988) with
values of 0.01, 0.06 and 0.14 indicating a small, medium and large amount of association
between variables, respectively. In addition, Scheffe and Games-Howell post hoc procedures
were conducted to determine the total number of pairs of the twelve LifeMode groups in which
means differed significantly.

Results
We report first on geocoding results of angler records (i.e. Michigan angler segment
classification assignments) using Esri Business Analyst 10.1. This initial geocoding process
included our entire sample, both survey respondents and nonrespondents. Subsequent analysis of
behavioral variable associations included only survey respondents.
Michigan Angler Segment Classifications
Over 90% of anglers from our sample are categorized by 7 LifeMode classifications,
High Society, Upscale Avenues, Senior Styles, Family Portrait, Traditional Living, Factories and
Farms, and American Quilt (see Table 1). The median household incomes for members of the
Factories and Farms and the American Quilt are among the lowest of the twelve LifeMode
groups. These groups are also characterized by rural living and rural activities (including fishing
and hunting), employment in the manufacturing and agricultural industries, and an aging

18

population with a greater proportion of retirees. In contrast, members of the well-educated
Upscale Avenues LifeMode group are characterized by “prosperity” with median household
incomes second only to the High Society LifeMode group. Broadly, Upscale Avenues and High
Society LifeMode groups are characterized by their greater purchasing power and discretionary
income reflected by their leisure activities, mobility and investments. The remaining dominant
groups are aptly named. Members of the older Senior Styles form one of the largest LifeMode
groups, living in households without children, with varying residential choices and lifestyles and
wide income ranges attributed to retirement income or Social Security payments. Members of
the Traditional Living LifeMode group “convey the perception of middle America” with modest
incomes falling in between the extremes reported by other LifeMode groups, a higher median
age, with many anticipating retirement and completion of child-rearing responsibilities. Finally,
the median incomes of the Family Portrait LifeMode group display the greatest range, and are
characterized by youth, family life and the presence of children (Table 1).
As was reported by Southwick Associates (2006) the license buying population generally
reflects Michigan’s population with some relatively minor differences. American Quilt and
Factories & Farms are better represented in the license buying population, while above average
income High Society customers are underrepresented. There was also a significant positive
correlation between classification proportions from our recent sample of Michigan anglers from
the 2009 - 2012 fishing seasons and classification proportions reported by Southwick Associates
(2006), (r = .912, p <.001, N=12). Despite differences in methodology (prior classifications by
Southwick Associates (2006), examined the total license purchase population for 2005) the
proportions of anglers assigned to LifeMode groups are comparable, suggesting consistency with
the classification approach. Slight differences between relative proportions may be accounted

19

for by changes in the State population or changes in Esri’s data resources given the dynamic
nature of their classification methodology (Table 1).

20

Table 1 Esri Tapestry LifeMode Classifications
ESRI Tapestry™ LifeMode classifications applied to randomly sampled licensed anglers from 2009, 2010, 2011 and 2012
fishing season (N = 42,433), Southwick Associates (2006) license purchasers from the 2005 Michigan license database
(N=1,163,767) and ESRI Tapestry™ profile descriptions. Significant positive correlation between the two classification
proportions, r =.912, p < 0.001, N=12)
LifeMode
Membership

Description

MI Sample
(2009 - 2012)

Michigan 2005
License
Purchasers

Percent of
Michigan
Population

1

High Society

10.9%

9.5%

13.2%

affluent, well educated, professional employment, highest income, married,
least ethnically diverse, fast growing, travel, active

2

Upscale Avenues

16.6%

17.5%

18.2%

prosperous, well educated, above-average income,

3

Metropolis

2.7%

2.6%

7.5%

city families, urban lifestyles, row houses, public transportation, service
related jobs, diverse housing, age and incomes

4

Solo Acts

2.0%

5.4%

4.3%

city life, single, young, well educated, professionals, urban lifestyles

5

Senior Styles

14.2%

11.6%

11.9%

large group, retires, income f/ Soc. Sec. & pensions,

6

Scholars and
Patriots

0.8%

1.5%

1.6%

young, lower incomes, college and military, low home ownership,

7

High Hopes

2.4%

4.6%

3.7%

young, mobile, college educated, couples, single parents, and singles

8

Global Roots

1.2%

2.3%

2.9%

ethnically diverse, young, modest income, renters, recent immigrants, mostly
with children

9

Family Portrait

4.6%

4.1%

3.7%

fastest growing, young families, ethnically diverse, 30% Hispanic descent,
single family homes

10

Traditional Living

11.2%

15.2%

14.5%

middle America, settled families, older, slow growing neighborhoods,
traditional media use

11

Factories and
Farms

18.8%

12.5%

8.9%

lower income, rural living, manufacturing and agriculture employment, small
communities, families

12

American Quilt

14.7%

13.3%

9.5%

small towns, rural areas, mobile homes, rural lifestyle

21

ESRI Tapestry™ profile description

Esri TapestryTM LifeMode and Behavioral Variable Associations
Of the 45,504 individuals who were contacted from our simple random sample 19,635
returned surveys. After accounting for eliminated cases and undeliverable addresses the final
adjusted response rate was 48% (Appendix B). A comparison of respondents and nonrespondent characteristics revealed statistically significant differences in age, residency, gender,
and license type. The mean age of respondents was significantly higher than nonrespondents
(p<0.001), Michigan residents (X2 =58.762, df=1, p <0.001) and females (X2 =4.542, df=1, p
=0.033) were somewhat more likely to respond and a smaller proportion of Resident Restricted
and Temporary (24-hour and 72-hour licenses) license holders responded to our survey, with the
remaining commonly issued license types (Resident All Species, Non-Resident All Species,
Senior Restricted, and Senior All Species comprising a slightly larger proportion of our
respondents (X2 =1441.437, df=7, p <0.001). In order to correct for possible
response/nonresponse bias, weights for respondents were computed according to the joint
distribution of age and license type (Appendix C). All analyses of the survey data use these
weights.
Chi-square test of independence and ANOVA performed to examine the relationship
between fishing behavior variables and LifeMode were significant (see Table 2 and Table 3).
However, chi-square directional measures of strength of association Proportional Reduction in
Error (i.e. Goodman and Kruskal tau) were generally under 1%, suggesting a very weak
association. Values of Cramer’s V also suggest a small effect, with all values under .30, with a
range between 0.05 and 0.17. One-way ANOVA analyses of fishing behaviors and license
purchase behaviors revealed statistically significant main effects. However eta squared (η²) and
estimated omega squared (

) for many variables were below .02, indicating a small amount of
22

the total variation in the average behavioral variables is attributable to differences in LifeMode
classification (Cohen 1992; Cohen 1988). Further, the number of significant differences
identified through post hoc comparisons using Scheffe and Games-Howell ranged between 13
and 56.
With few exceptions, our license segmentation approach using four aggregated license
types (resident, resident senior, nonresident and temporary) produced relatively larger

values,

(as well as eta squared (η²) and estimated omega squared (ω²)) suggesting this procedure may
provide greater discrimination and may account for more of the total variation in angler
behaviors compared to the Esri Tapestry™ segmentation strategy (Table 2 and Table 3).

23

Table 2 Comparison of Segmentation Approaches
Comparison of segmentation approaches. Comparisons using twelve Esri Tapestry™ LifeMode segments, four
aggregated license types segments (senior, non-senior, nonresident, temporary and other ). Measures of strength of
association and Post hoc tests: Eta squared (η²); omega squared (ω²); Goodman Kruskal Tau (τ); Cramer's V (V); GamesHowell (GH); Scheffe (S). All values p <0.001 unless noted. Variables with values of p > 0.05 have been omitted from this
table.
Behavior variable

12 Esri Tapestry™ LifeMode

4 License Types

General Fishing Behaviors
Fishing Experience (Mean, years)

η²= 0.012

Boat Ownership (0,1)
Catch Disposition (e.g. mostly keep, catch and
release)
Fishing with a companion (0,1)
Fishing Outside State (0,1)

2

ω = 0.011

GH=39/66

η²= 0.299

ω²=0.298

GH=5/6

τ= 0.02

V= 0.141

τ= 0.018

V= 0.134

τ= 0.015

V= 0.124

τ= 0.01

V= 0.102

τ= 0.022

V= 0.05

τ= 0.007

V= 0.086

τ= 0.013

V= 0.115

τ= 0.156

V= 0.396

Fishing Frequency (1 to 3 times in past 12 months)

τ= 0.004*

V= 0.055

τ= 0.037

V= 0.28

Fish Frequency (10 or more times in past 12 months)

τ= 0.004*

V= 0.055

τ= 0.037

Species Diversity (# of species fished)

η²= 0.008

ω²= 0.007

GH= 14/66

η²= 0.058

ω²=0.058

GH= 5/6

Method Diversity (# of methods used to fish)

η²= 0.009

ω²= 0.008

S= 13/66

η²= 0.065

ω²=0.065

GH= 5/6

Total Species and Method Diversity

η²= 0.007

ω²= 0.006

GH= 13/66

η²= 0.270

ω²=0.058

GH= 6/6

Michigan Fishing Experience (Mean, years)

η²= 0.020

ω²= 0.020

GH= 13/66

η²= 0.059

ω²=0.270

GH= 6/6

Waterbody Diversity (sum of waterbody indicators)

η²= 0.003

ω²= 0.003

S= 47/66

η²= 0.060

ω²=0.060

GH= 6/6

Great Lakes Fishing (0,1)

τ= 0.004

V= 0.06

τ= 0.006

V= 0.078

Inland Lakes Fishing (0,1)

τ= 0.006

V= 0.078

τ= 0.041

V= 0.203

Rivers (0,1)

τ= 0.009

V= 0.094

τ= 0.03

V= 0.174

Fishing behavior in last 12 months
V= 0.28

Water body type

* p < 0.05

24

Table 3 Comparison of Segmentation Approaches
Comparison of segmentation approaches. Comparisons using twelve Esri Tapestry™ LifeMode segments, four aggregated
license types segments (senior, non-senior, nonresident, temporary and other). Measures of strength of association and
Post hoc tests: Eta squared (η²); omega squared (ω²); Goodman Kruskal Tau (τ); Cramer's V (V); Games-Howell (GH);
Scheffe (S). All values p <0.001 unless noted. Variables with values of p > 0.05 have been omitted from this table.
Species targeted in last 12 months

12 Esri Tapestry™ LifeMode

4 License Types

Bass (0,1)

τ= 0.003

V= 0.056

τ= 0.032

V= 0.179

Catfish (0,1)

τ= 0.018

V= 0.133

τ= 0.016

V= 0.125

Northern Pike (0,1)

τ= 0.006

V= 0.075

τ= 0.028

V= 0.166

Panfish (0,1)

τ= 0.009

V= 0.094

τ= 0.05

V= 0.224

Salmon (0,1)

τ= 0.003

V= 0.052

τ= 0.006

V= 0.078

Suckers (0,1)

τ= 0.013

V= 0.112

τ= 0.014

V= 0.118

Trout (0,1)

τ= 0.006

V= 0.077

τ= 0.007

V= 0.081

Walleye (0,1)

τ= 0.008

V= 0.091

τ= 0.021

V= 0.145

Natural Bait (0,1)

τ= 0.012

V= 0.111

τ= 0.058

V= 0.241

Artificial Bait (0,1)

τ= 0.003

V= 0.053

τ= 0.014

V= 0.12

Trolling (0,1)

τ= 0.007

V= 0.085

τ= 0.008

V= 0.091

Casting from Boat (0,1)

τ= 0.004

V= 0.065

τ= 0.022

V= 0.15

Casting from Shore (0,1)

τ= 0.003

V= 0.056

τ= 0.033

V= 0.182

Fly Fishing (0,1)

τ= 0.003

V= 0.058

Ice Fishing (0,1)

τ= 0.02

V= 0.142

τ= 0.045

V= 0.212

Vendor (e.g. MDNR E-License; major grocery /
department store; major sporting goods; other)

τ= 0.014

V= 0.113

τ= 0.013

V= 0.13

License obtained (days since season start)

η²= 0.010

ω²= 0.010

GH= 40/66

η²= 0.120

ω²=0.120

GH= 6/6

License purchasing (2001-2012)

η²= 0.022

ω²= 0.022

GH= 56/66

η²= 0.188

ω²=0.188

GH= 6/6

License purchasing (2008-2012)

η²= 0.018

ω²= 0.018

GH= 49/66

η²= 0.172

ω²=0.172

GH= 6/6

Personal computer usage (0,1)

τ= 0.028

V= 0.167

τ= 0.072

Fishing methods used in last 12 months

License purchase behaviors

25

V= 0.268

Tables 4 and Table 5 contain a subset of the variables we tested. Relatively to other
variables we tested, a larger amount of the total variation was attributable to differences in
LifeMode classification. Interestingly, these fishing behavior and license purchase variables
further highlight a dichotomy between the lower-income Factories and Farms and American
Quilt LifeMode Groups and the above-average income High Society and Upscale Avenues
LifeMode groups. In contrast to Factories and Farms and American Quilt LifeMode groups,
High Society LifeMode members were the most likely to fish outside Michigan, fished the least,
were the most likely to fly fish, and were the most likely to practice catch and release. The
fishing behaviors of the High Society LifeMode group were mirrored by the similarly affluent
Upscale Avenues LifeMode group, which once again in contrast to Factories and Farms and
American Quilt LifeMode groups, were more likely to fish outside the state of Michigan, were
less likely to keep their catch and more likely to practice catch and release. In terms of license
purchase behaviors, High Society and Upscale Avenues were more likely to use purchase a
license online directly through the MDNR e-license system, and to make those license purchases
later in the season. For more detailed statistical output which includes group means, group
percents, chi-square, ANOVA and measures of strength of association, see Appendix D.

26

Table 4 Comparison of LifeMode fishing behavior characteristics
Comparison of LifeMode fishing behavior characteristics. One-way ANOVA, Chi-square tests of independence, directional
and strength of association measures.
LifeMode

a
b
c
d

Fishing
a
Frequency

Description

Catch
b
Disposition

1 to 3
times

10 or more
times

Mostly Keep

Mostly Catch
& Release

Fly
c
fishing

Fishing
Outside
d
State

1

High Society

27.1%

44.5%

22.3%

45.9%

21.5%

36.6%

2

Upscale Avenues

21.4%

51.8%

28.1%

35.3%

16.8%

30.0%

3

Metropolis

21.7%

51.8%

38.1%

32.5%

14.1%

27.4%

4

Solo Acts

30.3%

43.2%

20.7%

48.0%

24.8%

35.7%

5

Senior Styles

19.3%

53.8%

32.3%

31.6%

15.6%

23.7%

6

Scholars & Patriots

23.0%

54.9%

25.6%

41.0%

22.5%

23.1%

7

High Hopes

20.2%

50.8%

26.0%

38.9%

18.0%

27.1%

8

Global Roots

22.6%

52.8%

25.3%

30.2%

13.5%

18.6%

9

Family Portrait

27.0%

47.3%

22.6%

42.7%

16.0%

32.9%

10

Traditional Living

20.0%

53.4%

27.7%

35.2%

15.7%

25.3%

11

Factories and Farms

18.0%

55.5%

37.8%

23.3%

15.8%

22.0%

12

American Quilt

15.9%

60.6%

36.0%

23.9%

16.5%

22.2%

20.6%
53.0%
30.6%
Average
2
X =271.461, df=55, p <.001; Tau = .004, p <.05; Cramer's V = .055, p <.001, N=18,012

32.8%

16.9%

26.8%

2

X =573.750, df=22, p < .001; Tau = .015, p <.001; Cramer's V = .124, p <.001, N=18,528
2
X =59.659, df=11, p <0.001; Tau = .003, p <0.001; Cramer's V = .058, p <0.001, N=17,495
2

X =248.509, df =11, p <.001; Tau = .013, p <.001; Cramer's V = .115, p <.001, N=18,689

27

Table 5 Comparison of LifeMode license purchasing characteristics
Comparison of LifeMode Michigan license purchasing characteristics. One-way ANOVA, Chi-square
tests of independence, directional and strength of association measures.
LifeMode

Description

Days (since
b
season start)

Vendor*
DNR

a

Grocery /
a
Department

Sporting
a
Goods

Other

Computer Usage

c

a

90.4
12.5%
31.4%
15.0%
41.2%
90.1%
High Society
84.1
2
7.1%
35.2%
12.1%
45.6%
82.8%
Upscale Avenues
76.9
3
4.6%
36.4%
11.2%
47.8%
69.5%
Metropolis
90.7
4
9.6%
37.5%
13.4%
39.5%
88.1%
Solo Acts
78.9
5
4.5%
32.8%
9.5%
53.2%
76.4%
Senior Styles
Scholars and
90.2
6
7.7%
46.9%
12.8%
32.5%
90.8%
Patriots
87.2
7
5.5%
44.4%
9.4%
40.8%
83.0%
High Hopes
85.7
8
5.1%
43.3%
9.1%
42.5%
71.4%
Global Roots
87.9
9
8.2%
38.3%
11.6%
41.9%
85.0%
Family Portrait
81.9
10
4.3%
38.9%
11.8%
45.0%
76.7%
Traditional Living
Factories and
76.3
11
3.6%
33.3%
5.3%
57.7%
70.6%
Farms
75.9
12
3.8%
29.6%
6.2%
60.4%
70.5%
American Quilt
81.5
5.8%
34.4%
9.8%
50.0%
78.0%
TOTAL
a 2
X =1804.836, df=33,p <.001; Tau = .014, p <.001; Cramer's V = .113, p <.001, N=47,266
b
Lavene = (df1=11,df2=44,832),7.738, p<.001; Welch's F=(df1=11,df2=6340.987)= 40.905, p <.001; ω²= 0.0098; η²=0.0100;
Games-Howell 40/66, α =(.05)
c 2
X =516.492, df=33,p <.001; Tau = .028, p <.001; Cramer's V = .167, p <.001, N=18,542
* Vendor categories indicate the method used for purchasing fishing licenses: DNR= Michigan’s online e-license system;
Grocery/Department= large chain grocery and department store businesses; Sporting Goods: large chain sporting goods
businesses; Other= small and locally owned businesses
1

28

Discussion
Southwick Associates (2006) classified licensed anglers using Esri Tapestry™
Segmentation and concluded that the segment lifestyles descriptions can help discern likely
license purchasers and inform cost-effective marketing and recruitment programs. Here, we
build upon those efforts by using self-reported fishing behavior data and license purchase data to
evaluate this commercially available, generic segmentation approach as an angler market
segmentation methodology, specifically in developing distinct angler subgroups in terms of their
fishing and license purchase behaviors. Our discussion begins by summarizing our findings
regarding the market segment evaluation criteria mentioned above: segment accessibility,
stability, size and finally aspects of segment behavioral variables- identifiability and
responsiveness.
With respect to segment accessibility, geodemographic classifications provide locationbased market segment descriptions. Esri Tapestry™ Segmentation categories are developed by
combining Census data with consumer behavior characteristics to profile the aggregate
preferences and lifestyles of people who live in a geographic area (i.e. block groups, Census
tracts, ZIP codes). Therefore, these descriptions meet the accessibility criteria suggesting they
might be suitable for developing accessible market segments for use in direct mailing campaigns
to specific locations as part of outreach, education or promotional communication strategies, or
informing regional resource management decisions (e.g. county, watershed, management unit).
Comparisons of our classification results to previous research (Southwick Associates
2006) provided a measure of classification stability or durability. Our findings mirrored the
dichotomy or “split profile” reported by Southwick Associates (2006), with, broadly, a larger

29

segment comprised of anglers with average or lower than average incomes living in workingclass or rural communities (e.g. Factories and Farms and American Quilt), and another
suburban, above-average income segment (e.g. High Society and Upscale Avenues). Bivariate
correlations were positive and significant suggesting this classification methodology may be
stable overtime.
With respect to segment size, several LifeMode groups representing large portions of the
angling public were identified. Specifically, seven LifeMode classifications, High Society,
Upscale Avenues, Senior Styles, Family Portrait, Traditional Living, Factories and Farms, and
American Quilt encompassed over 90% of the angling population. These market segments each
represent substantial portions of the angler market, which may warrant tailored marketing
approaches. Of these seven LifeMode classifications, the Family Portrait LifeMode group is the
smallest segment, however given the focus of previous license campaigns on families and
children, this angler segment may be of particular importance to state agencies.
For several of the fishing behaviors and license purchase behavior variables examined,
we note modest differences between segments classified by Esri TapestryTM LifeMode.
Statistically significant and meaningful differences in fishing and license purchase behaviors
would suggest that these groups might respond differently to marketing mixes, necessitating the
development of tailored marketing approaches. However, as previously stated, additional
statistical tests examining of the association between variables (i.e. Goodman Kruskal Tau,
Cramer’s V and ANOVA) suggest weak relationships. Therefore, the LifeMode classification
approach did not meet identifiability or homogeneity segmentation criteria (i.e. maximized
between-segment variation and minimized within-segment variation) or responsiveness or
defensibility segmentation criteria (i.e. meaningful differences in behavioral characteristics

30

requiring different marketing approaches). Thus, the exclusive use of LifeMode classifications
to reach anglers with specific behaviors is unlikely to be effective. In their present form, the
aggregated and likely demographically-driven Esri TapestryTM LifeMode classification system
should perhaps be limited to inquiries projecting angling participation or targeting demographic
groups or geographic regions for general information and education purposes rather than as a
basis for reaching customers with more specific angling behaviors.
Weak associations identified in our research may be partially explained by (a) the
nonspecific nature of lifestyle measures and other segmentation variables used in this generic
commercial segmentation approach, (b) the diverse demographic and socioeconomic
characteristics of 65 Esri Tapestry™ segments encompassed by the LifeMode groups we tested,
(c) and the weak relationship between demographic variables and angling behavior. While the
Lifestyle construct is widely applied in market research, many researchers, using the Lifestyle or
similar constructs, have made a distinction between general and more service-specific and/or
salient attitude measures (e.g. Pierce, Manfredo, and Vaske 2001, Lawson and Todd 2002;
Gonzalez and Bello 2002, Crompton & Lamb 1986, Vaske and Manfredo 2012). The use of
generic lifestyle segmentation descriptions may present an additional challenge in determining
content as well as construct validity- or what is ultimately being measured (Tonks 2009).
Although we report weak associations between the behavior variables and market
segments we tested, further examination of these classification systems may be warranted. In
light of the availability of less-aggregated Esri Tapestry™ segments, it is possible that some of
the 65 less aggregated segments may meet the identifiability or homogeneity segmentation
criteria (i.e., they may produce some more homogeneous market segments that have
distinguishable behaviors). Moreover, with the growing volume of consumer data and advances

31

in large-volume data analysis (George et al. 2014; McKinsey Global Institute 2011), we expect
businesses, data compilers, and marketing firms will continue to be innovative in their collection
and application of consumer information resources. In the meantime, researchers and natural
resource managers interested in further developing angler segments for recruitment and retention
activities might consider developing statistical models that include spatially-explicit social
structures (Larson et al. 2014) and new and salient consumer behavior variables from the
burgeoning geodemographic field.

32

CHAPTER 2

Describing anglers using an inland lake classifications system

Motivation
Classification systems are fundamental to most branches of science in which researchers
use “data mining techniques” to uncover data structure and organize large data sets so that the
information may be more easily understood and analyzed (Everitt et al. 2011). Classification
systems for natural resources exist for both terrestrial and aquatic ecosystems and at multiple
scales (e.g. USFWS National Wetlands Inventory, Gawler (2008), USDA Watershed Condition
Framework, Krost et al. 2007). To predict and model species distributions, researchers have
utilized existing habitat categories or developed new classifications systems for a variety of
ecosystems, including: boreal wetlands (Morissette 2013), ocean reefs (Malcolm et al 2010)
streams systems (Frissell et al. 1986), estuaries (Schoch 2012), inland lakes (Wehrly et al. 2012),
woodland patches (Collier et al. 2012) and invertebrate habitat (Duffey 2010). Resource
managers faced with the challenge of managing heterogeneous systems across large landscapes
often with limited resources, utilize classification approaches to describe and inventory natural
resources, plan for and prioritize management and conservation activities, identify important
habitat-species associations or predict habitat conditions and species occurrences.
For large states with an abundance of inland lakes, lake distribution and lake diversity
presents managers and policy makers with additional challenges in terms of adequately
sampling, describing, and predicting responses from environmental changes and management
activities (Wehrly et al. 2012). Utilizing readily available spatially extensive variables
(including lake network position, morphometry, connectivity and lake thermal regime), Wehrly
33

et al. (2012) identified six distinct inland lake classes and fish assemblages for Michigan’s inland
lakes. The classifications will enable managers to characterize Michigan’s inland lake fishery
resources, allocate monitoring efforts, inform management decisions including stocking, predict
inland lake responses to management actions and communicate and validate management
decisions. The authors focused on lakes 4-ha and larger because of their ability to support
exploitable sportfish populations. Although these classifications are ecologically meaningful,
they do not directly account for angler behaviors.
Classification of inland lake anglers
Fisheries represent complex socio-ecological systems (Martin and Pope 2010) and
licensed anglers represent one of the most important constituent groups for state fishery
managers (Connelly et al. 2013). In lake-rich states such as Michigan, the multitude of inland
lakes, their diverse morphological, chemical and biological characteristics and differing fisheries
management strategies result in wide array of angling experiences available to anglers. Inland
lake fishing is the predominant angling activity reported by Michigan’s recreational anglers
(Jamsen 1985, Simoes 2009) and represents the majority of angler’s reported fishing trips
(Simoes 2014). Further, the species most often targeted by Michigan’s anglers include species
and species classes found largely in inland lakes (e.g. panfish, bass and walleye) (Simoes 2009).
The development of biological and social classification systems to describe complex
fishery problems has been identified as a major challenge in fisheries science and management
(Connelly et al. 2001). Human dimensions researchers have underscored the importance of
applying social data to fisheries management decision making, including the allocation of
management funds (Fisher 1997), clarification of management objectives (Dobson et al. 2005),
implementation of regulations with knowledge of which angler groups will be most affected

34

(Fisher 1997; Manfredo et al. 1998), legitimizing fishery policies with stakeholders (Sutton and
Tobin 2009, Kruegger and Decker 1999; Brown 1987, Hunt and Grado 2010) and, understanding
the impact of anglers on fishery resources (Lupi et al. 2005).
Fisheries management typically occurs on a species-by-species and lake-by-lake basis
(Tonn et al. 1983). Therefore, a useful segmentation approach involves identifying and
describing anglers who participate in specific fishery locations or fisheries (Romberg 1999).
When combined with information on other aspects of a fishery (resource integrity, catch rates,
physical setting, assess, etc.) managers can develop a reliable description of the fishery which
includes a description of its participating anglers, and the factors that make a given fishery
appealing to specific angler markets (Romberg 1999).
The goal of this research was to develop managerially-relevant inland lake angler
segments. Because many states have developed biological classification systems of fisheries
resources, our goal is to describe the characteristics of angler groups using a biological
classification system to describe inland lake anglers in terms that are useful to managers. The
research is motivated by the availability and scope of both the Michigan Recreational Angler
Survey data and landscape-based classification of inland lake fish assemblages (i.e. Wehrly et al.
2012). Further development of classification systems for Michigan’s fisheries using angler
survey data is warranted and may prove particularly informative for guiding management and
monitoring programs and communicating both the ecological and sociological reasons for inland
lake management decisions.
Our research included the following objectives: (1) Examine the characteristics of inland
lake anglers for each lake class; (2) and, Develop a model to predict the types of anglers that take
trips to particular lake classes.

35

Methods
Data Collection
The population of interest included anglers who reported trips to inland lakes in
Michigan. We used data from the Michigan Recreational Angler Survey (MRAS) a monthly
mail survey of anglers in Michigan. Our sample frame for the MRAS was the Michigan
Department of Natural Resources Retail Sales System database. For this research, we utilized
monthly simple random samples drawn from the Retail Sales System database during the 2008 –
2012 fishing seasons. Following a modified Tailored Design Method (Dillman 2007), four
contacts were made by mail, which included (1) an initial survey, cover letter and business reply
envelope, (2) a reminder/ thank you postcard, (3) a replacement survey, cover letter with a more
urgent appeal and business reply envelope, (4) a final postcard (Appendix A). The MRAS
survey response rate for the months of data used here was 46% after adjusting for undeliverable
mail and deceased persons.
The MRAS database provides a rich source of self-reported demographic information as
well as general fishing characteristics and fishing behaviors specific to Michigan. Additional
angler characteristics, specifically age, gender, and license type were obtained from the Retail
Sales System. Information for angler trips includes the location of the trip, whether fishing was
the primary purpose of the trip, whether the trip was an overnight trip, and whether the trip
involved multiple rivers and lakes. More specific trip details included the fishing mode and
which fish species were targeted.
Angler’s reported the inland lake name, the county the lake was located and the nearest
city. Nonresponse, incomplete answers and misspellings can confound the determination of the
exact inland lake a trip was taken. However we used an iterative process to match and decipher
angler’s trip location data to unique inland lake codes using Esri Python. Following this process,
36

unique lake identification codes were matched to the lake classification system developed by
Wehrly et al. (2012). See Appendix E for lake classification procedures and results of inland
lake coding and lake classification assignment. We cross-referenced the trips to identified inland
lakes against the classification system developed by Wehrly et al. (2012), assigning lake classes
(i.e. Lake Class 1...Lake Class 6) to 1,886 unique inland lakes. In summary, a total of 14,889
inland trips from our original sample 16,241 were assigned one of the six available lake
classification number and used in our analysis (Appendix E).
Model
Fishing trips to one of the six available lake classes constitutes an unordered, multinomial
discrete outcome. A multinomial probit model was used given our observed outcome variable
(lake class) and the distribution of our continuous and dichotomous predictor variables.
The stochastic error terms for multinomial probit are assumed to have independent,
standard normal distributions. As per Greene (2002):

[

Where

]

is an index function that describes angler i’s preference for alternative j, xi are

characteristics of the anglers (e.g., behaviors and demographics), and

are the parameters to be

estimated that relate the angler characteristics to the J alternatives. The term in the loglikelihood that corresponds to i’s choice of alternative q is

[

]

37

By normalizing the estimated parameters against an arbitrarily chosen baseline (

), we can

write the probability as follows:

[

( )(

)

Without loss of generality, one can then further normalize so that
estimation of the J-1 of the preference parameters relative to

( )(

)]

, which allows
Because the observed

outcome variable (lake class) represents six discrete lake classes (i.e. 1 - 6), the multinomial
probit model will identify five sets of estimated parameters that explain the relationship between
angler characteristics,

and factors related to an angler’s lake class choice. The most frequent

inland lake class outcome was Class 1 (41% of cases). Class 1 was thus chosen as the baseline
so that the predicted parameters measure the attractiveness of other inland lake classes relative to
Class 1 inland lakes. The predictor variables are defined in Table 6. In this paper we refer to
groups of species by their general association with water type. In our results and discussion,
largemouth bass and panfish are considered warmwater species, smallmouth bass, northern pike,
yellow perch and walleye are considered coolwater species, and trout and salmon are considered
coldwater species.

38

Table 6 Predictor variables for the multinomial probit model
Variable
Demographics
Age
Age_Squared
Gender
Michigan Resident (1/0)
Restricted License (1/0)
Trip Characteristics
Main Purpose (1/0)
Overnight Trip (1/0)
Multiple Lakes/Rivers (1/0)
Fishing Mode
Boat (1/0)
Wade/Shore (1/0)
Ice fishing (1/0)
Species Targeted
Warmwater
Largemouth Bass (1/0)
Panfish Species (1/0)
Coolwater
Smallmouth Bass (1/0)
Northern Pike (1/0)
Yellow Perch (1/0)
Walleye (1/0)
Coldwater
Trout Species (1/0)
Salmon Species (1/0)

Definition
Angler age (years)
Angler age (years), squared
1= male
1= Michigan resident
1= restricted (no trout and salmon)
1= fishing was main purpose of trip
1= overnight fishing trip
1= visited multiple waterbodies
1=angler used a boat for fishing
1= angler waded or shore fished
1= angler ice fished

Variable
Type
continuous
continuous
binary
binary
binary
binary
binary
binary
binary
binary
binary
binary
binary
binary

1= species was targeted
1= species was targeted

binary
binary

1= species was targeted
1= species was targeted
1= species was targeted
1= species was targeted

binary
binary
binary
binary

1= species was targeted
1= species was targeted

binary
binary

Marginal effects
The multinomial probit parameters provide information on the direction and statistical
significance of the parameter, but give no understanding of the practical size of the effect of the
variable. Marginal effects show how the likelihood shifts among lake class options due to an
incremental change in the predictor/independent variables, demonstrating the sensitivity of lake

39

class choice to the characteristics of anglers (Greene,2002). Because probabilities must sum to
one, the marginal effects across the J classes will also some to one (up to any rounding errors).

Results and Discussion
Descriptive Statistics
Before discussing the results of the multinomial probit model, we briefly discuss features
that characterize our data. Table 7 combines inland lake characteristics adapted from Wehrly et
al. (2012) with angler demographic and trip characteristics (means and percents) from MRAS
and license data. Inland lake classes were well represented by Lake Classes 1 and 2. Fishing
trips to these lake classes accounted for 41% and 26% of angler trips, respectively (Table 7).
Results of statistical tests from this analysis suggests there is a relationship between the outcome
variable (Lake Class) and several demographic and trip characteristics.
Inland lakes in Lake Classes 1 and 2 are characterized by high degree days, high mean
temperatures and distributions primarily within the lower peninsula of Michigan. In both lake
classes, fishing was more likely to be the main purpose of the trip, anglers were less likely to
make overnight trips and also somewhat less likely to fish at multiple sites. For the abundant,
deep, warmwater, southern lower peninsula inland lakes categorized as Lake Class 1, warmwater
species and especially panfish species were targeted at a relatively higher rate compared to other
lake classes. In Lake Class 2 a mix of warmwater, coolwater and coldwater species were
targeted. Compared to other lake classes, somewhat higher targeting of largemouth bass,
smallmouth bass and salmon species occurred in Lake Class 2. Inland lakes in Lake Class 3 are
concentrated in the western upper peninsula and are described as having low-degree days, low
mean temperatures, high depth and large surface area. In these lakes, coolwater species are the
40

prominent indicator species, but the lake class is also characterized by the presence of coldwater
species. Smallmouth bass, trout and especially walleye were targeted at a relatively higher rate
by anglers fishing in these lakes. Compared to other lake classes, anglers fishing in inland lakes
classified as Lake Class 3 were also less likely to hold a restricted license and were somewhat
more likely to make overnight trips. Inland lakes in Lake Classes 4 and 5 both occupy northern
climes (upper peninsula and northern lower peninsula). However compared to Lake Class 5,
inland lakes in Lake Class 4 are relatively smaller, more common and the prominent indicator
species include more coolwater species, coldwater species and fewer species tolerant of low
oxygen. Panfish species and most coolwater species comprised similar proportions of species
targeted by anglers in Lake Class 4 and 5, however anglers fishing in Lake Class 4 targeted
walleye at a relatively lower rate and trout were targeted at a much higher rate. Anglers also
reported similar trip characteristics to Lake Classes 4 and 5- for both inland lake classes about
half of the anglers fishing reported overnight trips and a relatively small percentage reported that
fishing was the main purpose of the trip. However compared to most other lake classes, anglers
fishing in the larger coolwater lakes (Lake Class 5) were more likely to be residents, use a boat,
hold restricted licenses, and more likely to fish exclusively at this site. Finally, inland lakes in
Lake Class 6 are found primarily in the upper peninsula, are described as shallow, having lowdegree days, intermediate mean temperatures and intermediate surface areas. The prominent
indicator species include white sucker, walleye, yellow perch, northern pike. Anglers reported
targeting the latter three species at somewhat higher rates compared to other lake classes.
Anglers fishing in these lakes were also much less likely to be residents, reported more overnight
trips, and trips to multiple sites.

41

These descriptive results suggest the lake classification developed by Weherly et al.
(2012) may also serve to classify anglers in terms of their fishing behaviors and trip
characteristics. The next section presents the results of the multinomial probit classification
model.

42

Table 7 Lake characteristics, classification, and angler characteristics
Lake class characteristics and sportfish species in order of prominence (adapted from Wehrly et al. 2012) and angler demographic and trip
characteristics from the Michigan Recreational Angler Survey.
Lake Class
Lake Class Number
Lake Characteristics
Species abundance
relative size
depth (mean)
temperature (mean)
percent of lakes
surface area %
location
Lake Classifications
Number (Total = 14,889)
Percent
Angler Characteristics
Age (mean)
Gender (% female)
Resident (% Yes)
License (% restricted)
Main Purpose (% Yes)
Overnight Trip (% Yes)
Number Days (mean)
Multiple Rivers / Lakes
Boat fishing
Shore/Wade fishing
Ice fishing

1

2

3

4

5

6

warmwater sp
small
relatively high depth
highest
59%
25%
SLP, Coastal NP

coolwater sp
large
deeper
similar
less abund
22%
LP

coolwater sp
large
deeper
lowest
less abund
22%
Western UP, N

warmwater sp
small

coolwater sp
intermediate

similar
33%
< 10%
UP, NLP

similar
rare
14%
NL, UP

coolwater sp
intermediate
lowest depth
similar
rare
8%
NL, UP

6,172
41.5%

3,816
25.6%

1,513
10.2%

822
5.5%

2,037
13.7%

529
3.6%

48.99
21%
90%
63%
75%
25%
1.79
16%
68%
18%
9%

48.98
22%
84%
56%
75%
31%
2.16
16%
69%
17%
9%

49.4
21%
74%
43%
67%
54%
2.87
23%
74%
9%
13%

48.92
23%
80%
50%
63%
51%
2.73
27%
67%
19%
8%

49.76
26%
90%
61%
62%
52%
2.79
17%
74%
11%
11%

50.98
28%
63%
48%
67%
59%
3.22
29%
76%
9%
11%

43

Table 7 (Cont’d)
Lake class characteristics and sportfish species in order of prominence (adapted from Wehrly et al. 2012) and angler demographic and trip
characteristics from the Michigan Recreational Angler Survey.
Lake Class
Lake Class Number
Species Targeted
Warmwater
Largemouth Bass
Panfish Species
Coolwater
Smallmouth Bass
Northern Pike
Yellow Perch
Walleye
Coldwater
Trout Species
Salmon Species
Total

1

2

3

4

5

6

25%
36%

22%
26%

9%
13%

21%
26%

16%
23%

13%
20%

11%
10%
12%
5%

15%
11%
13%
10%

16%
14%
18%
25%

13%
14%
13%
8%

14%
15%
15%
15%

12%
17%
18%
18%

1%
1%
100%

2%
2%
100%

4%
1%
100%

5%
0%
100%

1%
0%
100%

1%
0%
100%

44

Multinomial Probit Model
The multinomial probit model was estimated using the 14,889 total usable observations. The
observed distribution of trips for Lake Classes 1 through 6 were 42%, 26%, 10%,6% 14% and
4% respectively (shown in Table 7); the fitted class probabilities (shown in the column headings
of Table 8) closely match the data, although the model slightly over-predicts Class 2 and underpredicts Class 3. We present the marginal effects from the multinomial probit model in Table 8.
The majority of the variables in our model were discrete (1/0) regressors, with the exception of
our continuous age variables, Age and Age Squared. Holding other variables in the model
constant, we interpret the direction and magnitude of the marginal effect of each variable on the
probability of fishing in a particular lake class. The empirical analysis largely supported the
descriptive analysis of the survey. In general, angler demographic predictors in our model were
relatively less influential in predicting the lake class visited, compared to characteristics of the
trip and especially species targets.

45

Table 8 Marginal effects, standard errors and significance levels from probit estimates

Demographics
Age
Age_Squared / 1000
Gender
Michigan Resident (1/0)
Restricted License (1/0)
Trip Characteristics
Main Purpose (1/0)
Overnight Trip (1/0)
Multiple Lakes/Rivers (1/0)
Fishing Mode
Boat (1/0)
Wade/Shore (1/0)
Ice fishing (1/0)
Species Targeted
Warmwater
Largemouth Bass (1/0)
Panfish Species (1/0)
Coolwater
Smallmouth Bass (1/0)
Northern Pike (1/0)
Yellow Perch (1/0)
Walleye (1/0)
Coldwater
Trout Species (1/0)
Salmon Species (1/0)
* p < 0.10; **p < 0.05; ***p < 0.001

Lake Class 1
Pr(Class =1) =0.4235
Marginal effect
S.E.
0.0036**
0.0016
-0.0436**
0.0000
0.0665***
0.0105
0.0948***
0.0127
0.0833***
0.0091

Lake Class 2
Pr(Class =2) =0.2890
Marginal effect
S.E.
0.0010
0.0014
-0.0120
0.0170
0.0053
0.0096
-0.0745***
0.0123
-0.0172**
0.0082

Lake Class 3
Pr(Class =3) =0.0641
Marginal effect
S.E.
-0.0017**
0.0007
0.0000**
0.0000
-0.0131**
0.0054
-0.0514***
0.0074
-0.0376***
0.0045

0.0095
-0.1504***
-0.0095

0.0109
0.0101
0.0117

0.0480***
-0.0536***
-0.0413***

0.0095
0.0091
0.0101

-0.0049
0.0474***
0.0151**

0.0049
0.0053
0.0057

-0.0392*
-0.0203
-0.0694**

0.0200
0.0219
0.0238

-0.0050
0.0289
-0.0115

0.0182
0.0207
0.0222

0.0172*
-0.0023
0.0319**

0.0090
0.0110
0.0148

0.1057***
0.1694***

0.0097
0.0088

0.0298***
-0.0421***

0.0088
0.0082

-0.0770***
-0.0915***

0.0045
0.0051

-0.1434***
-0.0606***
-0.0988***
-0.2496***

0.0102
0.0109
0.0100
0.0100

0.0579***
-0.0287**
0.0042
-0.0006

0.0099
0.0096
0.0093
0.0100

0.0562***
-0.0039
0.0449***
0.1361***

0.0062
0.0047
0.0056
0.0077

-0.0330
-0.1020**

0.0243
0.0332

-0.0367*
0.2867***

0.0204
0.0337

0.0490***
-0.0314***

0.0133
0.0086

46

Table 8 (Cont’d)
Lake Class 4
Pr(Class =4) =0.0587
Marginal effect
S.E.
-0.0004
0.0007
0.0047
0.0000
-0.0047
0.0052
-0.0078
0.0060
-0.0178***
0.0044

Demographics
Age
Age_Squared / 1000
Gender
Michigan Resident (1/0)
Restricted License (1/0)
Trip Characteristics
Main Purpose (1/0)
Overnight Trip (1/0)
Multiple Lakes/Rivers (1/0)
Fishing Mode
Boat (1/0)
Wade/Shore (1/0)
Ice fishing (1/0)
Species Targeted
Warmwater
Largemouth Bass (1/0)
Panfish Species (1/0)
Coolwater
Smallmouth Bass (1/0)
Northern Pike (1/0)
Yellow Perch (1/0)
Walleye (1/0)
Coldwater
Trout Species (1/0)
Salmon Species (1/0)
* p < 0.10; **p < 0.05; ***p < 0.001

Lake Class 5
Pr(Class =5) =0.1367
Marginal effect
S.E.
-0.0018*
0.0011
0.0248**
0.0000
-0.0340***
0.0078
0.0842***
0.0063
0.0039
0.0063

Lake Class 6
Pr(Class =6) =0.0279
Marginal effect
S.E.
-0.0006
0.0005
0.0091*
0.0000
-0.0200***
0.0042
-0.0452***
0.0060
-0.0146***
0.0030

-0.0114**
0.0331***
0.0286***

0.0051
0.0052
0.0061

-0.0437***
0.1005***
-0.0081

0.0077
0.0076
0.0077

0.0026
0.0231***
0.0153***

0.0031
0.0036
0.0041

-0.0109
0.0016
-0.0114

0.0094
0.0101
0.0100

0.0308**
-0.0072
0.0547**

0.0134
0.0158
0.0206

0.0070
-0.0007
0.0056

0.0061
0.0075
0.0089

-0.0011
-0.0085**

0.0045
0.0043

-0.0401***
-0.0126**

0.0065
0.0062

-0.0173***
-0.0147***

0.0029
0.0030

-0.0089*
0.0156**
-0.0023
-0.0196***

0.0047
0.0055
0.0048
0.0046

0.0398***
0.0586***
0.0303***
0.1046***

0.0077
0.0080
0.0072
0.0087

-0.0016
0.0191***
0.0218***
0.0292***

0.0032
0.0040
0.0038
0.0044

0.0942***
-0.0489***

0.0169
0.0053

-0.0654***
-0.0833***

0.0126
0.0157

-0.0081
-0.0211***

0.0058
0.0043

47

Demographics
The marginal effect of our age variables (Age and Age Squared) was small (< 0.01), i.e.
less than 1 percentage point, and statistically significant for only 3 lake classes (see Table 8,
rows 3 and 4). Controlling for other variables, a one unit change in Age and Age Squared had a
small influence on the probability of fishing in a particular lake class. While our analysis of
group means demonstrated statistically significant age differences across lake classes, those
differences may not be of practical importance because mean ages were quite similar across lake
classes.
Compared to our age variables, our model suggests that Gender had a larger impact on
the propensity to fish at particular lake classes (Table 8, row 5). Holding all other variables
constant, being male significantly increased the probability of fishing in Lake Class 1 by 7
percentage points. For the remaining lakes, where the marginal effect was statistically
significant, being male reduced the likelihood of fishing. Specifically, in Lake classes 3, 5 and 6,
being male decreased the probability of fishing by 1 to 3 percentage points (-0.0131, -0.0340, 0.0200 respectively).
Our residency variable (Michigan Resident) was positive for Lake Classes 1 and 5 and
correspondingly negative for Lake Classes 2, 3 and 6 (Table 8, row 6). Being a resident
increased the probability of fishing in Lake Classes 1 and 5 by 8 to 9 percentage points
(0.0948;0.0842, respectively) and reduced the probability of fishing in Lake Classes 2, 3 and 6
by 5 to 7 percentage points (-0.0745; -0.0514; -0.0452, respectively). These results are also
consistent with our descriptive analysis (see Table 7).

48

We collapsed our license categories into one discrete license variable (Restricted
License) (Table 8, row 7)1. In broad terms, anglers in our sample possess some form of a
restricted fishing license or alternatively, an “all species” fishing license. The latter license
category permits trout and salmon fishing. Restricted license holders had a significantly higher
propensity for fishing in Lake Class 1 (0.0833), while for the remaining lakes the marginal
effects were negative or not significant. Holding a restricted license reduced the probability of
fishing in Lake Classes 3, 4 and 6 by 2 to 4 percentage points, (-0.0172; -0.0376; -0.0178; 0.0146). These results are somewhat consistent with the indicator species supported by these
lake classes and the estimated marginal effects of these targeted species variables (discussed
below).
Trip Characteristics
Our variable representing whether fishing was the main purpose of the trip (Main
Purpose) was positive for the southern inland lakes (Lake Class 2), and negative or not
significant for the remaining lakes (Table 8, row 9). The southern lakes in Lake Class 2 are
located relatively closer to large population centers. Controlling for other variables in the model,
trips focused on a single water body were more likely to occur in these lakes. In contrast, fishing
trips that might be part of some other travel to the region were more likely to occur in the
generally smaller and less common northern inland lakes (Lake Classes 4 and 5).
Our overnight variable (Overnight Trip), representing whether a fishing trip was part of
an overnight trip, was an important determinant in our model (Table 8, row 10). The marginal
1

In our sample, the fishing licenses issued to anglers included designations for residency (Michigan resident and
nonresident) and seniors (anglers 65 or older). These licenses are further specified as either ““Restricted” or “All
Species”. The latter permits fishing for trout and salmon. Combined, these licenses comprised 93% of our sample.
The remaining less commonly issued license types (e.g. Military, 24hour and 72 hour) permit fishing for trout and
salmon (i.e All Species). We aggregated all license types in our sample into one discrete license variable based on
the Restricted and All Species designation. Across all license types, 57% are Restricted, with the remaining 43%
All Species.

49

effect of our overnight variable was negative and significant for the less remote inland lakes
classified as Lake Classes 1 and 2 (15 and 5 percentage points, respectively) and correspondingly
positive for the remaining relatively remote and/or less abundant inland Lake Classes (effects
ranged from 2 to 10 percentage points).
The primary mode of fishing was disaggregated from one categorical variable, into three
separate discrete variables, specifically: those involving a boat (Boat); those involving fishing
from the shore or wading (Wade/Shore); and, ice fishing (Ice) (Table 8, rows 13, 14 and 15).
The marginal effects were significant only for Lake Classes 1, 3 and 5 and only for the Boat and
Ice fishing variables. Given the general climate and ice free periods, the results are not
surprising-- anglers who reported ice fishing had a higher propensity for fishing in lakes
predominantly found in the north (Lake Classes 3 and 5), and a lower propensity for fishing in
the south (Lake Class 1). Similarly, fishing from a boat increased the probability of fishing in
Lake Classes 3 and 5 and decreased the probability of fishing in Lake Class 1.
Species Targeted
Eight variables representing fish species and species groups targeted on a trip were
examined in our model. The marginal effects for each of the species-targeted variables were
statistically significant in at least 4 of the 6 inland lake classes. Marginal effects were on average
larger compared to our other variables, and ranged from 1 percentage point to 29 percentage
points. Generally the results support the results of our descriptive analysis and show that specific
angler-targeted species or classes of species (i.e. warmwater, coolwater or coldwater) were
significant predictors of lake class choice in our model.
Controlling for other variables in our model, anglers targeting largemouth bass and
anglers targeting panfish were more likely to fish in Lake Class 1, which are generally southern

50

lower peninsula and coastal northern peninsula lakes. The marginal effects for these variables
were also relatively large (11 and 17 percentage points, respectively) while for the remaining
coolwater and cold water species, marginal effects were large and negative for Lake Class 1. In
contrast to inland lakes classified by Lake Class 1 the marginal effects of the targeted species
variables for Lake Class 2 lacked a clear warmwater, coolwater or cold water species focus.
However, targeting salmon significantly increased the probability of fishing in Lake Class 2 by
29 percentage points. Although our descriptive analysis demonstrated that salmon comprise just
2% of the total species targeted by anglers, that proportion was relatively large compared other
lake classes. For Lake Class 3, the marginal effects for warmwater species (largemouth bass and
panfish) were negative. However, fishing for most of the coolwater or coldwater species in our
model (with the exception of Northern Pike and Salmon) increased the likelihood of fishing in
Lake Class 3. In particular, fishing for walleye significantly increased the probability of fishing
in Lake Class 3 (by 14 percentage points). Results from our model suggest Lake Class 4
supports an inland lake coldwater fishery for trout. The marginal effect for targeting trout was
large (10 percentage points) and positive for this lake class. The marginal effects for most of the
remaining species were negative or only marginally positive. As was the case with targeting
salmon for inland lakes classified as Lake Class 2, trout do not comprise a large proportion of the
total species targeted by anglers in Lake Class 4; however the 4% that targeted trout at Lake
Class 4 was relatively large compared to the proportions that targeted trout at most other lake
classes. This was further supported by the propensity of “all species” license holders to fish in
Lake Class 4 (see above).
The species profiles from our descriptive analysis for Lake Classes 5 and 6 were
relatively uniform and do not reveal a unique fishery for Lake Classes 5 and 6 as was the case in

51

other lakes classes (e.g. the dominance of warmwater species in Lake Class 1, salmon in Lake
Class 2, walleye in Lake Class 3, or trout in Lake Class 4). However, whereas targeting
particular species or warmwater species or a mix of coolwater and coldwater species increased
the likelihood of fishing in other lake classes, controlling for other variables in the model,
anglers targeting several coolwater species were more likely to fish in Lake Class 5 and 6.
Fishing for walleye, northern pike and to some degree smallmouth bass and yellow perch
increased the propensity to fish in inland lakes classified by Lake Class 5. Marginal effects for
the remaining coldwater (trout and salmon) and warmwater species (panfish and largemouth
bass) where negative. Similarly, although the marginal effects for Lake Class 6 were all
relatively small, fishing for coolwater species (northern pike, yellow perch and walleye)
increased the likelihood of fishing in these lakes. The marginal effects for the remaining
coldwater and warmwater species were either negative or not significant.

Summary and Conclusions
Although previous classifications of inland lakes are ecologically meaningful, they do not
directly account for angler behavior. Recreational angling is an important activity on inland
lakes, and anglers have significant impacts on the resource. The goal of this research was to
develop managerially-relevant segments of anglers associated with the biological lake
classifications developed in Wehrly et al (2012). We identified several important angler
characteristics related to fishing in particular lake classes. The next section includes a summary
description of the six inland lake classes including the characteristics of anglers fishing at each
class.

52

Lake Class 1
This lake class consists of small, deep, warm inland lakes, and resident anglers, anglers
possessing a restricted license, anglers making day trips and anglers targeting warmwater species
(largmouth bass and panfish) comprised a larger proportion of the clientele making trips to this
lake class. Controlling for other variables in the model, anglers with these characteristics also
had a higher propensity to fish in these lakes. The species that were the strongest predictors of
fishing in this lake class (Largemouth bass and panfish species) were also prominent indicator
species in the earlier lake classification by Wehrly et al (2012).
Lake Class 2
This lake class consists of large, deep, warm inland lakes and anglers making day trips,
trips focused one a single waterbody, trips where fishing was more likely to be the main purpose
of the trip, and anglers targeting a cross section of warmwater, coolwater and coldwater species
comprised a larger proportion of the clientele making trips to this lake class. Controlling for
other variables in the model, anglers with these characteristics as well as nonresident anglers and
anglers possessing all species licenses had a higher propensity to fish in these lakes. The topranking indicator species from the previous lake classifications by Wehrly et al. (2012) are
bowfin, longnose gar, common carp and channel catfish. On the whole, these species are
generally not considered highly desirable sportfish species. However, despite their lower
abundance and frequency of occurrence in this lake class, largemouth bass, smallmouth bass and
especially salmon were important predictors of fishing in this lake class. Results from our model
suggest Lake Class 2 supports an inland lake coldwater fishery for salmon—anglers targeting
inland lake salmon where much more likely to fish in this lake class. Finally, despite the

53

occurrence of panfish species as an ecological indicator, our model suggests fishing for panfish
did not increase the likelihood of fishing in these lakes.
Lake Class 3
Inland lakes in this category are large, are deep, are among the coldest, and are
predominantly located in the western upper peninsula and northern Michigan. Nonresidents are
somewhat more likely to fish in these lakes, as are anglers holding all species licenses, anglers
using a boat, anglers that are ice fishing, anglers making overnight trips, anglers fishing in more
than one waterbody and anglers targeting smallmouth bass, yellow perch as well as trout.
Walleye were a prominent indicator species in earlier ecological classifications (Wehrly et al.
2012), and they comprised a larger proportion of the species targeted by anglers, and, controlling
for other variables in the model, they are an important predictor of fishing trips to this lake class.
Lake Class 4
Although similar in some ecological aspects to Lake Class 1, the anglers differed since
anglers possessing “all species” licenses, anglers making overnight trips, anglers visiting
multiple fishing sites, and anglers combing their fishing trips with other non-fishing activities
have a higher propensity of fishing in these small somewhat abundant northern inland lakes. Our
model suggests fishing for northern pike modestly increases the likelihood of fishing at these
lakes, however the marginal effects for trout were considerably larger. Largemouth bass, yellow
perch and panfish species were important ecological indicators in these lakes. However results
of our descriptive analysis and model showed that these species comprised a relatively small to
modest proportion of species targets and were also not strong predictors of fishing in this lake
class. Further our descriptive analysis and results from our model suggests this lake class
supports a non-negligible inland lake cold water fishery for trout.

54

Lake Class 5
This lake class consists of rare, intermediately-sized, northern, coolwater lakes and
resident anglers, female anglers, anglers using a boat, anglers that are ice fishing, anglers making
overnight trips, anglers making trips in which fishing was not the main purpose, and anglers
targeting coolwater species are more likely to fish these lakes. Although a relatively large
proportion of anglers fishing in these lakes possessed restricted licenses, this was not a
significant variable in our model which controls for the effects of other variables. The species
targeted in these lakes are somewhat inconsistent with the dominant indicator sportfish found in
these lakes. Northern pike and yellow perch are both dominant indicator species (Wehrly 2012)
and strong predictors of fishing in this lake class. However the remaining coolwater species (i.e.
walleye and smallmouth bass) which were also strong predictors when targeted by anglers, are
not prominent ecological indicators in these lakes. Finally, despite their importance as
ecological indicator species, the marginal effects of targeting warmwater species were negative
for Lake Class 5.
Lake Class 6
While inland lakes in this class are somewhat similar to Lake Class 5, in terms of their
rarity, size, and northern distribution, nonresident anglers, female anglers, anglers holding an all
species license, anglers making overnight trips, and anglers making trips to multiple waterbodies
are more likely to fish in these lakes. Walleye, yellow perch and northern pike were both
prominent indicator species (Wehrly et al. 2012) and when targeted by anglers were important
predictor of fishing in these lake classes.

55

A key challenge to human dimensions researchers has been to identify and quantify the
different angler segments to help managers estimate the relative demand for different types of
opportunities (Fisher 1997). While effective at describing the diversity and size of particular
angler subgroups, human dimensions researchers have noted problems with applying angler
profiles based on specialization, desired recreational experiences and preferred site conditions
(Connelly et al. 2013). Fully integrated management perspectives include collecting and
understanding the interdependence of biological and social data in order to make informed and
justifiable management decisions (Hunt and Grado 2010, Johnston 2010).
Our integration of angler use provides a richer description of existing lake classifications
and can also be used for prioritizing management efforts including stocking, facility
development, and communication. For example, where multiple management options are
available to achieve the same biological outcome, managers may design rules and regulations
inclusive of the motives and preferences of angler subgroups that dominate the angler
constituency at particular waterbodies (e.g. Hutt and Bettoli 2007; Hunt and Grado 2010 pp430).
Although the influence of lake characteristics on angler behavior may seem intuitive (i.e.
the effect of lake size, lake remoteness, and species profiles), our results show that the
relationship is not completely congruent. In particular, our model suggests that angler species
targets do not always mirror the ecological profiles of species found in some lakes. Additionally,
our research is novel in meeting the call for development of managerially relevant angler
classifications based on inland lake fishery resources (Connelly et al. 2013), and it does so in a
manner that builds upon a tradition of ecological classification research in fisheries and natural
resources more generally.

56

CHAPTER 3

Management preferences of Michigan’s Great Lakes Anglers

Motivation
Covering 94,000 square miles and holding 5,500 cubic miles of water, the Great Lakes
contain roughly 21% of the world’s freshwater supply and are the largest freshwater system on
Earth (EPA 2014). The Great Lakes support shipping networks, provide recreational
opportunities and support tourism, agricultural and industrial sectors of the economy that are
dependent on these water resources. Due in part to their diverse geology, the Great Lakes
provide a rich diversity of habitats supporting commercial and recreational fisheries (Moll et al.
2013). Four Great Lakes and nearly 3,200 miles of coastline define Michigan’s border.
Fisheries managers, researchers and stakeholders tasked with developing and
implementing management plans are often faced with conflicts among management objectives to
support this large and diverse system. For example, following the reduction and extirpation of
native salmonids in the Great Lakes and due in part to extensive stocking programs and
naturalization, Pacific salmonine have effectively replaced native top-predator salmonids in
much of the Great Lakes. The proliferation of these species in the Great Lakes accompanied
important economic, biological and ecological consequences. In the wake of what was an
unprecedented ecosystem-level biomanipulation of a freshwater system, Great Lakes managers
are now faced with management conundrum Claramunt et al. (2013). The authors highlight
conflicting and often polarizing management strategies related to Pacific salmonines and native
species rehabilitation, a discussion intensified in the wake of Lake Michigan’s Chinook salmon

57

crash in the late 1980’s. Central to the differences in management perspectives is determining
which species (or species mix) will achieve a more balanced, stable, resilient ecosystem, limiting
the risk of future fish stock collapse.
Policy decisions and management strategies are further complicated by a lack of Great
Lakes angler preference information (Thayer and Loftus 2013). Although “public expectations
and desires factor prominently” in the goals and objectives of Michigan Department of Natural
Resources Fish Community Objectives (circa 1995 and 2003), researchers acknowledge the need
for reassessment and adjustment (Claramunt 2013). Our objective was to inform the
development of fishery management plans in four Great Lakes (Erie, Huron, Michigan and
Superior) and Lake St. Clair, using a stated-preference choice model to examine anglers’
preferred management outcomes. In our model, an angler’s preferred management outcome is
explained by the various attributes embodied in the management outcome alternative. We use
the model to estimate relative preferences, willingness to make tradeoffs between attributes, and
to illustrate likely angler support for Great Lakes management strategies differentiated by
emphasis on Pacific salmon, prey base, and risk of ecosystem collapse.
Choice Experiment
Following initial development of statistical models employing a behavioral theory
component (random utility theory) (McFadden 1974), choice experiments have been widely used
by marketing researchers to measure consumer preferences for different levels of attributes in a
given product (Louviere et al. 2000) and later within resource and environmental economics
where no markets exist (Hanley et al. 1998). Hunt (2005) includes a review of published
literature applying choice models to recreational fishing (and more specifically fishing site
choice), that began in the 1980’s.

58

In a choice experiment, participants are asked to choose their preferred alternative from a
set of two or more alternatives comprised of different levels of attributes. The stated preference
choice experiment format measures consumer preference for different levels of attributes in a
given alternative (Louviere and Woodworth 1983, Louviere et al. 2000), can be designed to
minimize unwanted correlation between attribute levels (Hunt 2005, Hanley et al. 1998) and
offers the ability to examine hypothetical goods that may not currently exist Adamowicz et al.
1994).
Participant choices reveal tradeoffs between the levels of attributes presented in each of
the alternatives. Attributes and attribute levels in choice experiments vary according to
researcher’s design and therefore fall somewhere within in the spectrum of “laboratory
experiments” and non-experimental observational studies (Harrison and List 2004).
Hypothetical management outcomes for stated preference choice models may be constructed
using a suite of relevant attributes and attribute levels to facilitate the estimation of all model
parameters, permitting researchers to estimate the effects of attributes independently from each
other (Hanely et al. 1998, Adamowicz et al. 1998).
Hunt et al. (2010) implemented a discrete choice stated-preference choice model that
solicited Canadian respondent’s most preferred outcome among the status quo and two
competing hypothetical outcomes for Lake Huron. Hunt et al. (2010) employed a choice model
with species abundance levels for 5 species, fish size, risk of fish stock collapse and prey fish
community. Each attribute contained three levels informed by consultations with fisheries
managers and survey testing. Results indicated that anglers generally preferred outcomes that
decreased risk of fish stock collapse and increased native prey fish communities, average fish
size and species abundances. To further illustrate management implications, Hunt et al. (2010)

59

also estimated Ontarian’s support for two contrasting management outcomes- specifically one
outcome emphasizing Chinook salmon abundance and another emphasizing lake trout and
walleye abundance.
We implement a stated-preference choice models to examine Michigan angler’s preferred
outcomes from management activities in four Great Lakes (Erie, Huron, Michigan and Superior)
and Lake St. Clair, and use attributed and a choice format that adapted and replicated for
Michigan anglers the stated preference choice model Hunt et al. (2010) implemented with
Ontarian anglers for Lake Huron. Figure 1 shows an example of a choice used in our study.

60

Figure 1: Image of example Great Lakes Angler Survey Choice Experiment.
If these were the only three possible outcomes from managing Lake Superior, which would you
prefer?
Please indicate below by checking your most preferred outcome (A, B, or C) for a future of Lake Superior.

61

Random Utility Theory
The choice model /statistical analysis in our choice experiment is motivated by a
respondent’s underlying utility. Random utility models have been widely used in situations
where a single product is chosen from a finite set of alternatives and the alternatives can be
characterized by a set of attributes. Random utility theory applies utility maximization theory to
choices among discrete alternatives to estimate preferences based on choices made by
respondents (McFadden 1974). Here the choices are gathered through a choice experiment.
The underlying assumption of a choice experiment is that when confronted with sets of
alternatives composed of several attributes, respondents will choose the alternative that leaves
them best off, maximizing their wellbeing or utility (utility maximization). All of the relevant
product attributes (i.e. attributes of the alternatives in the choice experiment) are known to the
consumer when the choice is made. Randomness enters the model because not all of the relevant
attributes are measured by the researcher.
Within the random utility model approach, angler i’s utility for alternative j is given by
. Faced with J alternatives, if the angler ranks a particular alternative, k, as most preferred,
then

is assumed to be the maximum among the J utilities. Let

be a variable indicating the

best alternative for individual i. The probability that alternative k is best is given by:

(

)

(

)

(1)

Because utility is latent researchers specify utility in two components, one that is
deterministic (observable) and another that is stochastic (unobserved). The deterministic
component is the Preference for the vector of attributes which describes the product (i.e.

62

management outcome alternative) measured by the researcher, denoted,
component represents a random error term, denoted

. The stochastic

. Since individual errors cannot be

observed from the perspective of researchers, each alternative has a probability of being chosen.
In our case, the ith angler is faced with J management outcome alternatives. Following
McFadden (1974) and Adamowicz (1998) the utility of alternative j,

, can be represented as:

(2)

The joint density of the random vector

) is denoted as ( ). With this density,

(

the researcher can make probabilistic statements about the decision maker’s choice. Different
discrete choice models are obtained from different specifications of this density, in other words
different distribution for the unmeasured portion of utility. If the error terms, , are independent
and identically distributed and follow a standard type I extreme-value distribution, a conditional
logit model can be used to estimate the probability of a respondent’s choice (McFadden 1974;
Louviere et al. 2000, Alberini 2007). The probability that k is best is:

(

)

(
∑

)
(

(3)

)

By presenting anglers with alternative management scenarios consisting of different
characteristics, x, we can use the choice probabilities to estimate the preference parameters

by

maximum likelihood. In our case, respondents were presented a choice scenario and were asked
to choose between the status quo or “same as today” and two alternative management outcomes
with attributes of risk of fish stock collapse, prey fish community, species abundance levels for
63

fish species and average fish size. Three levels were used to describe each attribute (see Table
9).

64

Table 9 Attributes and attribute levels for states preference choice model
Attribute Name
Risk of fish stock collapse (RISK)
Prey fish community (PREY)
Number of Fish
Channel Catfish
Largemouth Bass
Panfish
Smallmouth Bass
Walleye
White Bass
White Perch
Yellow Perch
Average Size of Fish (SIZE)

3
3

Model Attribute
Levels
0,1
0,1

slightly increased risk ; slightly decreased risk ; same as today
mostly introduced prey; mostly native prey; same as today

3
3
3
3
3
3
3
3
3

-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0
-0.5, 0.5, 0

50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% less; 50% more; same as today
50% smaller; 50% larger; same as today

Levels

65

Corresponding Survey Attribute

Carrying forward with the Lake Erie example and following the attribute levels used in the
choice experiment (and described above) equation (4) shows the model we are estimating:
(4)

where

is an alternative specific constant for the status quo,

on various levels for each alternative within a choice set,

is a vector of attributes that take

is a vector of unknown preference

parameters associated with those attributes, and is an error term. All coefficients in the model
are estimated relative to the base case “same as today”.
Marginal Rate Substitution (MRS)
Once the model is estimated, the rate of tradeoff between two attributes is the ratio of
their respective

coefficients, in economics these tradeoffs are referred to as the Marginal Rate

Substitution. Holding all other attributes constant, the MRS is equal to the change in one
attribute

required to compensate the individual for a one unit change in another attribute
66

,

(the amount of
change in

required to keep the individual at the same level of utility before the one unit

). Equation (5) and (6) illustrate how to calculate MRS as the total derivative of

utility with respect to

and set this derivative to zero so that utility doesn’t change, as

and

per Train (2009):
(5)
Solving for a change in

that keeps utility constant when

.

changes gives
(6)

The negative sign indicates that the two changes are in the opposite direction: assuming both
attributes are desirable, to keep utility constant,

must rise when

decreases. Examining

these ratios provides further insight into the relative importance anglers place on attributes at
each of the five lakes and the trade-offs anglers make between different levels of attributes,
holding all of the other attributes constant (all else equal).
In our results that follow, respondents generally preferred management outcomes that
both increased average fish size and the abundance of several sport fish species. We calculate
MRS for abundance of a species (e.g. an X1) relative to an increase or decrease in overall
average fish size (X2). Assuming a change in the abundance level of a particular species, the
MRS estimates change in overall average fish size necessary to keep anglers at the same level of
utility. For example, an MRS of 2.5 for a change in walleye abundance from “same as today” to
50% more, means that all else equal, management outcomes with 50% more walleye, would be
equally preferred to management outcomes with walleye levels the “same as today” and overall
average fish size 2.5 levels larger. For each lake, MRS are calculated relative to ther average
size of fish and are presented in Appendix J.

67

Data
Survey Development: Pretests
The mixed mode survey was conducted using both mail and internet questionnaires. The
questions were developed using an iterative process guided by the results of focus group and
one-on-one cognitive interviews with 54 individuals (Kaplowitz et al. 2004). Initial testing was
conducted in a focus group setting with nine licensed Michigan Great Lakes anglers. A
subsequent version of the survey, which included modified versions of the stated-preference
choice question implemented by Hunt et al. (2010), was tested in one-on-one in-person cognitive
interviews with 20 Great Lakes anglers. Finally, following methods reported by Weicksel
(2012), 25 cognitive interviews were conducted remotely with resident Great Lakes and nonGreat Lakes anglers recruited from a web survey panel. The latter pretesting approach more
closely resembled the actual experience actual survey respondents would face, thus providing
important insights and opportunities for viewing the choice selection process. Pretests were
concluded once it was clear that respondents comprehended survey tasks and that survey
information was clearly communicated to respondents.
Survey Sample, Survey Strategy, Response Rate
Our sample frame was the MDNR Retail Sales System database. We randomly drew
3,095 licensed Michigan anglers from a list of resident anglers, 18 years and older. Following a
modified Tailored Design Method (Dillman 2007), four contacts were made. For each of the
first three contacts, anglers were provided with the web-based survey address and a unique code
to access the web survey. A final paper survey package was mailed to remaining nonrespondents. After accounting for undeliverable addresses the final response rate was 36%
(Appendix F).

68

In addition to the choice experiment, both the web and paper questionnaires included
sections with included behavioral and demographic questions and additional questions related to
opinions, preferences and awareness. Details of the survey strategy and examples of survey
materials are provided in Appendix F.
The choice experiment section of the questionnaires began with a separate introductory
page containing an explanation of the format and purpose of the proceeding choice experiment
section. The page also contained an explanation of each of the attributes (i.e. “Glossary”).
Following the opening explanation and glossary, respondents were shown five different choice
sets (one for each lake: Lake Huron, Lake Erie, Lake Superior, Lake Michigan, Lake St. Clair.).
Respondents were asked to choose their preferred management outcome from two possible
outcomes and a baseline, or “same as today” condition. Each outcome was described using
eleven attributes that provided information about risk of fish stock collapse, prey fish
community, abundance of eight fish species and the overall average size of fish species. The
format closely follows Hunt et al. (2010). The suite of eight fish species presented in each
outcome scenario were not uniform across all lakes for each lake and included species with the
highest reported creel effort as well as species of management interest (personal communication,
MDNR).
Experimental Design
The choice experiment approach is essentially a structured method of data generation,
relying on careful design of choice tasks and definition of attribute space (Hanley et al 1998).
The set of attributes and levels displayed in Table 11 (above) can be viewed as setting the space
to be spanned in the choice experiment (Adamowicz 1994, pg 276). Our experimental design
was generated using Ngene software (Choice Metrics 2011). Our chosen orthogonal design, as

69

generated by Ngene, allows for isolating the separate effects of individual attributes on choice.
The experimental design consisted of 40 distinct choice sets for each of the five lakes. We used
a random number generator to assign choice sets. Each angler viewed one of the 40 distinct
choice sets, describing management outcomes for each of the five lakes..

Results and Discussion
A comparison of respondents and non-respondent characteristics revealed statistically
significant differences in age and license (there were no significant differences in gender).
Specifically, the mean age of respondents was significantly higher than nonrespondents
(p<0.001) and a smaller proportion of resident restricted, military and temporary (24-hour and
72-hour licenses) responded to our survey, with the remaining commonly issued license types
(Senior Restricted, Resident All Species, Senior All Species comprising a slightly larger
proportion of our respondents (X2 =126.057, df=4, p <0.001). To correct for possible
response/nonresponse bias, post-stratification weights for respondents were computed according
to the joint distribution of age and license type following Holt and Smith (1979) (Appendix G).
All results we report on use the generated weights.
Management preferences of Michigan’s Anglers
Respondents were asked to choose their preferred management outcome from two
hypothetical outcomes and a baseline, or “same as today” condition for four Great Lakes (Erie,
Huron, Michigan and Superior) and Lake St. Clair. Each outcome was described using eleven
attributes that provided information about risk of fish stock collapse, prey fish community,
abundance of eight fish species and the overall average size of fish species. All coefficients in
the model are estimated relative to the baseline or “same as today” attribute (Table 10).
70

The number of individuals completing the choice experiment question for each lake
ranged from 925 to 951. Across all lakes statistically significant parameter estimates followed
the expected direction, with positive parameter estimates for ecological health attributes
(decreasing risks of fish stock collapse, a more native prey fish community), and recreational
opportunities (increased average fish size and greater abundance of most fish species). Relative
to other species, increased abundance of walleye (and to a lesser extent yellow perch and salmon
species) most affected choice for management options. However the number of significant
parameter estimates and their relative magnitude varied across models for each of the five lakes.
We interpret these parameter estimates and estimate tradeoffs below.
Relative to other species, increased walleye abundance was particularly important to
angler management choices for Lake St. Clair, Lake Michigan and Lake Erie. In Lake Huron,
where there was relatively less distinction among species preferences, angler preferences for
increased abundance of Chinook salmon were somewhat greater than walleye, followed closely
by northern pike. For Lake Superior, where walleye were not part of the choice set, parameter
estimates for all eight species were statistically significant, with increased abundance of Chinook
and Coho salmon most preferred. Across lakes, parameter estimates for lake trout were not
significant, with the exception of Lake Superior, where anglers placed much less emphasis on
increased abundances of lake trout relative to other species.
In addition to interpreting relative preferences for management outcomes, we examined
tradeoffs between attributes implied by the parameter estimates. Although increased average
fish size was an important component of management outcomes for anglers across all lakes,
increasing the abundance of many native species (most notably walleye) was relatively more
important. For those systems, tradeoff calculations demonstrate that all else equal, anglers would

71

require some magnitude of an increase in overall fish size in order to be indifferent toward a
decreased abundance of those native species. For three of the lakes (Lake Huron, Lake Michigan
and Lake Superior), nonnative salmonid species were included in the choice set. Only in Lake
Huron were nonnative salmonid species (i.e. Chinook salmon and rainbow trout) relatively more
important than average fish size.

72

Table 10 Parameter estimates for conditional logit model
Lake Erie
observations
2802
Log likelihood
-983.62511
Pseudo R2
0.0566
Attribute
Status quo
0.1287
Risk of fish stock collapse (RISK)
slightly decreased risk
0.2404*
slightly increased risk
-0.0709
Prey fish community
(PREY)
mostly native prey
0.0615
mostly introduced prey -0.2579**
Number of Fish
Brown Trout
Channel Catfish
-0.1843
Chinook Salmon
Coho Salmon
Lake Herring
Lake Trout
Lake Whitefish
Largemouth Bass
0.3384**
Muskellunge/Muskie
Northern Pike
Panfish
0.3866**
Rainbow Trout
Smallmouth Bass
0.7077***
Splake
Sturgeon
Walleye
0.7895***
White Bass
0.1180
White Perch
0.2662**
Yellow Perch
0.3034**
Average Size of Fish
0.3113**
(SIZE)
* p < .10; **p < .05; ***p < .001

2787
-974.47639
0.0631

Lake
Michigan
2853
-1012.4024
0.0397

Lake St
Clair
2775
-961.80954
0.0695

Lake
Superior
2790
-987.67065
0.0501

0.3841**

0.0310

0.8492**

0.3848**

0.44013**
0.0175

0.0662
-0.1035

0.3489**
0.0226

0.0429
-0.2947**

0.4925**
0.1362

0.0223
-0.0953

0.2957**
0.2746**

0.2092
-0.0307

Lake Huron

0.3853**
0.0451
0.6359**
0.1444
0.1083

0.3012**
0.2932**

0.4473**
0.4226**
0.3022**
0.2368**
0.2774*

0.1487
0.3240**
0.2169*

0.5975***
0.2645**
0.5034***
0.3013**

0.1864
0.1746

0.2934**
0.2030
0.2624**

0.6064***

0.6307***

-0.0261
0.8524***

0.4759***

0.4491***

0.4304**

0.3983**

0.4425**

0.4767**

0.3614**

0.4848**

73

Predicted choice probabilities are calculated using the parameter estimates from our
model, and the relevant attribute levels (characteristics of the alternative). The predicted choice
probabilities for each scenario can be taken as a measure of the average respondent’s preference
for the alternatives and in that sense predicted support for each of the three alternatives. To
further illustrate managerial implications of the results, we developed scenarios to explore the
average angler support for the status quo or “same as today” option as compared with two
polarizing management scenarios differentiated by their emphasis on Pacific salmon versus
native sportfish species. In addition to species abundances, the characteristics of the alternatives
describe possible ecological conditions likely to accompany these management scenarios, with
prey base, risk of ecosystem collapse and average fish size taking on the relevant attribute levels
(e.g. +1,0,-1). Specifically, for the salmon focus, prey base is set to “mostly introduced,” risk of
fish stock collapse is set to “slightly increased,” fish size is set to 50% smaller, salmon
abundances are raised by 50% and other species are reduced by 50%. For the native species
focus, prey base is set to “mostly native,” risk of fish stock collapse is set to “slightly decreased,”
fish size is set to 50% larger, salmon abundances are reduced by 50% and other species are
raised by 50%. For the status quo, variables are set to their mid-point, or same as today levels.
See (Appendix H) for exact lake-specific management alternatives including attribute levels and
choice probability calculations. Because they each include a Pacific salmon component, we
focus our scenario analyses on to Lake Huron, Lake Michigan and Lake Superior. In general,
choice probabilities, interpreted as predicted support, were greater for outcomes with a native
species emphasis, as compared to a Pacific salmon emphasis and the status quo (i.e. “same as
today option”). Although the general scenarios for each lake are similar, the level of predicted
support for each of the three lakes varied because the estimated preference parameters and exact

74

species vary across lakes. It is important to note that these results apply to the average angler
represented by our study and are for the hypothetical management scenarios that were examined,
Subsets of anglers are likely to have substantially different preferences, and alternative
assumptions about the attribute values that underlie the scenarios (e.g., changes in average fish
size) would lead to different levels of predicted support for each scenario.

Table 11 Predicted probabilities
Predicted percentage of anglers choosing management alternatives focused on:
Pacific Salmon, native species or the "same as today".*
Pacific Salmon Focus

Native Species Focus

Same as Today

Lake Huron

7%

73%

20%

Lake Michigan

18%

49%

33%

Lake Superior

11%

57%

32%

* For the Pacific Salmon Focus, prey base is set to mostly introduced, risk of fish
stock collapse is set to slightly increased, fish size is set to 50% smaller, salmon
abundances are raised by 50% and other species are reduced by 50%. For the
Native Species Focus, prey base is set to mostly native, risk of fish stock collapse
is set to slightly decreased, fish size is set to 50% larger, salmon abundances are
reduced by 50% and other species are raised by 50%. For the Same as Today or
status quo, variables are set to their mid-point, or same as today levels.

Lake Erie
In the following sections, we review the results for each lake. For Lake Erie, anglers
preferred management outcomes that favored ecological health attributes, and that also increased
average fish size. In terms of species abundances, walleye had the largest effect on outcome
choice, followed by smallmouth bass, panfish, largemouth bass, yellow perch, and white perch.
In Lake Erie, relative preferences for walleye and smallmouth bass far exceeded preferences for
all other species. The magnitude of the relative importance of walleye and smallmouth bass was
75

unique to the Lake Erie system; in the remaining lakes, species preferences were either not as
distinct, or were clearly dominated by walleye alone. Trade-off calculations further underscored
the importance of walleye and smallmouth bass in particular to management outcomes in Lake
Erie. Holding other variables constant (all else equal), anglers were willing to make relatively
greater trade-offs in the average size of fish in Lake Erie to maintain or increase walleye and
smallmouth bass abundance.
Given the potential for more restrictive walleye and yellow perch regulations (Lake Erie
Committee 2013), new management strategies, regulatory changes or outreach activities that
both enhance and promote the Lake Erie smallmouth bass fishery could serve to offset the lower
abundance of walleye and regulation restrictions. Additionally, enhancement of the white perch
fishery, or promotion of more preferred species (e.g. largemouth bass and panfish) may also
serve to offset the more restrictive yellow perch regulations.
Lake Huron
For Lake Huron, our findings in part mirrored the angler preferences reported for
Canadian waters of Lake Huron by Hunt et al (2010), despite the differences in some of the
species attributes and the sample frame between the two studies. In our model results for Lake
Huron, Chinook salmon and walleye abundance also had a large effect on choice probabilities
and respondents also expressed preferences for decreasing risks of fish stock collapse and more
native prey fish communities. However, species preferences were somewhat uniform.
Parameter estimates for Chinook salmon were closely followed by walleye and northern pike.
Trade-off calculations suggest anglers were willing to tradeoff average fish size to maintain or
increase the abundance of five species: Chinook salmon, walleye, northern pike, rainbow trout
and to a lesser degree, yellow perch.

76

The comparable parameter estimates for Chinook salmon, walleye and northern pike in
our model suggest that these species are viewed as closer substitutes for one another in Lake
Huron. In contrast to other lake systems, fisheries managers may consider a broader suite of
management strategies that increased the abundance of a combination of these three species. For
both yellow perch and smallmouth bass recent evidence of strong recruitment and significant
increase in abundance (Riley 2013) may explain relatively small effect on management outcome
choice probabilities.
To illustrate managerial implications of the results, we estimated choice probabilities for
competing hypothetical management strategies, one focused on increasing the abundance of
Chinook salmon and rainbow trout and another focused on increasing abundance of the
remaining native species. Given large and significant parameter estimates for native species,
Table 11 shows that nearly three quarters of anglers were predicted to support a native species
emphasis outcome (see Appendix H for the exact computations).
Lake Michigan
For Lake Michigan, parameter estimates for the status quo and ecological considerations
were not significant. Estimates were significant for four species, and were strongest walleye,
followed by yellow perch, brown trout, Chinook salmon and Coho salmon. The relative
importance of increasing overall fish size for Lake Michigan is notable. With the exception of
walleye, anglers placed more emphasis on average fish size than increases in fish species
abundance. Preferences for yellow perch were second only to walleye; therefore tradeoff
calculations for yellow perch mirrored the tradeoff calculations for increased fish size.
Management outcomes that increased yellow perch and decreased abundances of Coho salmon,
Chinook salmon, brown trout would be equally preferred to management outcomes that

77

maintained these species abundances. To illustrate managerial implications of the results, we
estimated choice probabilities for competing hypothetical management strategies, one focused on
increasing the abundance of Chinook salmon, Coho salmon and rainbow trout and another
focused on increasing abundance of the remaining native species (see Table 11). Compared to
Lake Huron and Lake Superior, the predicted support was less clear for Lake Michigan, with just
under 50% of anglers predicted to support the native species emphasis outcome, and nearly onethird preferring the status quo.
Lake St. Clair
Anglers indicated preferences for the status quo, decreased risk of fish stock collapse, and
increased average fish size for Lake St Clair. Preferences for prey fish community were mixed.
Preferences were greatest for walleye, followed by yellow perch, largemouth bass, panfish and
muskellunge. Increases in fish abundance for two species (walleye and yellow perch) were more
preferable than fish size. For the remaining species (channel catfish, smallmouth bass and
sturgeon) angler’s preferences were not significantly different than zero.
Lake Superior
Anglers indicated preferences for the status quo, decreased risk of fish stock collapse and
larger fish size for Lake Superior. Angler’s preferences for prey fish community outcomes were
not significantly different from zero. Preferences were greatest for Chinook salmon, followed by
Coho Salmon, yellow perch, lake herring, rainbow trout, lake whitefish, splake and lake trout.
Anglers placed greater emphasis on increased average fish size than on increases in abundance of
any of the fish species. For Lake Superior 56% of anglers were predicted to support the native
species emphasis outcome for the management scenarios examined in Table 11, with just under
one-third choosing the status quo.

78

Comparing Preferences of Great Lakes and Non Great Lakes Anglers
All of the results and interpretations presented so far are for models that use all of the
survey respondents and capture the average preferences for those respondents. However, not all
anglers actually fish in Great Lakes. Slightly less than half of respondents (47%) reported
fishing in the Great Lakes and connecting waterways in the last 12 months. We tested whether
interaction with our Great Lakes fishing variable (Great Lakes and connecting waterways fishing
in last 12 months) provided any significant improvement to account for preference heterogeneity
between anglers that fished the Great Lakes in the past year and those that did not. Results of
log likelihood ratio tests failed to reject the null hypothesis of equal preferences, suggesting
preferences for management outcomes were not significantly different between groups
(Appendix I).
However, from a managerial perspective, we felt it was important to examine the
outcome preferences for Lake Huron and Lake Michigan, controlling for anglers who indicated
they focused their Great Lakes fishing on these lakes in the last 12 months. These had the two
largest shares of our sampled anglers that fished them, and thus provided adequate model sample
sizes to separately estimate preferences for anglers that fished these lakes. These lakes also are
of special management interest. Separate models controlling for Lake Huron and Lake Michigan
anglers who indicated the mostly fished at these lakes are presented in (Appendix I). In sum,
these models showed that anglers that recently fished at Lake Huron placed more weight on
Chinook and walleye than the average respondent, and that anglers that recently fished at Lake
Michigan placed more weight on yellow perch, and Coho salmon, and brown trout, and less
weight on walleye than the average respondent.

79

Conclusions and Management Implications
The goal of this paper was to investigate the preferences that Michigan anglers have for
fisheries management outcomes in four Great Lakes (Lake Erie, Lake Huron, Lake Michigan,
Lake Superior) and Lake St. Clair. In general, anglers were supportive of management outcomes
that reduced fish stock collapse, provided a prey fish community with more native species, and
increased the abundance and size of most sport fish species. Management outcomes that
increased the abundance of walleye, yellow perch and salmon species were also more preferred.
An important outcome of this study is the ability to estimate the tradeoffs anglers make
among different attributes. Across all lakes, tradeoff calculations demonstrate angler preferences
for increased abundance of native sport fish species. For these systems, anglers have stronger
preferences for increasing the abundance of these native species, than for management outcomes
that produce larger average fish size. In Lake Huron, preferences for Chinook salmon and
rainbow trout were accompanied by strong preferences for walleye and northern pike.
Estimated choice probabilities for hypothetical management scenarios with a native
species focus v. pacific salmon focus were estimated to illustrate the managerial implications of
the results. Given large and significant parameter estimates for native species in Lake Huron,
nearly three quarters of anglers were predicted to support a native species emphasis outcome.
For Lake Michigan, the predicted support was less clear, with just under 50% of anglers
predicted to support the native species emphasis outcome, and nearly one-third choosing the
status quo. For Lake Superior 56% of anglers were predicted to support the native species
emphasis outcome, with just under one-third choosing the status quo.

80

The choice experiment we employed provides a mechanism for future inquiries regarding
management preferences. The choice experiment format could be implemented in future angler
inquiries to elicit angler preferences toward the complex nature of biological management,
including uncertain outcomes, competing management outcomes or as part of a standardized
protocol to measure Great Lakes angler preferences (evaluate trends, gauge support for
management changes, etc.).
Limitations of this study include the hypothetical nature of the management scenarios
examined and the pooled or average preferences presented above. The choice scenarios
presented to anglers were effective for revealing tradeoffs between levels of attributes presented
in each of the alternatives. However, when we illustrated the predicted support for the
hypothetical management scenarios we made assumptions about the attributes to enter into the
scenario, but these attribute levels may not be biological feasible for some or all of the Great
Lake systems in this study. Further, we highlight that we did find evidence of relevant
preference heterogeneity within Lake Huron and Lake Michigan and suspect similar differences
in preferences exist between subgroups of anglers differentiated by fishing frequency and
perhaps by other characteristics (e.g. proximity to the lake, avidity, commitment). Thus, the
preferences of management-relevant subgroups of anglers are likely to reveal different
preferences than those reported here for the average angler in our study.

81

APPENDICES

82

APPENDIX A
Michigan Recreational Angler Survey (MRAS)
Mailing Procedures
Monthly samples for the MRAS originate from the Michigan Department of Natural
Resources, Retail Sales System database. The sample population is Licensed Anglers (all license
types, age 18+). The simple random sample is conducted monthly.
Once the sample has been received from the MDNR, the “Customer ID” is used to crosscheck the current sample against all previously sampled anglers. Removing repeats should
ensure that the survey is not sent to the same individual more than once. After repeats are
removed, the remaining list is examined and entries missing key information (e.g., name of
licensee, address of licensee) are removed. Cases containing extraneous information are edited
for correctness, where possible, or removed. Licensees with addresses outside the U.S. or
Canada are also removed from the sample. Once the final sample is established, each licensee in
the sample is assigned a unique project code (Angler ID) which is used for the duration of the
project.
An outside vendor (currently ASAP Printing, Okemos, MI) is contracted to conduct all
outgoing mailing procedures. The outside vendor conducts a National Change of Address
(NCOA) check of licensees, mail merge, prints and packages all components of the mailings and
applies the appropriate postage to all mailings. The completed packages are then pre-sorted and
mailed First-Class through the United States Postal Service. Each business reply envelope
contains the unique project code (Angler ID) assigned to each individual in our sample. After

83

conducting the NCOA, but prior to printing, packaging and mailing materials, the vendor
contacts the researcher for a proof approval.

Mailing Content and General Timeline
The procedures for the mail survey were adapted from on the methods in Mail and
Internet Surveys (Dillman 2007). The first mailing includes: 1) a cover letter printed on MSU
letterhead explaining the importance and objectives of the survey; 2) the four page survey
instrument; and, 3) a self-addressed, postage paid, business reply envelope. The first mailing is
followed approximately 5 days later by a reminder / thank you postcard. Approximately 30 days
after the first mailing, non-respondents receive a third mailing containing: 1) a second cover
letter printed on MSU letterhead which again explains the importance and objectives of the
survey with a slightly more urgent appeal; 2) a replacement survey instrument; and, 3) a selfaddressed, postage paid, business reply envelope. Approximately 14 days after the third mailing,
non-respondents received a fourth and final reminder / thank you postcard with a slightly more
urgent appeal. The survey instrument itself varies monthly only in Part B, fishing activities
during the most recent month. The researcher updates the calendar to reflect the most recently
completed month for the survey round.

Data Collection and Management Procedures
Returned business reply envelopes and undeliverable mailings are grouped by the week
of their receipt. All returned mailings are coded to inform follow-up mailings to nonrespondents. Additional communications from respondents directed to the principal investigator
including short notes, letters, pictures and refusals are coded and filed appropriately. Initially,
returned business reply envelopes were opened and the project code appearing on the business

84

reply envelope is written on the top of the corresponding survey. Current procedures include a
preprinted Angler ID on the survey. Some small percentage of surveys may be unidentifiable
because they are returned in non-project envelopes, or the survey code is illegible (i.e. crossed
out). All of the returned surveys and additional communications are stored in locked offices on
the Michigan State University campus.

Sampling History
Simple random sample of 5,000 individuals who purchased fishing licenses between
April 1, 2008 and July 31, 2008 was drawn on August 27, 2008. A second random sample of
6,000 individuals who purchased fishing licenses between April 1, 2008 and August 31, 2008
was drawn on September 8, 2008. Each sample included all possible license types. The
sampling continued as above, with modifications only to the monthly sample amount. Simple
random samples for the following periods were as follows:




September 2008 through November 2009: 2,500 individuals
December 2009 through June 2010: 1,250 individuals
July 2010 through February 2013: 500 individuals

Survey Instrument
They survey instrument was designed to capture angling seasonality, cover all fishing
activity within Michigan, incur low costs per respondent and collect both short term and longer
term trip information balancing the benefits of long recall periods with the needs for accuracy.
Longer recall (activities over the last 12 months) are limited to general angler behavior
questions: fishing in Michigan (yes/no); participation in fishing events (yes/no); fishing in other
countries or states besides Michigan (yes/no); types of waterbodies fished; number of trips
taken (range); and fish species sought and methods used.

85

To meet our objectives, the survey instrument contains 21 multi-part questions over four major
sections: Part A) general fishing activities during the last 12 months; Part B) fishing activities
during the most recent month (e.g., June); Part C) details of the recent fishing trip(s); Part D)
usual fishing activities, background and demographic information (see Figure A1).

86

Figure A.1: Image of Michigan Recreational Angler Survey.

87

Figure A.1 (Cont’d)

88

Figure A.1 (Cont’d)

89

Figure A.1 (Cont’d)

90

Figure A.2: Michigan Recreational Angler Survey Contact letter
DATE
<Name>
<Street>
<City>, <State> <ZIP>
Your help is needed with a study of fishing in Michigan. The study is being conducted by
Michigan State University’s Department of Fisheries and Wildlife for the Michigan Department
of Natural Resources, Fisheries Division. The results from this survey will help natural resource
agencies make fisheries management decisions that better reflect the needs of people that fish in
Michigan.
You are part of a small sample of people being asked about their fishing activities. Your answers
are needed to help ensure the results accurately represent the people who fish in Michigan.
Whether you go fishing often or only occasionally, your input is important. Please let us know
what you think by completing the enclosed questionnaire and returning it in the prepaid
envelope.
Your individual views will be completely confidential and your privacy will be protected to the
maximum extent permitted by law. Also, your participation in the survey is voluntary, and you
may refuse to answer certain questions. If you have any concerns or questions about this
research study, such as scientific issues, how to do any part of it, or if you believe you have been
harmed because of the research, please contact Frank Lupi, Department of Fisheries and
Wildlife, Michigan State University, 13 Natural Resources Building, East Lansing, MI 488241222; lupi@msu.edu, 517-432-3883.
If you have any questions or concerns about your role and rights as a research participant, or
would like to register a complaint about this research study, you may contact, anonymously if
you wish, Michigan State University Human Research Protection Program at 517-355-2180,
FAX 517-432-4503, or e-mail irb@msu.edu, or regular mail at: 202 Olds Hall, Michigan State
University, East Lansing, MI 48824.
Thank you very much for helping with this important study.
Sincerely,

Frank Lupi
Associate Professor
Enclosure

91

Figure A.2: (Cont’d)

DATE
<Name>
<Street>
<City>, <State> <ZIP>
I recently sent you a survey about your fishing activities in Michigan. To the best of my
knowledge, I have not heard from you.
I am writing to you again because your input is vital! You are part of a small sample of people
who are being asked about their fishing activities.
Your answers are needed to help ensure the results accurately represent the people who fish in
Michigan. Your answers will help natural resource agencies make management decisions that
better reflect the needs of people that fish in Michigan.
Please take a few minutes to share your viewpoint by filling out this short survey.
Your individual views will be completely confidential and your privacy will be protected to the
maximum extent permitted by law. Also, your participation in the survey is voluntary, and you
may refuse to answer certain questions. If you have any concerns or questions about this
research study, such as scientific issues, how to do any part of it, or if you believe you have been
harmed because of the research, please contact Frank Lupi, Department of Fisheries and
Wildlife, Michigan State University, 13 Natural Resources Building, East Lansing, MI 488241222; lupi@msu.edu, 517-432-3883.
If you have any questions or concerns about your role and rights as a research participant, or
would like to register a complaint about this research study, you may contact, anonymously if
you wish, Michigan State University Human Research Protection Program at 517-355-2180,
FAX 517-432-4503, or e-mail irb@msu.edu, or regular mail at: 202 Olds Hall, Michigan State
University, East Lansing, MI 48824.
Thank you very much for helping with this important study.
Sincerely,

Frank Lupi
Associate Professor
Enclosure

92

APPENDIX B
Chapter 1 Response Rate

Table B 1 Michigan Recreational Angler Survey Response Rate
Adjusted response rate [19,635 / 45,504-(1851+2524+3)]= 47.7%
Number
Cases Eliminated
Undeliverable
Refusals
Deceased
Surveys Returned
Surveys Not Returned
Total

93

Percent
1851

4.1%

2524
14
3
19635
21477
45504

5.5%
0.0%
0.0%
43.2%
47.2%
100%

APPENDIX C
Chapter 1 Respondent and Non-Respondent characteristics and Data Weights
Of the 45,504 individuals who were contacted from our simple random sample from
March 2009 through February 2012, 19,635 returned surveys. After accounting for eliminated
cases and undeliverable addresses, the final adjusted response rate was 47.7%. A comparison of
respondent and non-respondent characteristics revealed statistically significant differences in
age, residency, gender, and license type. The mean age of respondents was significantly higher
than nonrespondents (p<0.001), Michigan residents (X2 =58.762, df=1, p <0.001) and females
(X2 =4.542, df=1, p =0.033) were somewhat more likely to respond and a smaller proportion of
resident restricted, temporary (24-hour and 72-hour licenses) license holders responded to our
survey, with the remaining commonly issued license types (Resident All Species, Non-Resident
All Species, Senior Restricted, and Senior All Species comprising a slightly larger proportion of
our respondents (X2 =1441.437, df=7, p <0.001) (Table C.1).
These results are suggestive of age and license type differences in respondents to the
survey. To correct for possible response/nonresponse bias, weights for the survey respondents
are computed according to the distribution of age, gender, residency and license type. For the
continuous variable age and the multiple-category variable license type, the distribution of the
variables was examined and categories were imposed to create relatively even distributions. Post
stratification case weights are arrived at by normalizing the percent in sample and percent of
respondents for jointly distributed categories (Holt and Smith 1979). For example, the
percentage of resident, males, age 18-24, with restricted licenses in our sample (2.4%) was
divided by the percent of respondents for that joint distribution (1.69%) to arrive at a case weight
of 1.4484. Case weights were applied to respective jointly distributed categories. There are 6
age ranges (18-24, 25-34, 35-44, 45-54, 55-64, 65-94), two license categories (Restricted and All
Species), two gender categories, and two resident categories (MI resident, NonMichigan
Resident) (See Table C.2 through C.5).

94

Table C 1 MDNR sample, respondents and nonrespondent characteristics.
MDNR Sample

Respondents

NonRespondents

Michigan Residents

84.0%

85.5%

82.9%

Non-Residents

16.0%

14.5%

17.1%

Females

18.1%

18.5%

17.8%

Males

81.9%

81.5%

82.2%

Mean Age

47.14

51.69

43.658

Standard Deviation (Age)

16.221

15.699

15.743

Resident Restricted

40.0%

34.1%

44.6%

Resident All Species

29.5%

32.9%

26.9%

Non-Resident Restricted

4.9%

4.9%

4.9%

Non-Resident All Species

2.6%

3.0%

2.3%

Senior Restricted

5.2%

6.8%

4.0%

Senior All Species

6.9%

10.2%

4.3%

Temporary (24hr & 72hr)

10.4%

7.8%

12.3%

Other Licenses

0.5%

0.3%

0.7%

statistic

df

p

58.762c

1

<0.001

4.542c

1

0.033

1.442f

<0.001

License Type*

1441.437

7

<
0.001

A total of 45,504 anglers were contacted from the 2009 through 2012 fishing season.
Subscript c denotes chi-square. Subscript f denotes ANOVA F statistic.

95

Table C 2 Post-stratification weights: Age and License Distributions
(males residents).
Age and License Type Distribution for Male Residents from the MDNR Retails Sales System
sample
License
Age 18Age 25Age 35Age 45Age 55Age 65Total
Category
24
34
44
54
64
94
Restricted
5.7%
8.4%
9.4%
11.0%
8.9%
6.5%
49.9%
All Species
3.9%
7.4%
8.8%
10.9%
9.5%
9.7%
50.2%

Age and License Type Distribution for Male Resident Survey Respondents
License
Age 18Age 25Age 35Age 45Age 55Category
24
34
44
54
64
Restricted
2.9%
5.0%
6.7%
9.7%
10.5%
All Species
2.6%
6.1%
8.2%
12.7%
13.2%

Table C.4. Age and License Type Sample Weights for Male Residents
License
Age 18- Age 25- Age 35- Age 45- Age 55Category
24
34
44
54
64
Restricted
1.9838
1.6846
1.3997
1.1330
0.8526
All Species
1.4736
1.2077
1.0740
0.8591
0.7200

96

Age 6594
8.2%
14.3%

Age 6594
0.7931
0.6802

Total
42.8%
57.2%

Table C 3 Post-stratification weights: Age and License Distributions
(female residents)
Age and License Type Distribution for female residents from the MDNR Retails Sales
System sample
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
8.8%
12.8%
12.2%
16.7%
12.8%
8.4% 71.7%
All Species
3.2%
4.7%
4.9%
6.9%
4.8%
3.8% 28.3%

Age and License Type Distribution for female resident Survey Respondents
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
5.5%
8.1%
10.0%
17.7%
16.5%
11.4% 69.2%
All Species
1.9%
3.9%
4.4%
8.5%
6.9%
5.3% 30.8%

Age and License Type Sample Weights for female residents
License
Age 18Age 25Age 35Age 45Category
24
34
44
54
Restricted
1.591337 1.580754 1.22828 0.94229
All Species
1.637317 1.220612 1.119879 0.807801

97

Age 5564
0.777402
0.694536

Age 6594
0.742324
0.713903

Table C 4 Post-stratification weights: Age and License Distributions
(male nonresidents)
Age and License Type Distribution for male nonresidents from the MDNR Retails Sales
System sample
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
2.2%
4.1%
5.5%
6.1%
6.5%
5.2% 29.6%
All Species
5.8%
12.9%
13.6%
15.9%
13.1%
9.0% 70.3%

Age and License Type Distribution for male nonresident Survey Respondents
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
1.3%
2.5%
4.3%
6.8%
9.4%
8.4% 32.7%
All Species
3.0%
6.3%
11.2%
16.6%
17.5%
12.8% 67.4%

Age and License Type Sample Weights for male nonresidents
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Restricted
1.6693 1.6289 1.2741 0.9095 0.6958 0.6218
All Species
1.9677 2.0336 1.2189 0.9579 0.7501 0.7060

98

Table C 5 Post-stratification weights: Age and License Distributions
(female nonresidents)
Age and License Type Distribution for female Nonresidents from the MDNR Retails
Sales System sample
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
2.4%
5.7%
5.8%
9.8%
7.9%
5.8% 37.4%
All Species
7.9%
14.2%
12.2%
14.4%
9.6%
4.4% 62.7%

Age and License Type Distribution for female NonResident Survey Respondents
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
1.7%
3.6%
4.7%
12.5%
10.2%
9.3% 41.9%
All Species
3.2%
11.0%
10.8%
14.8%
12.9%
5.5% 58.1%

Age and License Type Sample Weights for female NonResidents
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Restricted
1.4484 1.5823 1.2415 0.7856 0.7759 0.6207
All Species
2.4829 1.2892 1.1279 0.9754 0.7462 0.7958

99

APPENDIX D
Chapter 1 Esri Tapestry segmentation strategy evaluation.
Examination of LifeMode segments and general fishing behaviors (D.1), Michigan-specific fishing behaviors (D.2 through D.4) and
license purchase behavior (D.5). Chi-square tests of independence, Cramer’s V, Goodman and Kruskal Tau, One-Way ANOVA F
statistic, Welch’s adjusted F ratio, and eta-squared (η²) and an adjusted omega square ( ).
Table D 1 Comparison of LifeMode general fishing behavior characteristics

1

High Society

40.60

58.4%

Catch
Disposition
Mostly
Keepc
22.3%

93.3%

36.6%

2

Upscale Avenues

39.94

63.6%

28.1%

35.3%

93.4%

30.0%

3

Metropolis

40.33

38.2%

38.1%

32.5%

92.8%

27.4%

4

Solo Acts

36.66

43.3%

20.7%

48.0%

89.3%

35.7%

5

Senior Styles

42.00

59.1%

32.3%

31.6%

90.5%

23.7%

6

Scholars and Patriots

31.20

42.0%

25.6%

41.0%

91.0%

23.1%

7

High Hopes

35.20

50.4%

26.0%

38.9%

89.0%

27.1%

8

Global Roots

35.30

47.2%

25.3%

30.2%

91.5%

18.6%

LifeMode

a

Description

Fishing
Experiencea
(Mean)

Boat
Ownershipb

Catch
Disposition
Mostly
Catch&Releasec
45.9%

Companiond

Fishing
Outside
Statee

9

Family Portrait

37.20

52.8%

22.6%

42.7%

92.7%

32.9%

10

Traditional Living

39.26

51.7%

27.7%

35.2%

92.4%

25.3%

11

Factories and Farms

40.28

66.0%

37.8%

23.3%

92.8%

22.0%

12

American Quilt

42.04

67.8%

36.0%

23.9%

90.3%

22.2%

Average

40.23

60.2%

30.6%

32.8%

92.1%

26.8%

Lavene = (df1=11,df2=18,127),7.492, p<.001; Welch's F=(df1=11,df2=1825.797)= 19.878, p <.001; ω²= 0.0111; η²= 0.0117; Games-Howell 39/66, α =(.05)

b

X2=2008.108, df = 11, p < .001; Tau = .020, p <.001; Cramer's V = .141, p <.001, N=18,688

c

X2=573.750, df=22, p < .001; Tau = .015, p <.001; Cramer's V = .124, p <.001, N=18,528

d

X2=40.488, df=11, p <.001; Tau = .002, p <.001; Cramer's V = .050, p <.001, N=16510

e

X2=248.509, df =11, p <.001; Tau = .013, p <.001; Cramer's V = .115, p <.001, N=18,689

100

Table D 2 Comparison of LifeMode Michigan-specific fishing behavior characteristics
Life
Mode

a

Description

Fishing
Frequecy a
1 to 3
times

Fishing
Frequecya
10 or more
times

Species
Diversityb

Method
Diversityc

Total Species
and Method
Diversityd

Michigan
Fishing
Experiencee

Waterbody
Diversityf

Great
Lakesg

Inland
Lakesh

Riversi

1

High Society

27.1%

44.5%

3.11

3.61

8.51

30.69

1.70

49.0%

76.8%

43.8%

2

Upscale Avenues

21.4%

51.8%

3.31

3.79

9.20

32.40

1.72

49.4%

79.1%

43.7%

3

Metropolis

21.7%

51.8%

3.39

3.18

8.40

29.57

1.70

43.5%

63.4%

63.1%

4

Solo Acts

30.3%

43.2%

3.28

3.37

8.37

27.06

1.71

49.8%

70.6%

50.5%

5

Senior Styles

19.3%

53.8%

3.42

3.71

9.38

34.38

1.76

52.3%

74.4%

49.3%

6

Scholars&Patriots

23.0%

54.9%

3.59

3.85

10.12

21.03

1.88

46.1%

80.9%

60.9%

7

High Hopes

20.2%

50.8%

3.53

3.64

9.29

27.53

1.81

44.4%

78.2%

58.2%

8

Global Roots

22.6%

52.8%

3.69

3.54

9.75

25.81

1.73

34.4%

78.1%

60.0%

9

Family Portrait

27.0%

47.3%

3.20

3.62

8.84

28.15

1.67

46.6%

78.2%

42.4%

10

Traditional Living

20.0%

53.4%

3.49

3.78

9.81

31.48

1.77

48.8%

78.6%

49.6%

11

Factories&Farms

18.0%

55.5%

3.41

3.86

9.55

34.20

1.76

44.1%

79.3%

53.0%

12

American Quilt

15.9%

60.6%

3.63

4.02

10.40

36.08

1.83

49.7%

82.3%

51.3%

TOTAL

20.6%

53.0%

3.39

3.77

9.41

32.66

1.75

48.3%

78.1%

49.0%

2

X =271.461, df=55, p <.001; Tau = .004, p <.05; Cramer's V = .055, p <.001, N=18,012

b

Lavene = (df1=11,df2=17,483),6.983, p<.001; Welch's F=(df1=11,df2=1777.895)= 12.701, p <.001; ω²= 0.0070; η²= 0.0076; Games-Howell 14/66, α =(.05)

c

F=(df1=11,df2=17493)= 14.391, p <.001; ω²=0.0083; η²= 0.0090; Scheffe 13/66, α =(.05)

d

Lavene = (df1=11,df2=17,482),6.792, p<.001; Welch's F=(df1=11,df2=1782.877)= 10.505, p <.001; ω²= 0.0060; η²= 0.0066; Games-Howell 13/66, α =(.05)

e
f

Lavene = (df1=11,df2=17,983),3.881, p<.001; Welch's F=(df1=11,df2=1817.258)= 34.572, p <.001; ω²= 0.0195; η²=0.0201; Games-Howell 47/66, α =(.05)

F=(df1=11,df2=18080)= 5.614, p <.001; ω²=0.0028; η²= 0.0034; Scheffe 3/66, α =(.05)

g

X2=64.079, df=11 p <.001; Tau = .004, p <.05; Cramer's V = .060, p <.001, N=18,081

h

X2=109.202, df=11, p <.001; Tau = .006, p <.05; Cramer's V = .078, p <.001, N=18,084

i

X2=158.101, df=11, p <.001; Tau = .009, p <.05; Cramer's V = .094, p <.001, N=18,084

101

Table D 3 Comparison of LifeMode and species fished for in last 12 months
Life
Mode

Description

Bassa

Catfishb

Nortern
Pikec

1
2
3
4
5
6
7
8
9
10
11
12

Panfishd Salmone

High Society
74.1%
12.7%
41.4%
64.7%
31.0%
Upscale Avenues
72.1%
17.3%
44.9%
73.7%
32.4%
Metropolis
80.6%
40.5%
35.5%
62.8%
23.8%
Solo Acts
71.2%
21.2%
40.4%
60.3%
39.7%
Senior Styles
69.9%
18.4%
47.0%
70.8%
30.5%
Scholars and Patriots
73.9%
11.7%
50.5%
73.0%
35.1%
High Hopes
76.7%
29.4%
46.7%
72.4%
33.2%
Global Roots
80.1%
37.2%
44.9%
71.2%
29.5%
Family Portrait
74.3%
18.5%
43.4%
68.4%
27.0%
Traditional Living
76.0%
26.6%
46.0%
73.7%
28.0%
Factories and Farms
70.3%
23.4%
44.7%
75.9%
29.9%
American Quilt
70.8%
18.7%
53.3%
76.3%
33.5%
72.3%
20.2%
45.8%
72.2%
30.9%
TOTAL
a 2
X =55.382, df=11, p <0.001; Tau = .003, p <0.001; Cramer's V = .056, p <0.001, N=17,493
b 2
X =307.492, df=11, p <0.001; Tau = .018, p <0.001; Cramer's V = .133, p <0.001, N=17,493
c 2
X =97.464, df=11 p <0.001; Tau = .006, p <0.001; Cramer's V = .075, p <0.001, N=17,495
d 2
X =153.450, df=11, p <0.001; Tau = .009, p <0.001; Cramer's V = .094, p <0.001, N=17,495
e 2
X =46.787, df=11, p <0.001; Tau = .003, p <0.001; Cramer's V = .052, p <0.001, N=17,496
f 2
X =221.102, df=11, p <0.001; Tau = .013, p <0.001; Cramer's V = .112, p <0.001, N=17,492
g 2
X =102.700, df=11, p <0.001; Tau = .006, p <0.001; Cramer's V = .077, p <0.001, N=17,494
h 2
X =144.096, df=11, p <0.001; Tau = .008, p <0.001; Cramer's V = .091, p <0.001, N=17,497

102

Suckersf
4.5%
8.5%
14.1%
9.3%
9.9%
16.2%
10.3%
20.5%
7.6%
11.4%
14.5%
14.3%
10.8%

Troutg Walleyeh
34.1%
31.6%
29.0%
46.0%
36.0%
51.4%
36.1%
32.1%
31.5%
30.6%
30.2%
38.1%
33.5%

47.8%
49.9%
50.7%
39.4%
58.9%
43.2%
47.5%
53.2%
47.9%
56.1%
51.3%
57.9%
52.7%

Table D 4 Comparison of LifeMode and bait types used in the last 12 months
Life
Mode
1
2
3
4
5
6
7
8
9
10
11
12

Description
High Society
Upscale Avenues
Metropolis
Solo Acts
Senior Styles
Scholars and Patriots
High Hopes
Global Roots
Family Portrait
Traditional Living
Factories and Farms
American Quilt
TOTAL

Natural
Baita

Artificial Baitb

Trollingc

Casting From
Boatd

Casting
From Shoree

Fly
Fishingf

Ice
Fishingg

75.2%
82.4%
82.4%
66.9%
83.7%
82.9%
81.7%
85.3%
80.8%
86.1%
86.0%
87.0%
83.1%

76.2%
77.4%
63.9%
67.5%
73.8%
75.7%
75.4%
73.1%
76.0%
75.4%
75.4%
76.7%
75.5%

50.5%
51.3%
32.0%
47.4%
49.0%
42.3%
43.4%
35.9%
42.9%
45.9%
49.4%
55.5%
49.3%

65.8%
67.7%
49.6%
60.3%
62.2%
64.0%
63.0%
64.1%
67.2%
64.7%
67.0%
67.9%
65.6%

49.4%
52.6%
59.2%
50.0%
53.4%
67.6%
59.0%
58.3%
53.4%
58.3%
53.7%
54.9%
53.9%

21.5%
16.8%
14.1%
24.8%
15.6%
22.5%
18.0%
13.5%
16.0%
15.7%
15.8%
16.5%
16.9%

20.8%
29.2%
17.0%
18.5%
30.8%
28.8%
21.2%
19.9%
23.4%
29.1%
35.7%
40.5%
30.3%

a 2

X =215.506, df=11, p <0.001; Tau = .012, p <0.001; Cramer's V = .111, p <0.001, N=17,495

b 2

X =48.328, df=11,p <0.001; Tau = .003, p <0.001; Cramer's V = .053, p <0.001, N=17,495

c 2

X =126.157,df=11, p <0.001; Tau = .007, p <0.001; Cramer's V = .085, p <0.001, N=17,496

d 2

X =73.887, df=11, p <0.001; Tau = .004, p <0.001; Cramer's V = .065, p <0.001, N=17,495

e 2

X =54.102, df=11, p <0.001; Tau = .003, p <0.001; Cramer's V = .056, p <0.001, N=17,490

f 2

X =59.659, df=11, p <0.001; Tau = .003, p <0.001; Cramer's V = .058, p <0.001, N=17,495

g 2

X =351.569, df=11, p <0.001; Tau = .020, p <0.001; Cramer's V = .142, p <0.001, N=17,494

103

Table D 5 Comparison of LifeMode and license purchasing behavior

1

High Society

12.5%

Vendor
Marjor
Grocery /
Departmenta
31.4%

2

Upscale Avenues

7.1%

35.2%

3

Metropolis

4.6%

36.4%

LifeMode

4
5
6
7

Description

Vendor
DNRa

Vendor
Marjor
Sporting
Goodsa
15.0%

41.2%

Days
(since
season
start)b
90.44

12.1%

45.6%

84.13

6.43

3.29

82.8%

11.2%

47.8%

76.91

5.55

2.93

69.5%

2.59

88.1%
76.4%

Vendor
Othera

Fidelity
20012012 c

Fidelity
20082012d

5.89

3.05

90.1%

Computer
Usage

Solo Acts

9.6%

37.5%

13.4%

39.5%

90.75

4.54

Senior Styles
Scholars and
Patriots

4.5%

32.8%

9.5%

53.2%

78.87

6.59

3.33

7.7%

46.9%

12.8%

32.5%

90.22

4.38

2.66

High Hopes

5.5%

44.4%

9.4%

40.8%

87.24

5.24

2.91

83.0%

2.69

71.4%

90.8%

Global Roots

5.1%

43.3%

9.1%

42.5%

85.70

4.74

9

Family Portrait

8.2%

38.3%

11.6%

41.9%

87.94

5.81

3.06

85.0%

10

Traditional Living
Factories and
Farms

4.3%

38.9%

11.8%

45.0%

81.86

6.10

3.13

76.7%

3.6%

33.3%

5.3%

57.7%

76.33

6.84

3.42

American Quilt

3.8%

29.6%

6.2%

60.4%

75.89

7.03

3.49

8

11
12

70.6%
70.5%

6.35
3.24
5.8%
34.4%
9.8%
50.0%
81.45
78.0%
TOTAL
X =1804.836, df=33,p <.001; Tau = .014, p <.001; Cramer's V = .113, p <.001, N=47,266
b
Lavene = (df1=11,df2=44,832),7.738, p<.001; Welch's F=(df1=11,df2=6340.987)= 40.905, p <.001; η²= 0.0100; ω²= 0.0098; GamesHowell 40/66, α =(.05)
c
Lavene = (df1=11,df2=47,254),16.403, p<.001; Welch's F=(df1=11,df2=6687.155)= 106.097, p <.001; η²=0.0224; ω²= 0.0222; GamesHowell 56/66, α =(.05)
d
Lavene = (df1=11,df2=47,254),13.737, p<.001; Welch's F=(df1=11,df2=6659.125)= 79.317, p <.001; η²=0.0182; ω²= 0.0179; GamesHowell 49/66, α =(.05)
a 2
X =516.492, df=33,p <.001; Tau = .028, p <.001; Cramer's V = .167, p <.001, N=18,542
a 2

104

APPENDIX E
Inland Lake Classification Process and Response Rate

Inland lake trip locations where identified using some combination of the water body
name and nearest city and county. Identified inland lakes were coded with a unique
identification number developed by the Michigan Department of Natural Resources, Institute
for Fisheries research. Identified lakes were then matched to the lake classification system
developed by Wehrly et al. (2012). A total of 2,125 unique inland lake identification
numbers were generated. Of those, 1,886 were assigned a lake classification number (1-6)
resulting in 14,889 individual trips with an inland lake classification code reported by 13,053
anglers. General Survey Response is total number of responses to the survey. Many
respondents did not report trips because they (1) did not fish in last 12 months or (2) did not
complete the section (i.e. item nonresponse).

Table E 1 Inland Lake Classifications

Michigan Recreational Angler Survey. Survey dates begining July 2008 through
February 2012. Raw data, adjusted response rate, identification results and
number and percent of identified trips. General Survey Response is response
to the survey. Adjusted response rate (Response / ((Total Sample (Undeliverables + Deceased)).

Sample
General Survey Response
Adjusted Survey Response
Number of anglers reporting at least one inland lake trip
Unique Inland Lakes
Unique Inland Lakes classification number
Analyzed inland lake trips

105

number
70,273
31,046

percent

46%
13,053
2,125
1,886
14,889

89%

APPENDIX F
Chapter 3 Great Lakes Angler web-survey, paper survey, contact letters and postcards.
Following a modified Tailored Design Method (Dillman 2007), we mailed three contacts
over the course of one month (from November 7 to November 21, 2012): an invitation letter, a ¼
sheet reminder postcard (in color) and, a 1/2 sheet reminder postcard (in color). A fourth
contact, consisting of a final survey package with most of the original web survey questions, was
mailed 3.5 months later (March 28, 2013) to non-respondents. During all four contacts, anglers
were provided with the web-based survey address and a unique code to access the web survey.
In total, 459 individuals responded the survey through the internet, and 577 individuals
responded to the final mail survey. After accounting for undeliverable addresses the final
adjusted response rate was 36% (Table H.1).
Both the web survey and paper survey instruments included 4 main sections (A) a section
gathering general fishing behavior and fishing activities over the last 12 months, (B) a choice
experiment section, (C) a section soliciting levels of awareness, levels of agreement, importance
rankings, and preferences related to a variety of management programs, issues and challenges,
and (D) a section of background and demographic questions. The survey instrument mailed to
non-respondents to the web survey largely duplicated the original web survey instrument, with a
reduction primarily in part (C).

106

Table F 1 Great Lakes Angler survey response
Great Lakes Angler survey. Adjusted response rate [3095 /1036-256] = 36%
Number
256
459
577
1036
767
3095

Undeliverables
Web-based survey response
Mail survey response
Total Surveys Returned
Surveys Not Returned
Total

107

Percent
8.3%
14.8%
18.6%
33.5%
24.8%
100.0%

APPENDIX G
Chapter 3 Data Weights
The mean age of respondents was significantly higher than nonrespondents (p<0.001) and a
smaller proportion of resident restricted and temporary or 24-hour and 72-hour licenses
responded to our survey, with the remaining commonly issued license types (Senior Restricted,
Resident All Species, Senior All Species comprising a slightly larger proportion of our
respondents (X2 =142.18, df=7, p <0.001) (see Table I.1).

These results are suggestive of age and license type differences in respondents to the
survey. To correct for possible response/nonresponse bias, post-stratification weights for the
survey respondents are computed according to the distribution of age and license type. For the
continuous variable age and the multiple-category variable license type, the distribution of the
variables was examined and categories were imposed to create relatively even distributions. Post
stratification case weights are arrived at by normalizing the percent in sample and percent of
respondents for jointly distributed categories (Holt and Smith 1979). The base is our sample
frame the MDNR Retail Sales System database (RSS). There are 6 age ranges (18-24, 25-34,
35-44, 45-54, 55-64, 65-94) and two license categories (Restricted and All Species).

108

Table G 1 MDNR sample, respondents and nonrespondent characteristics

Female
Mean Age
Standard Deviation (Age)
License Typec
Resident Restricted
Senior Restricted
Resident All Species
Senior All Species
Military & Temporary
(24Hour&72Hour)

MDNR
Sample
20.7%
46.58
16.439

Respondents

Nonrespondents

statistic

df

p

20.6%
52.88
15.277

20.8%
43.39
16.085

250.506f

1

<0.001

50.5%
5.8%
32.0%
7.8%

40.2%
7.9%
36.1%
13.6%

55.6%
4.8%
30.0%
4.9%

127.445

4

<0.001

3.8%

2.2%

4.6%

N=3,095. Subscript c denotes chi-square. Superscript f denotes ANOVA F statistic
Table G 2 Post-stratification weights: Age and License type distributions.
Joint Age and License Type Distributions from the MDNR Retail Sales System sample
Age Category
License
Age 18- Age 25- Age 35- Age 45- Age 55- Age 65Category
24
34
44
54
64
94
Total
Restricted
8.14%
9.14%
10.18% 11.57% 11.02%
6.37%
56.41%
All Species
4.43%
6.91%
7.30%
8.69%
8.11%
8.17%
43.62%

Joint Age and License Type Distributions of Survey Respondents
License
Category
Restricted
All Species

Age 1824
3.67%
2.70%

Age 2534
4.54%
4.15%

Age Category
Age 35- Age 4544
54
7.14%
11.39%
5.98%
10.81%

Age 5564
12.93%
14.09%

Age 6594
8.78%
13.80%

Total
48.46%
51.54%

Age and License Type Sample weights
License
Category
Restricted
All Species

Age Category
Age 18Age 25Age 35Age 45Age 55Age 6524
34
44
54
64
94
2.21981 2.01552 1.42488 1.015547 0.851822 0.724643
1.637803 1.665883 1.220157 0.803958 0.575466 0.592221
109

APPENDIX H
Chapter 3 Predicted Choice Probabilities
Lake Huron, Lake Michigan and Lake Superior

Predicted Choice Probabilities
Lake Huron, Lake Michigan and Lake Superior
To further illustrate managerial implications of the results, we used the model to predict
support for Great Lakes management strategies differentiated by their emphasis on Pacific
salmon, prey base, risks of ecosystem collapse and average fish. Recall that the probability of
preferring option k is:
(

)

(
∑

(

)
)

where x are the observed variables relating to the alternative management strategies, β are the
parameters that were estimated. Tables J1 through J3 represent three management alternatives
for which choice probabilities are calculated: a Pacific Salmon focus with increased risk of fish
stock collapse, more nonnative prey, and smaller average size fish (column A), a native species
focus (column B) with decreased risk of fish stock collapse, more native prey, and larger average
size fish and the status quo or “same as today” (column C). For each lake, the tables show the
specific values assumed for each of the attributes under each of the three alternatives. In general,
choice probabilities, interpreted as predicted support, were greater for outcomes with a native
species emphasis, as compared to a Pacific salmon emphasis and the status quo or “same as
today option”.

110

Table H 1 Predicted Probabilities: calculations for Lake Huron

Status Quo
Risk of fish stock collapse (RISK)
slightly decreased risk
slightly increased risk
Prey fish community (PREY)
mostly native prey
mostly introduced prey
Number of Fish
Chinook
Lake Herring
Lake Trout
Northern Pike
Rainbow Trout
Smallmouth Bass
Walleye
Yellow Perch
Average Size of Fish (SIZE)
TOTALS (Σ B* X)
Exp of Σ β*X

Pacific Salmon Focus
A
X
β*X
0
0

Native Focus
B
X
β*X
0
0

X
1

Status Quo
C
β*X
0.3841

0
1

0
0.0175052

1
0

0.440075
0

0
0

0
0

0
1

0
0.1361696

1
0

0.492458
0

0
0

0
0

0.5
-0.5
-0.5
-0.5
0.5
-0.5
-0.5
-0.5
-0.5

0.31793765
-0.07221215
-0.05414565
-0.29875695
0.2517238
-0.15064415
-0.30319165
-0.23793475
-0.221272
-0.61482105
0.540737652

-0.5
0.5
0.5
0.5
-0.5
0.5
0.5
0.5
0.5

-0.31794
0.072212
0.054146
0.298757
-0.25172
0.150644
0.303192
0.237935
0.221272
1.701029
5.479582

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0.3841
1.468293

ExpA + ExpB+ ExpC

7.48861237

7.488612

7.488612

Choice probabilities

0.072207991

0.731722

0.19607

111

Table H 2 Predicted Probabilities: calculations for Lake Michigan

Status Quo
Risk of fish stock collapse (RISK)
slightly decreased risk
slightly increased risk
Prey fish community (PREY)
mostly native prey
mostly introduced prey
Number of Fish
Brown Trout
Chinook Salmon
Coho Salmon
Lake Trout
Rainbow Trout
Smallmouth Bass
Walleye
Yellow Perch
Average Size (SIZE)
TOTALS (Σ B* X)
Exp of Σ β*X

Pacific Salmon Focus
A
X
β*X
0
0

Native Focus
B
X
β*X
0
0

Status Quo
C
β*X
0.030972

X
1

0
1

0
-0.1034979

1
0

0.066193
0

0
0

0
0

0
1

0
-0.0953338

1
0

0.02227
0

0
0

0
0

0.5
0.5
0.5
-0.5
0.5
-0.5
-0.5
-0.5
-0.5

0.1926342
0.1505957
0.146603
-0.07434915
0.09317505
-0.08727995
-0.31534315
-0.22453205
-0.23836335
-0.5556914
0.573675485

-0.5
-0.5
-0.5
0.5
-0.5
0.5
0.5
0.5
0.5

-0.19263
-0.1506
-0.1466
0.074349
-0.09318
0.08728
0.315343
0.224532
0.238363
0.445323
1.560994

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0.030972
1.031456

ExpA + ExpB+ ExpC

3.166125645

3.166126

3.166126

Choice probabilities

0.181191636

0.49303

0.325779

112

Table H 3 Predicted Probabilities: calculations for Superior

Status Quo
Risk of fish stock collapse (RISK)
slightly decreased risk
slightly increased risk
Prey fish community (PREY)
mostly native prey
mostly introduced prey
Number of Fish
Chinook Salmon
Coho Salmon
Lake Herring
Lake Trout
Lake Whitefish
Rainbow Trout
Splake
Yellow Perch
Average Size (SIZE)
TOTALS (Σ B* X)
Exp of Σ β*X

Pacific Salmon Focus
A
X
β*X
0
0

Native Focus
B
X
β*X
0
0

Status Quo
C
X
β*X
1
0.384753

0
1

0
-0.2946737

1
0

0.042865
0

0
0

0
0

0
1

0
-0.03065

1
0

0.209188
0

0
0

0
0

0.5
0.5
-0.5
-0.5
-0.5
0.5
-0.5
-0.5
-0.5

0.2236385
0.2112997
-0.15105595
-0.11837985
-0.138666
0.1466966
-0.13118715
-0.19910945
-0.2423972
-0.7244845
0.484574303

-0.5
-0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5

-0.22364
-0.2113
0.151056
0.11838
0.138666
0.146697
0.131187
0.199109
0.242397
0.944607
2.571802

0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0.384753
1.469252

ExpA + ExpB+ ExpC

4.525628723

4.525629

4.525629

Choice probabilities

0.107073366

0.568275

0.324651

113

APPENDIX I
Great Lakes and Non-Great Lakes Anglers
Lake Huron and Lake Michigan Model Comparison
Great Lakes and Non-Great Lakes Anglers
Lake Huron and Lake Michigan Model Comparison
Slightly less than half of respondents (47%) reported fishing in the Great Lakes and
connecting waterways in the last 12 months. Compared to anglers who did not report fishing in
the Great Lakes, these Great Lakes Anglers were slightly less likely to be female, were more
likely to hold All Species licenses, fished more frequently, less likely to fish inland lakes and
more likely to fish in rivers. More Great Lakes Anglers reported owning a boat used for fishing,
and held membership in fishing organizations or clubs and attended fisheries-related public
meetings, citizen advisory committee meetings or other fisheries meetings. Great Lakes Anglers
targeted many species, but were nearly twice as likely to target trout and walleye and were three
times more likely to target salmon species. Great Lakes anglers also employed a greater
diversity of bait types and fishing techniques, in particular Great Lakes anglers were more likely
to report offshore boating, near shore boating and trolling techniques (Table K 1 and K2).
We tested whether interaction with our Great Lakes fishing variable (Great Lakes and
connecting waterways fishing in last 12 months) provided any significant improvement to
account for preference heterogeneity in the model. Results of log likelihood ratio tests failed to
reject the null hypothesis of equal preferences, suggesting preferences for management outcomes
were not significantly different between groups (Table K3). However, from a managerially
perspective, we felt it was important to examine the outcome preferences for Lake Huron and
Lake Michigan, controlling for anglers who focus their fishing on these lakes.

114

Separate models controlling for Lake Huron and Lake Michigan anglers who indicated
they mostly fished at these lakes are presented in Table K4 and K5. The results in Table K4 and
K5 highlight the different emphasis placed on increasing the average size and abundance of
Pacific salmon and native species. Anglers who predominately focused their fishing in Lake
Huron (see table K4) placed relatively greater importance on management outcomes that
increased average fish size as well as the numbers of Chinook salmon, northern pike and
especially walleye. Lake Huron anglers placed relatively less emphasis on smallmouth bass, and
for rainbow trout the parameter estimates were not statistically significant. In terms of
ecosystem variables, non-Lake Huron anglers placed more importance on decreasing risk of fish
stock collapse and more native prey fish, while Lake Huron anglers were indifferent. Although
broadly the relative species preferences of Lake Huron angler mirror our full model, the
magnitude of the parameter estimates for the most preferred species (i.e. walleye, Chinook
salmon and northern pike) stand in contrast to the relatively more uniform species preferences of
our full model. Lake Huron angler utility will be disproportionality negatively affected by
management outcomes that decrease Chinook salmon. However, given Lake Huron angler’s
strong preferences for management outcomes favoring northern pike and walleye and lower
preferences for rainbow trout the predicted probabilities over the illustrative management
scenarios are quite similar in spirit to those in Table J1 (our calculations demonstrate slightly
higher preferences for the “Native Focus” [73%] and slightly lower for “status quo” [20%]).
Anglers who predominately focused their fishing in Lake Michigan (see Table K5)
placed more emphasis on the status quo, increased average fish size, and greater abundance of
yellow perch, brown trout and Coho salmon. Lake Michigan anglers placed relatively less
emphasis on walleye. Although not statistically significant, preferences for Chinook salmon and

115

smallmouth bass were smaller. For all lakes, including Lake Michigan, we reported that
outcomes increasing numbers of walleye most affected the choice of management outcomes. For
Lake Michigan anglers, yellow perch, brown trout and Coho salmon dominate outcome
preferences. Lake Michigan anglers are somewhat more likely to support management outcomes
that favored these nonnative species (our calculations demonstrate somewhat higher preferences
for “Native Focus” [58%] and slightly lower for “status quo [30%]).

116

Table I 1 Comparison of Great Lakes and non-Great Lakes anglers
NonGreat
Lakes
Anglers
25.1%
46.84
16.502

Great
Lakes
Anglers

Total

Statistic

18.9%
46.45
16.154

22.2%
46.66
16.333

5.911

1

0.015

33.1%

33.5%

31.2%

60.7%
6.7%
21.8%
7.6%
3.1%

39.1%
4.7%
46.3%
8.2%
1.6%

50.6%
5.8%
33.3%
7.9%
2.4%

74.415c

4

<0.001

7.4%
20.1%
17.2%
12.7%
18.4%
24.2%

2.1%
6.2%
14.8%
11.3%
19.1%
46.6%

4.8%
13.3%
16.0%
12.0%
18.7%
35.1%

88.495c

5

<0.001

Michigan Inland Lakes

82.2%

70.4%

76.7%

19.935c

1

<0.001

Michigan Rivers

38.1%

53.7%

45.5%

25.425c

1

<0.001

3.6%
4.4%
19.9%
4.2%
1.1%

12.1%
22.7%
42.6%
11.5%
6.6%

7.6%
13.0%
30.6%
7.6%
3.7%

441.729c

5

<0.001

66.8%

4.5%

37.5%

36.26
25.563
54.4%

37.61
15.948
73.3%

36.91
21.468
63.6%

37.799c

1

<0.001

16.311

c

1

<0.001

24.693

c

1

<0.001

Female
Mean Age
Standard Deviation (Age)
Education (% Bachelor's or Graduate
Degree)
License Type
Resident Restricted
Senior Restricted
Resident All Species
Senior All Species
Temporary (24Hour&72Hour)
Fishing Frequency (past 12 months)
1 time
2 or 3 times
4 to 5 times
6 to 9 times
10 to 19 times
20 or more time
Waterbodies Fished (past 12 months)

Great Lakes Most Fished (past 12 months)
Lake Erie
Lake Huron
Lake Michigan
Lake St. Clair
Lake Superior
No Response or Did NOT fish in the Great
Lakes
Michigan Fishing Experience
Standard deviation (Experience)
Boat Ownership
Fishing Organization Member
Public Meeting Attendance
c

8.1%
2.9%

Chi square statistic; df degrees of freedom;

117

16.4%
10.9%

12.1%
6.8%

df Probability

Table I 2 Comparison of Great Lakes and non-Great Lakes anglers

Species Targeted (last 12 months)
Bass
Catfish
Panfish
Pike
Salmon
Suckers
Trout

NonGreat
Lakes
Anglers

Great
Lakes
Anglers

Total

76.9%
19.1%
72.0%
48.1%
17.3%

73.1%
23.8%
74.8%
52.4%
54.0%

75.1%
21.3%
73.3%
50.1%
34.6%

9.3%

15.6%

26.0%

47.0%

12.2%
35.9%

Statistic

df

Prob.

3.388c

1

0.066

153.702c

1

<0.001

1

0.002

1

<0.001

1

<0.001

9.493

c

49.336

c
c

Walleye
Methods Used (last 12 months)

38.9%

75.6%

56.1%

140.934

Natural Bait

78.1%

89.7%

83.6%

25.278c

1

< 0.001

76.8%

c

1

< 0.001

1

< 0.001

1

< 0.001

Artificial Bait
Trolling
Boat Near Shore
Boat Off Shore

68.7%

85.8%

32.2%

67.1%

60.3%

78.4%

40.4%

73.7%

42.360

48.6%

126.402

68.8%

c

56.1%

39.612

116.045

1

< 0.001

1

0.003

1

0.011

1

< 0.001

46.9%

46.7%

51.3%

9.046

Fly Fishing

12.4%

18.1%

15.1%

6.518c

c

23.3%

39.6%

Chi square statistic; df degrees of freedom;

118

31.0%

c

c

Shore Pier
Ice Fishing

c

32.125

c

Table I 3 Log likelihood ratio tests: Great Lakes and non-Great Lakes anglers
Log likelihood ratio test comparing the fit of our unconstrained model
(controlling for Great Lakes fishing in the last 12 months) to our
constrained model. Observed differences in model fit for each of the
five lakes were not statistically significant. df=13. Log likelihood ratio=
-2(L(constrained)-L(unconstrainted)). Data used calculated poststratification weights.
Constrained Unconstrained
Lake
Statistic
df
Model
model
Lake Erie
983.62511
972.22234
22.80554
13
Lake Huron
974.47639
965.33448
18.28382
13
Lake Michigan
1012.4024
1004.63313
15.53854
13
Lake St Clair
961.80954
953.01705
17.58498
13
Lake Superior
987.67065
983.15323
9.03484
13

119

Table I 4 Lake Huron model comparison.
Parameter estimates for conditional logit model for Lake Huron. Full model repeats the information in
Table 10. Non Lake Huron anglers (column 3) did not indicate they fished most in Lake Huron in the
past 12 months. Lake Huron anglers (column 4) indicated they fished most in Lake Huron in the past
12 months.

observations
Log likelihood
Pseudo R2
Attribute
Status quo
Risk of fish stock collapse (RISK)
slightly decreased risk
slightly increased risk
Prey fish community (PREY)
mostly native prey
mostly introduced prey
Number of Fish
Brown Trout
Channel Catfish
Chinook Salmon
Coho Salmon
Lake Herring
Lake Trout
Northern Pike
Rainbow Trout
Smallmouth Bass
Walleye
Yellow Perch
Average Size (SIZE)
* p < .10; **p < .05; ***p < .001

Lake Huron
Full Model
2787
-974.47639
0.0631

Lake Huron
Non Lake Huron Anglers
2409
-845.56626
0.0609

Lake Huron
Lake Huron Anglers
378
-116.87023
0.1635

0.3841**

0.4999***

0.2638

0.44013**
0.0175

0.4764***
-0.0033

0.3100
0.2963

0.4925**
0.1362

0.5889***
0.1649

-0.1325
-0.1148

0.6359**

0.5870***

0.8854**

0.1444
0.1083
0.5975***
0.5034***
0.3013**
0.6064***
0.4759***
0.4425**

0.1906
0.1485
0.5617***
0.5461***
0.3051**
0.5131***
0.4501***
.3940**

-0.1465
-0.1325
0.8517**
0.2637
0.2522**
1.2596***
0.5469
0.6607**

120

Table I 5 Lake Michigan model comparison.
Parameter estimates for conditional logit model for Lake Michigan Full model repeats the information
in Table 10. Non Lake Michigan anglers (column 3) did not indicate they fished most in Lake Michigan
in the past 12 months. Lake Michigan anglers (column 4) indicated they fished most in Lake Michigan
in the past 12 months.
Lake Michigan

2853
-1012.4024
0.0397

Lake Michigan
Non Lake Michigan
Anglers
1920
-685.95221
0.0416

Lake Michigan
Lake Michigan
Anglers
933
-316.67244
0.0639

0.0310

0.2178

-0.3788*

0.0662
-0.1035

0.0734
-0.0920

-0.0098
-0.0989

0.0223
-0.0953

0.0752
-0.0205

-0.0732
-0.2750

0.3853***

0.3174**

0.5706**

0.3012**
0.2932**

0.3330**
0.1959

0.2161
0.5187**

0.1487

0.0891

0.2662

0.1864
0.1746
0.6307***
0.4491***
0.4767***

0.1435
0.2576*
0.7158***
0.3346**
0.4761***

0.2992
0.0033
0.4481**
0.7841***
0.5347**

Full Model
observations
Log likelihood
Pseudo R2
Attribute
Status quo
Risk of fish stock collapse (RISK)
slightly decreased risk
slightly increased risk
Prey fish community (PREY)
mostly native prey
mostly introduced prey
Number of Fish
Brown Trout
Channel Catfish
Chinook Salmon
Coho Salmon
Lake Herring
Lake Trout
Northern Pike
Rainbow Trout
Smallmouth Bass
Walleye
Yellow Perch
Average Size (SIZE)
* p < .10; **p < .05; ***p < .001

121

APPENDIX J
Chapter 3: MRS Calculations
Table J 1 Marginal Rate Substitution calculations relative to Average Size
Marginal Rates Substitution calculation (calculations relative to Average Size; ∂X₁/∂X₂= -β₂/β₁)
Lake
Lake
Lake
Lake
Lake
Erie
Huron
Michigan
St Clair
Superior
Attribute
Status quo
-0.4134
-0.0650
-2.3495
-0.7936
-0.8679
Risk of fish stock collapse (RISK)
slightly decreased risk
-0.7724
-0.9944
-0.1388
-0.9654
-0.0884
slightly increased risk
0.2278
-0.0396
0.2171
-0.0626
0.6078
Prey fish community (PREY)
mostly native prey
-0.1976
-1.1128
-0.0467
-0.8182
-0.4315
mostly introduced prey
0.8285
-0.3077
0.2000
-0.7598
-0.0307
Number of Fish
Brown Trout
-0.8082
Channel Catfish
0.5920
-0.1248
Chinook Salmon
-1.4369
-0.6318
-0.9226
Coho Salmon
-0.6150
-0.8717
Lake Herring
-0.3264
-0.6232
Lake Trout
-0.2447
-0.3119
-0.4884
Lake Whitefish
-0.5721
Largemouth Bass
-1.0870
-0.8963
Muskellunge/Muskie
-0.6001
Northern Pike
-1.3502
Panfish
-1.2419
-0.7318
Rainbow Trout
-1.1376
-0.3909
-0.6052
Smallmouth Bass
-0.6808
-0.3662
-0.5616
-2.2734
Splake
-0.5412
Sturgeon
0.0722
Walleye
-1.3702
-1.3230
-2.3583
-2.5363
White Bass
-0.3791
White Perch
-0.8550
Yellow Perch
-0.9748
-1.0753
-0.9420
-1.1908
-0.8214
Average Size (SIZE)
-1.0000
-1.0000
-1.0000
-1.0000
-1.0000

122

REFERENCES

123

REFERENCES

Aas, O. & Arlinghaus, R.. (2009). New Markets for Recreational Fishing. In M. Manfredo, J.
Vaske, P.J. Brown, D.J. Decker, & E.A. Duke (Eds.). Wildlife and Society, The Science of
Human Dimensions (pp. 229-243). Island Press: Washington, DC.
Adamowicz, W., Louviere, J. & Williams, M. (1994). Combining revealed and stated preference
methods for valuing environmental amenities. Journal of Environmental Economics and
Management, 26, 271-292.
Agresti, A. (1979). Statistical methods for the social sciences. San Francisco: Dellen Publishing
Company
Alberini, A., Longo, A., & Veronesi, M. (2007). Basic Statistical Models for Stated Choice
Studies. In B.J. Kanninen (Ed.). Valuing Environmental Amenities Using Stated Choice Studies
A Common Sense Approach to Theory and Practice (pp. 203-227). Springer: AA Dordecht, The
Netherlands.
American Sportfishing Association (2007). American Sportfishing Association, Association of
Fish & Wildlife Agencies, U.S. Fish and Wildlife Service, Sport Fish Restoration. Lifestyles and
License Buying Habits of America’s Anglers. Accessed February 2014.
http://asafishing.org/facts-figures/studies-and-surveys/lifestyles-and-license-buying/
http://asafishing.org/uploads/Lifestyles_and_License_Buying_Habits.pdf
Anderson, D. K., Ditton, R. B., Hunt, K. M. (2007). Measuring Angler Attitudes Toward CatchRelated Aspects of Fishing, Human Dimensions of Wildlife, 12,181-191.
Arlinghaus, R. (2006). Understanding recreational angling participation in Germany: Preparing
for demographic change. Human Dimensions of Wildlife, 1:229-240
Arlinghaus, R., Tillner, R. & Bork, M. (2014). Explaining participation rates in recreational
fishing across industrialized countries. Fisheries Management and Ecology.
doi: 10.1111/fme.12075
Arthur, L. (2013). Big Data Marketing: Engage Your Customers More Effectively and Drive
Value. Hoboken, New Jersey: John Wiley & Sons.
Birkin, M., Clarke, G., Clarke, M. & Wilson, A. (1996). Intelligent GIS: location decisions and
strategic planning. GeoInformation International: Cambridge.
Bissel, S J., Duda, M.D., Young, K.C. (1998). Recent studies on hunting and fishing
participation in the United States, Human Dimensions of Wildlife, 3:1,75-80.

124

Brown, T. L. (1987). Typology of Human Dimensions Information Needed for Great Lakes
Sport-Fisheries Management. Transaction of the American Fisheries Society, 116, 320-324.
Bruwer, J. & Li, E. (2007). Wine-Related Lifestyle (WRL) Market Segmentation: Demographic
and Behavioural Factors. Journal of Wine Research 18, 19-34.
Bryan, H. (1977). Leisure value systems and recreational specialization: the case of trout
fishermen. Journal of Leisure Research 9, 174-187.
Chase, L. C., Decker, D. J. & Lauber, T. B. (2004). Public participation in wildlife management:
What do stakeholders want? Society and Natural Resources 17, 629-639.
Chipman, B. D., & Helfrich, L. A. (1988). Recreational specializations and motivations of
Virginia river anglers. North American Journal of Fisheries Management 8, 390-398.
Choice Metrics. (2011). “NGene 1.1 User Manual Reference Guide.” Version 13/05/2011.
http://www.choice-metrics.com/index.html
Claramunt, R.M., Madenjian, C.P., & Clapp, D.F. (2013). Pacific Salmonines in the Great Lakes
Basin. In B. Taylor, A. J. Lynch, & N.J. Leonard (Eds.). Great Lakes Fisheries Policy &
Management (2nd edition) (pp.609-650). East Lansing, MI: Michigan State University Press.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Mahwah, NJ:
Lawrence Erlbaum Associates
Cohen, J. (1992). A Power Primer. Psychological Bulletin, 112 (1) 155-159.
Collier, B.A., Groce, J.E, Morrison, M.L, Newman, J.C, Campomizzi, A.J, Farrell, S.L.,
Mathweson, H.A., Snelgrove, R.T, Carroll, R.J., & Wilkins, R.N. (2012). Predicting patch
occupancy in fragmented landscapes at the rangewide scale for an endangered species: an
example of an American warbler. Diversity and Distributions18, 158-167
Connelly, N. A., Brown, T. L., & Knuth, B. A. (2000). A Multiple Market Segmentation of
Great Lakes Anglers in New York. North American Journal of Fisheries Management 20(2),
399-407.
Connelly, N. A., Keeler, S., & Knuth, B. A. (2013). Understanding angler preferences for
species and water bodies: Importance for New York’s Future Fisheries Management. Human
Dimensions of Wildlife 3(5), 386-402.
Connelly, N. A., Knuth, B. A., & Brown, T.L. (2001). An angler typology based on angler
fishing preferences. Transactions of the American Fisheries Society 130(1),130-137.
Crawford, D. W., Jackson, E. L., Godbey, G. (1991). A hierarchical model of leisure
constraints. Leisure Sciences, 13:309-320.

125

Crompton, J. L.& Lamb, C.W. (1986). Marketing Government and Social Services. Hoboken,
NJ: John Wiley & Sons, Inc.
Curcione, N. (1980). A reconceptualization of angler typologies. Review of Sport & Leisure, 5,
97-113.
Dann, S.L., Alvarado, A., Palmer, D., Schroeder, B., Stephens, M. (2008). Angler Participation,
Recruitment, and Retention in Michigan, 1995-2004: Using Data-mining Techniques for
Custmer Relationship Management. Michigan Department of Natural Resources, Fisheries
Research Report 2088, Ann Arbor.
Dillman, D. A. (2007). Mail and Internet Surveys, The Tailored Design Method. Hoboken, New
Jersey: John Wiley & Sons, Inc.
Ditton, R. B. (1996). Understanding diversity among largemouth bass anglers. In L. E. Miranda
Editor & D. DeVries, Editor (Eds.). Multidimensional Approaches to Reservoir Fisheries
Management: Symposium 16 (pp. 135-144). Bethesda, MD: American Fisheries Society
Ditton R. B. (2004). Human dimensions of fisheries. In M. J. Manfredo, J.J. Vaske, B. L.
Bruyere, D. R. Field, and P.J. Brown (Eds.). Society and natural resources: a summary of
knowledge prepared from the 10th international symposium on society and resource management
(pp.199-208). Jefferson City, Missouri: Modern Litho,
Dobson, T., Riley, S., Gaden, M. (2005). Human Dimensions of Great Lakes Fishery
Management: New Research Thrust of the Great Lakes Fishery Commission. Society and
Natural Resouces,18(5)487-491
Driver, B. L. (1977). Item pool for scales designed to quantify the psychological outcomes
desired and expected from recreation participation. Unpublished report. USDA Forest Service,
Rocky Mountain Forest and Range Experiment Station. Fort Collins.
Duda, M. D., Bissell, S. J., & Young, K. C. (1998). Wildlife and the American Mind: Public
opinion on and attitudes toward fish and wildlife management (1st edition). Harrisonburg, VA:
Responsive Management.
Duffey, E. Spider Habitat Classification and the Development of Habitat Profiles.
Arachnology,15(1)1-20
Environmental Protection Agency (2014). Great Lakes Basic Information: Geography &
Hydrology. Accessed April 2014
http://www.epa.gov/greatlakes/basicinfo.html
Esri TapestryTM Segmentation Reference Guide. Accessed February 2014.
http://www.esri.com/library/brochures/pdfs/tapestry-segmentation.pdf

126

Everitt, B.S., Landau, S., Leese, M, Stahl, D. (2011). Cluster Analysis. 5th Edition. John Wiley
& Sons, Ltd.
Fedler, A. J. (2000). Participation in boating and fishing: a literature review. Report of Human
Dimensions Consulting to the Recreational Boating and Fishing Foundation. Alexandria,
Virginia.
Fedler, A.J.& Ditton, R. B.. (2001). Dropping out and dropping in: a study of factors for
changing recreational fishing participation. North American Journal of Fisheries Management
21:283-292.
Finn, K. L, & Loomis, D. K. (2001). The importance of Catch Motives to Recreational Anglers:
The Effects of Catch Satiation and Deprivation. Human Dimensions of Wildlife, 6, 173-187.
Fisher, M. R. (1997). Segmentation of the Angler Population by Catch Preference, Participation,
and Experience: A Management-Oriented Application of Recreation Specialization. North
American Journal of Fisheries Management, 17(1) 1:10
Fisher, W. L. and Burroughs, J. P.. (2003). Stream fisheries Management in the United States,
Fisheries, 28:2, 10-18.
Fishbein, M & Ajzen, I. (1975). Belief, Attitude, Intention and Behavior; An Introduction to
Theory and Research. Reading, Massachusetts: Addison-Wesley.
Floyd, M.F. & Lee, I. (2002). Who buys Fishing and Hunting Licenses in Texas? Results from
a Statewide Household Survey. Human Dimensions of Wildlife,7(2),91-106
Floyd, M. F., Nicholas, L, Lee, I., Lee, J. and Scott, D. (2006). Social stratification in
recreational fishing participation: research and policy implications. Leisure Sciences 28:351368.
Frissell, C.A, Liss, W.J., Warren, C.E., & Hurley, M.D. (1986). A Hierarchical Framework for
Stream Habitat Classification: Viewing Streams in a Watershed Context. Environmental
Management,10 (2)199-214.
Gawler, S. C. (2008). Northeastern Terrestrial Wildlife Habitat Classification. Report to the
Virginia Department of Game and Inland Fisheries on behalf of the Northeast Association of
Fish and Wildlife Agencies and the National Fish and Wildlife Foundation. NatureServe,
Boston, MA
George, G., Hass, M., Pentland, A. (2014). From the Editors: Big Data and Management.
Academy of Management Journal 57:2,321-326
Gilner, J. A., Vaske, J.J., & Morgan, G.A. (2001). Null hypothesis significance testing: Effect
size matters. Human Dimensions of Wildlife an International Journal, 6, 291-301,

127

Gigliotti, L. M., and Peyton, R. B. (1993). Values and behaviors of trout anglers, and their
attitudes toward fishery management, relative to membership in fishing organizations: a
Michigan case study. North American Journal of Fisheries Management 13, 492-501
Gonzalez, A. M. & Bello, L. (2002). The construct “lifestyle” in market segmentation The
behaviour of tourist consumers. European Journal of Marketing, 36 (1/2) 55-85.
Greene, W. H. (2002) Econometric Analysis. Prentice Hall: Upper Saddle River, New Jersey
Hanley, N., R. E. Wright, V. Adamowicz. (1998). Using Choice Experiments to Value the
Environment. Environmental and Resource Economics 11(3-4): 413-428
Harris, R., Sleight, P. & Webber, R. (2005). Geodemographics, GIS and Neighbourhood
Targeting. West Sussex, England; Hoboken, N.J. John Wiley & Sons LTD
Harrison, G. W. (2007) Making Choice Studies Incentive Compatible. Pages 67-110 in B.J.
Kanninen (ed.), Valuing Environmental Amenities Using Stated Choice Studies, © 2007
Springer.
Harrison, G.W. and J.A. List (2004). Field Experiments. Journal of Economic Literature, 42,
1009-1055.
Hawkins, D.I. and Mothersbaugh, D. L. (2010). Consumer Behavior: Building Marketing
Strategy (11th edition). New York, NY: McGraw-Hill/Irwin.
Heath, E. & Wall, G. (1992). Marketing Tourism Destinations. New York, NY: John Wiley and
Sons.
Heberlein, T. A., Serup, B. & Ericsson, G. (2008). Female Hunting Participation in North
America and Europe, Human Dimensions of Wildlife, 13:6, 443-458.
Holt, D., & Smith, T.M.F. (1979). Post-stratification. Journal of the Royal Statistical Society
142(1):33-46.
Hunt, K. M. & Ditton, R. B. (2002). Freshwater fishing participation patterns of racial and
ethnic groups in Texas. North American Journal of Fisheries Management 22:52-65.
Hunt, K.M, & Grado S.C. (2010). Use of Social and Economic Information in Fisheries
Management. In W. Hubert & M. Quist (Eds.) Inland Fisheries Management in North America
(3rd edition) (pp. 425-447). American Fisheries Society: Bethesda, MD.
Hunt, K. M, Floyd, M.F. & Ditton R. B. (2007). African-American and Anglo Anglers’
Attitudes toward the Catch-Related Aspects of Fishing. Human Dimensions of Wildlife, 12(4)
227-239.

128

Hunt, L. M. (2005). Recreational Fishing Site Choice Models: Insights and Future Opportunities.
Human Dimensions of Wildlife, 10:3, 153-172.
Hunt, L.M., D. Gonder, W. Haider. (2010). Hearing voices from the Silent Marjority: A
Comparison of Preferred Fish Stocking Outcomes for Lake Huron by Anglers from
Representative and Convenience Samples. Human Dimensions of Wildlife,15;1,27-44.
Hutt, C. P., & Bettoli, P. W. (2007). Preferences, Specialization, and Management Attitudes of
Trout Anglers Fishing in Tennessee Tailwaters. North American Journal of Fisheries
Management 27:1257-1267.
Hvenegaard, G. T. (2002). Birder specialization differences in conservation involvement,
demographics, and motivations. Human Dimensions of Wildlife, 7 21-26.
Illinois Department of Natural Resources (2009). Report of the Youth Recruitment and Retention
Committee of the Natural Resources Advisory Board to the Illinois Department of Natural
Resources and Conservation Congress. Retrieved from
http://dnr.state.il.us/nrab/children.htm
Jamsen, G. (1985). Michigan’s 1981 and 1982 Sport Fishery. Michigan Department of Natural
Resources Fisheries Technical Report 85-4 and 85-5, Ann Arbor, MI.
Johnston, F., Arlinghaus, R., & Dieckmann, U. (2010). Diversity and complexity of angler
behavior drive socially optimal input and output regulations in a biocentric recreational-fisheries
model. Canadian Journal of Fisheries and Aquatic Science, 67, 1507-1531
Kaplowitz, M., Lupi, F., Hoehn, J., (2004). Multiple methods for developing and evaluating a
stated preference survey to value wetland ecosystems. In S. Presser, J.M. Rothgeb, M. P. Couper,
J.T. Lessler, E. Martin, J. Martin & E. Singer (Eds.). Methods for Testing and Evaluating
Survey Questionnaires (pp. 503–524). John Wiley and Sons: Hoboken, NJ.
Kotler, P. (1984). Marketing management: analysis, planning and control. 5th edition.
Prentice-Hall Inc.: Englewood Cliffs, New Jersey.
Krost, M. A., Albert, D.A., Cohen, J.G., Slaughter, B.S., Schillo, R. K., Weber, C.R. &
Chapman, K. (2007). Natural Communities of Michigan: Classification and Description.
Michigan Natural Features Inventory, 2007-21, V1.2. Updated July 9, 2010.
Krueger, C.C., & D.J. Decker. (1999). The process of fisheries management. In C. C. Kohler &
W.A. Hubert (Eds.). Inland fisheries management in North America, 2nd edition (pp. 31-59).
American Fisheries Society: Bethesda, Maryland.
Kuehn, D., Luzadis, V., & Brincka, M. (2013). An Analysis of the Factors Influencing Fishing
Participation by Resident Anglers. Human Dimensions of Wildlife, 18:5, 322-339

129

Kyle, G., Norman, W., Jodice, L., Graefe, A., & Marsinko, A. (2007). Segmenting Anglers
Using Their Consumptive Orientation Profiles. Human Dimensions of Wildlife 12, 115-132.
Lake Erie Committee (2013). 2013 Walleye and Yellow Perch Catch Levels Recommended for
Lake Erie.
http://www.glfc.org/lakecom/lec/YPTG_docs/other_reports_and_docs/YPTG_release_3-2813.pdf
Larson, L R., Stedman, R. C., Decker, D.J., Siemer, W. F. & Baumer, M.S. (2014). Exploring
the Social Habitat for Hunting: Toward a Comprehensive Framework for Understanding Hunter
Recruitment and Retention. Human Dimensions of Wildlife 19, 105-122
Lawson, R. & Todd, S. (2002) Consumer Lifestyles: A Social Stratification Perspective.
Marketing Theory, 2(3), 295-307
Liu, Y., Kiang M., & Brusco, M. (2012). A unified framework for market segmentation and its
applications. Expert Systems with Applications, 39, 10292-10302.
Louviere, J.J., Hensher, D. A., & Swait, J.D. (2000). Stated Choice Methods: Analysis and
Applications. New York: Cambridge University Press.
Louviere, J.J., & Woodworth, G. (1983). Design and analysis of simulated consumer choice or
allocation experiments: An approach based on aggregated data. Journal of Marketing Research,
20, 350-367.
Lupi, F., D. Hayes, Z. Su. 2005. F-81-R-4, Study 230548-4. Michigan Department of Natural
Resources. A statewide survey of Michigan’s licensed anglers.
Malcolm, H.A., Smith, S.D.A, Jordan, A. (2010). Using patterns of reef fish assemblages to
refine a Habitat Classification System for marine parks in NSW, Australia. Aquatic
Conservation: Marine and Freshwater Ecosystems,20:83-92
Manfredo, M. J., D. J. Decker, M. Duda. (1998) What is the Future for Human Dimensions of
Wildlife? Transactions of the North American Wildlife and Natural Resources Conference,
64, 278-292.
Martin,D.R. and K.L. Pope. (2010). Luring anglers to enhance fisheries. Journal of
Environmental Management, 92 1409-1413.
Martin, D.R., Pracheil, B.M., DeBoer, J.A., Wilde, G., Pope, K.L. (2012). Using the Internet to
Understand Angler Behavior in the Information Age. Fisheries,37(10): 458-463.
Mather, M. E., Parrish, D.L., Stein, R.A, & Muth, R.M. (1995). Management Issues and Their
Relative Priority within State Fisheries Agencies. Fisheries 20(10):14- 21

130

McCann, J.M. (1974). Market segment response to the marketing decision variables. Journal of
Marketing Research,11,4: 399-412.
McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. In P. Zarembka
(Ed.). Frontiers in econometrics (pp. 105-142). Academic Press: New York, NY.
McKinsey Global Institute. (2011). Big data: The next frontier for innovation, competition, and
productivity. KcKinsey & Company: Lexington, KY.
Michigan Department of Natural Resources (2013). MiScorecard Performance Summary.
Retrieved from: https://www.michigan.gov/documents/dnr/DNR-Scorecard_422577_7.pdf
Minnesota Department of Natural Resources. (2013). Recruiting and Retaining Minnesota
Hunters and Anglers. Retrieved from: http://www.dnr.state.mn.us/harr/index.html
Milligan, G. W. (1996). Clustering validation: Results and implications for applied analyses. In
P. Arabie, L.J. Hubert & G. de Soete (Eds.). Clustering and Classification (pp. 341-375). World
Scientific Publishers, New Jersey.
Mitchell, S. (1995). Birds of a feather. American Demographics, 17,( 2):40-48.
Moll, R.A., Sellinger, C., Rutherford, E.S., Johnson, J.L., Fainter, M.R., & Gannon, J.E. (2013).
The Great Lakes: An Overview of Their Formation, Geology, Physics and Chemistry. In B.
Taylor, A. J. Lynch, & N.J. Leonard (Eds.). Great Lakes Fisheries Policy & Management (2nd
edition) (pp.3-29). East Lansing, MI: Michigan State University Press.
Morissette, J.L., Kardynal, K.J., Bayne, E.M, & Hobson, K.A. (2013). Comparing Bird
Community Composition Among Boreal Wetland: Is Wetland Classification a Missing Piece of
the Habitat Puzzle. Wetlands,33:653-665
Moscardo, G., Pearce, P., & Morrison, A. (2001): Evaluating Different Bases for Market
Segmentation: A Comparison of Geographic Origin versus Activity Participation for Generating
Tourist Market Segments. Journal of Travel & Tourism Marketing, 10(1), 29-49
Murdock, S. H., K. Backman, R. B. Ditton, M. N. Hoque, D. Ellis (1992). Demographic Change
in the United States in the 1990s and the Twenty-first Century: Implications for Fisheries
Management. Fisheries, 17:(2),6-13.
Oregon Department of Fish and Wildlife (2012). Recruitment, Retention and Re-activation
Action Plan. Retrieved from:
http://www.dfw.state.or.us/agency/commission/minutes/13/06_june
Organ, J.F., Geist, V., Mahoney, S.P., Williams, S., Krausman, P.R., Batcheller, G.R.,…Decker,
D. J. (2012). The North American Model of Wildlife Conservation (No. Technical Review 1204). Bethesda, MD: The Wildlife Society and The Boone and Crockett Club.

131

O’Shaughnessy, J. (2013). Consumer Behaviour: Perspectives, Findings and Explanations,
Palgrave Macmillan: Hampshire, England
Pierce, C.L., Manfredo, M., & Vaske, J. (2001). Social Science Theories in Wildlife
Management. In J.D. Decker, T.L. Brown, & W.F. Siemer (Eds.). Human Dimensions of
Wildlife Management in North America (pp. 39-56). The Wildlife Society: Bethesda Maryland.
Responsive Management. (2009). Increasing Freshwater Fishing Participation and Fishing
License Sales in Washington State: A Marketing Plan for the Washington Department of Fish
and Wildlife. Duda, M.D., Jones, M., Beppler, T., Bissell, S.J., Criscone, A., Herrick, J.B.,
Leftwich, B., Nobile, J., Ritchie, A., Schilli, C.L., Wilkes, M., Winegord, T., Lanier, A.
Harrisburg, VA.
Riley, S.C. [ED]. 2013. The state of Lake Huron in 2010. Great Lakes Fish. Comm. Spec. Pub.
13-01
Romberg, W. J. (1999). Market segmentation, preferences and management attitudes of Alaska
nonresident anglers. Master’s thesis. Virginia Polytechnic Institute and State University,
Blacksburg.
Ross, M. R. & Loomis, D.. (1999). State Management of Freshwater Fisheries Resources: Its
Organizational Structure, Funding, and Programmatic Emphases. Fisheries. 24(7):8-14.
Schroeder, S. A., Fulton, D.C., Currie, L., & Goeman, T. (2006). He Said, She Said: Gender and
Angling Specialization, Motivations, Ethics and Behaviors. Human Dimensions of Wildlife,
11(5), 301-315.
Schroeder, S. A., M. L. Nemeth, R. E, Sigurdson,& R. J. Walsh. (2008). Untangling the
Line:Constraints to Fishing Participation in Communities of Color. American Fisheries Society
Symposium, 67.
Schroeder, S.A., D.C. Fulton, M.L. Nemeth, R.E. Sigurdson & R J. Walsh (2008). Fishing in the
neighborhood: Understanding Motivations and Constraints for Angling Among MinneapolisSt.Paul Minnesota Metro Residents. American Fisheries Society Symposium 67,
Schoch, G.C, Albert, D.M.&Shanley, C.S. (2014). An Estuarine Habitat Classification for a
Complex Fjordal Island Archipelago. Estuaries and Coasts, 37:160-176
Scott, D., Ditton, R.B., Stoll, J.R. & Eubanks, T.L. (2005). Measuring Specialization among
Birders: Utility of a Self-Classification Measure. Human Dimensions of Wildlife, 10(1)53-71.
Shafer, E. L. (1969). The average camper who doesn’t exist. U.S. Forest Service Research
Paper NE-142.

132

Simoes, J. C. (2009). Recreational Angler Surveys: Their role and importance nationally and
the 2008 Michigan Angler Survey. Unpublished Thesis, Michigan State University, East
Lansing, MI.
Smith, S.L.J. (1995). Tourism Analysis: a handbook. Harlow, Essex, England: Longman
Southwick Associates. (2006). IAFWA-ASA Data Mining Project: Initial Michigan Findings.
http://www.southwickassociates.com/products/projects-by-us-state/
Accessed February 2014.
Southwick Associates. (2007). Lifestyles and License Buying Habits of America’s Anglers.
http://www.southwickassociates.com/products/projects-by-us-region/
Accessed February 2014.
Southwick Associates. (2012). Annual 2012 Angler Participation and Trends-Sample Report.
http://www.southwickassociates.com/services/surveys/
Accessed March 2014
Sutton, S. (2003). Personal and Situational Determinants of Catch-and-Release Choice of
Freshwater Anglers. Human Dimensions of Wildlife, 8,109-126.
Sutton, S.G., Dew, K., Higgs, J. (2009). Why do People Drop Out of Recreational Fishing? A
Study of Lapsed Fishers from Queensland, Australia. Fisheries, 34:9, 443-452.
Sutton, S. G., & Tobin, R. C. (2009). Recreational fishers’ attitudes towards the 2004 rezoning of
the Great Barrier Reef Marine Park. Environmental Conservation 36:245-252.
Tabachnick, B. G, & Fidell, L. S. (2001). Using multivariate statistics. 4th edition. Allyn &
Bacon, Needham Heights, MA.
Thach, E. C. & Olsen, J. E. (2004). The Search for New Wine Consumers: Marketing Focus on
Consumer Lifestyle or LifeCycle. International Journal of Wine Marketing, 16(3)44-57.
Thayer, S.A., & Loftus, A.J. (2013). Great Lakes Recreational Fisheries and Their Role in
Fisheries Management Policy. In B. Taylor, A. J. Lynch, & N.J. Leonard (Eds.). Great Lakes
Fisheries Policy & Management (2nd edition) (pp. 399-440). East Lansing, MI: Michigan State
University Press.
Todd, S. & Lawson, R. (2001). Lifestyle segmentation and museum/gallery visiting behaviour.
International Journal of Nonprofit and Voluntary Sector Marketing, 6(3)269-277.
Tonks, D. G. 2009. Validity and the design of market segments. Journal of Marketing
Management, 25(3/4), 341-356.

133

Tonn, W.M., Magnuson, J.J, Forbes, A.M. (1983). Community Analysis in Fishery
Management: An Application with Northern Wisconsin Lakes. Transactions of the American
Fisheries Society,112(3)368-377
Train, K. (2009). Discrete choice methods with simulation. Cambridge University Press.
U.S. Department of Agriculture (2011). Watershed Condition Framework. United States
Department of Agriculture Forest Service. FS-977.
U.S. Fish & Wildlife Service (2012). 2011 National Survey of Fishing, Hunting and WildlifeAssociated Recreation: U.S. Department of the Interior, Fish and Wildlife Service, and U.S.
Department of Commerce.
U.S. Fish & Wildlife Service (2014). Historical fishing license data. Retrieved from
http://wsfrprograms.fws.gov/Subpages/LicenseInfo/Fishing.htm
U.S. National Wetlands Inventory. (2014). http://www.fws.gov/wetlands/Data/WetlandCodes.html
Vaske, J.J., & Manfredo, M. J. (2012). Social psychological considerations in wildlife
management. In D. J. Decker, S. J. Riley, & W. F. Siemer (Eds.). Human dimensions of wildlife
management (2nd edition.) (pp. 43-57). Baltimore, MD: Johns Hopkins University Press.
Vogt, C. A. (2011). Customer Relationship Management in Tourism: Management Needs and
Research Applications. Journal of Travel Research, 50(4), 356-364.
Walker, G. J. and Virden, R. J. (2005). Constraints on outdoor recreation. In E. L. Jackson
(Ed.). Constraints to leisure (pp. 201-209). State College, Pennsylvania: Venture Publishing.
Warshaw, P.R. (1980a): Predicting Purchase and Other Behaviors from General and
Contextually Specific Intentions. Journal of Marketing Research, 17:(1) 26-33
Warshaw, P.R. (1980b): A New Model for Predicting Behavioral Intentions: An Alternative to
Fishbein, Journal of Marketing Research, 17:(2)153-172
Wehrly, K.E., Breck, J.E., Wang, L. & Szabo-Kraft, L. (2012). A Landscape-Based
Classification of Fish Assemblages in Sampled and Unsampled Lakes, Transactions of the
American Fisheries Society, 141(2),414-425
Weicksel, S. A. (2012). Measuring Preferences for Changes in Water Quality at Great Lakes
Beaches using a Choice Experiment. Unpublished Thesis. Michigan State University
Wilde, G. R., Ditton, R. B., Grimes, S. R., & Riechers, R. K. (1996). Status of Human
Dimensions Surveys Sponsored by State and Provincial Fisheries Management Agencies in
North America. Fisheries. 21(11):12-17.

134

Wilkie, W. L. (1994). Consumer Behavior, (3rd edition). New York: John Wiley and Sons.
Zhu, H., Wang, Q., Yan, L., & Wu, G. (2009). Are consumers what they consume?--Linking
lifestyle segmentation to product attributes: an exploratory study of the Chinese mobile phone
market. Journal of Marketing Management, 25(3/4), 295-314.

135