RECREATIONAL SPECIALIZATION AMONG MICHIGAN SNOWMOBILERS:
DIFFERENTIATION, RISK PERCEPTION, AND GEOGRAPHIC INTERACTION
By
I-Chun Wu

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Sustainable Tourism and Protected Area Management – Doctor of Philosophy
2014

ABSTRACT
RECREATIONAL SPECIALIZATION AMONG MICHIGAN SNOWMOBILERS:
DIFFERENTIATION, RISK PERCEPTION, AND GEOGRAPHIC INTERACTION
By
I-Chun Wu
Snowmobiling is an integral part of Michigan’s tourism economy with over two million
snowmobile use days annually. The extensive designated snowmobile trail system and abundant
snowfall have attracted passionate snowmobilers who travel long distances and stay overnight.
The theoretical construct of recreational specialization was applied to understand the withingroup social world of the snowmobilers. To foster safer riding experiences and reduce the
number and severity of snowmobiling accidents, snowmobilers’ perceptions of key riding risks
were examined by looking at the effect of levels of specialization on the perception of risk. The
spatial interaction between snowmobilers’ residential origin and snowmobiling destination was
also investigated to better understand the association between recreational specialization and
spatial interaction.
A total of 807 research subjects came from a statewide survey of Michigan snowmobile
trail permit purchasers in the winter of 2007-08. Daily snow precipitation for the winter of 200708 and Michigan snowmobile trail GIS shapefiles were also utilized. Summative indices and
cluster analysis were applied to segment different snowmobile subgroups. One-way ANOVA,
effect size Eta, coupled with a gravity model and spatial statistics were applied to investigate the
relationships among snowmobilers’ within-group social world, their perception of risks, and
spatial interaction with the snowmobile trail network.

The results showed that Michigan snowmobilers were heterogeneous, with distinct
behavioral, skill/knowledge, and psychological attachment characteristics. Some became more
specialized in snowmobiling and spent increasingly more time and money on the activity. Others
focused on different aspects of snowmobiling such as trail riding or conspicuous consumption,
once they acquired enough snowmobiling experience. Recreational specialization did not
effectively predict perception of risks as Michigan snowmobilers perceived similar threats in all
nine hazardous riding stimuli. While expert snowmobilers perceived significantly lower risk on
“Speed of snowmobile,” novice snowmobilers underestimated the risk of “Cars/trucks on
seasonal roads.” Neither did recreational specialization explain snowmobilers’ spatial interaction
with snowmobiler trail network. Some minimal to moderate associations were found between
snowmobilers travel from region to region. While Michigan’s Upper Peninsula attracted expert
snowmobilers from the Lower Peninsula, the northern Lower Peninsula was the favorite
destination for intermediate snowmobilers.
Key managerial recommendations from the analysis of this data include establishing a
mandatory snowmobile training course; more robust mechanisms of disseminating snowmobile
information; establishing a voluntary safety patrol program; creating a digital profile for
snowmobile accidents involving personal injury or fatality; and more comprehensive
snowmobile marketing strategy for residents and visitors.

Copyright by
I-CHUN WU
2014

This Doctoral dissertation is dedicated to my family.
The sincere gratitude to my loving father, Ei-Kwang, Wu and mother, Chi-Fang, Huang for their
devotion and endless support that made this seven lonely years of study abroad possible. My
sister, Hsing-Yin, Wu, a talented animator, who always gives me courage and laughter, helped
me sail through much difficult time along the way. My baby brother, Che-An, Wu, an earnest
man, reminds me of myself 10 years younger, whose passion and determination always inspire
me to work harder.
There is no doubt that without their continued encouragement and counsel, I could not have
completed this process.

v

ACKNOWLEDGEMENTS

I would never have been able to complete my dissertation without the guidance of my
committee members, help from friends, and support from family.
I would like to thank my committee members who were more than generous with their
expertise and precious time. My deepest gratitude goes to Dr. Charles Nelson, my committee
chairman for his countless hours of reflecting, reading, encouraging, caring, and most of all
patience throughout the entire process. I would also like to thank Dr. Richard Paulsen, Dr. Sarah
Nicholls, and Dr. Ashton Shortridge for agreeing to serve on my committee. Dr. Paulsen has
been invaluable on both an academic and a personal level. He served as the academic adviser for
my first two years at MSU, whose kindness and generosity helped me get through those years of
accommodation, for which I am extremely grateful. I would like to thank Dr. Nicholls and Dr.
Shortridge for guiding my research for the past several years and helping me to develop my
background in spatial analysis, not to mention their advice and unsurpassed knowledge of
Geographic Information Science.
Special thanks go to other faculty and staff members of Community Sustainability
department for their continued support.
Last, but by no means least, I thank my family and friends in Taiwan, America, and
elsewhere for their support and encouragement. It has been a humble journey.

vi

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... ix
LIST OF FIGURES ..................................................................................................................... xi
Chapter One
Introduction ................................................................................................................................... 1
Background Information ............................................................................................................. 1
Problem Statement ...................................................................................................................... 6
Research Objectives .................................................................................................................... 8
Research Questions ..................................................................................................................... 8
Conceptual Framework and Null Hypotheses ............................................................................. 9
Research Significance ............................................................................................................... 12
Chapter Two
Literature Review ....................................................................................................................... 15
Recreational Specialization ....................................................................................................... 15
Original Concept.................................................................................................................... 15
Recreational Specialization and Behavior, Attitude, and Perception .................................... 17
Recreational Specialization Measurements ........................................................................... 19
Perception of Risk ..................................................................................................................... 22
Concept and Measurement .................................................................................................... 22
Snowmobile Accidents and Fatalities.................................................................................... 24
Spatial Interaction ..................................................................................................................... 27
Application of the Gravity Model ......................................................................................... 29
Spatial Interaction Measurements ......................................................................................... 30
Chapter Three
Research Methods ....................................................................................................................... 34
Research Data ............................................................................................................................ 34
Snow Precipitation Data ........................................................................................................ 34
Michigan Snowmobile Trail Shapefiles ................................................................................ 35
Survey Data ........................................................................................................................... 35
Survey Instrument.................................................................................................................. 36
Data Analysis ............................................................................................................................ 38
Identifying Snowmobile Specialization................................................................................. 41
Measuring Perception of Risk ............................................................................................... 42
Examining the Influence of Specialization on the Perception of Risk .................................. 43
Measuring Spatial Interaction................................................................................................ 45
Calculating Winter Snowfall in Region 𝑗 .............................................................................. 49
Assessing the Relationships between Spatial Interaction and Specialization ....................... 49
Definition and Treatment of Outliers ........................................................................................ 50
vii

Treatment of Missing Data........................................................................................................ 51
Chapter Four
Research Findings ....................................................................................................................... 52
Descriptive Statistics ................................................................................................................. 52
Surveyed Michigan Snowmobilers ........................................................................................ 52
The Year of 2007-08 Winter Snowfall .................................................................................. 53
Michigan Snowmobile Trail .................................................................................................. 54
Most Snowmobile Friendly Point in Each Region ................................................................ 55
General Characteristics of Michigan Snowmobilers ............................................................. 56
Michigan Snowmobilers’ General Perception of Risk and Internal Consistency ................. 57
Michigan Snowmobilers’ Actual Participation ..................................................................... 58
Travel Distance to Michigan Snowmobile Trail Network .................................................... 60
Spatial Interaction between Snowmobiler’ Residential Region and Snowmobile Trail
Network ................................................................................................................................. 60
Inferential Statistics ................................................................................................................... 63
Michigan Snowmobilers’ Recreational Specialization.......................................................... 63
Summative Index for Recreational Specialization ............................................................. 63
Cluster Analysis for Recreational Specialization .............................................................. 68
Summary of Comparison of Summative Index and Cluster Analysis................................. 73
Influence of Specialization on the Perception of Risk .......................................................... 75
Influence of Specialization on Spatial Interaction ................................................................. 84
Spatial Interaction Result by Destination Regions ............................................................ 85
Spatial Interaction Result by Snowmobilers’ Residential Regions .................................... 87
Section Summary ................................................................................................................ 95
Chapter Five
Discussion, Implications, and Conclusions ............................................................................... 97
Research Background ................................................................................................................ 97
Discussion ................................................................................................................................. 98
Michigan Snowmobilers’ Within-group Social World ....................................................... 100
Influence of Specialization on the Perception of Risk ........................................................ 104
Relationships between Recreational Specialization and Spatial Interaction ....................... 109
Managerial Implications and Recommendations .................................................................... 112
General Managerial Implications ........................................................................................ 112
Administrative Recommendations ...................................................................................... 113
Research Limitation ................................................................................................................ 119
Future Research ....................................................................................................................... 121
Conclusion............................................................................................................................... 124
APPENDIX ................................................................................................................................ 127
BIBLIOGRAPHY ..................................................................................................................... 136

viii

LIST OF TABLES

Table 1 Dimension, Data Type, and Scale of Snowmobile Specialization Measurement ............ 39
Table 2 Scale and Data Type of the Perception of risks Measurement ........................................ 43
Table 3 Winter Snowfall by Region (2007-08) ............................................................................ 53
Table 4 Michigan Snowmobilers’ General Characteristics .......................................................... 56
Table 5 Descriptive Statistics and Reliability of Snowmobiling Perception of Risk ................... 57
Table 6 Michigan Snowmobilers’ Actual Participation in Five Michigan Regions ..................... 59
Table 7 Average Travel Distance for All Snowmobilers to Trail Network in each Region ......... 60
Table 8 Total Spatial Interaction Received by Snowmobile Trail Network Destination Region . 61
Table 9 Spatial Interaction between Snowmobilers and Trail Network by Region ..................... 62
Table 10 Summative Index Result for Snowmobilers’ Recreational Specialization .................... 64
Table 11 Pairwise Comparison Result of Summative Index for Snowmobilers’ Specialization . 67
Table 12 Cluster Analysis Result for Snowmobilers’ Recreational Specialization...................... 69
Table 13 Pairwise Comparison Result of Cluster Analysis for Snowmobilers’ Specialization ... 72
Table 14 Michigan Snowmobilers’ Perception of risks by Specialization Subgroup .................. 79
Table 15 Comparison of Perception of Risk Across Specialization Subgroups ........................... 81
Table 16 Spatial Interaction between Michigan Snowmobilers and Five Regions ...................... 86
Table 17 Comparison Result of Spatial Interaction Among Specialization Subgroups ............... 87
Table 18 Spatial Interaction between Region One and Two Snowmobilers and Trail Network . 88
Table 19 Relationship between Region One and Two Specialization and Spatial Interaction ..... 89
Table 20 Spatial Interaction between Region Three Snowmobilers and Trail Network .............. 91
Table 21 Comparison Result of Spatial Interaction for Region Three Specialization Subgroups 91
ix

Table 22 Spatial Interaction between Region Four Snowmobilers and Trail Network ................ 92
Table 23 Comparison Result of Spatial Interaction for Region Four Specialization Subgroups . 93
Table 24 Spatial Interaction between Region Five Snowmobilers and Trail Network ................ 94
Table 25 Comparison Result of Spatial Interaction for Region Five Specialization Subgroups .. 95
Table 26 Research Hypotheses, Tests, and Outcome ................................................................... 99

x

LIST OF FIGURES

Figure 1 Michigan Snowmobile Trail Network .............................................................................. 2
Figure 2 Trends of Michigan Snowmobile Fatalities ..................................................................... 5
Figure 3 Conceptual Framework .................................................................................................... 9
Figure 4 Recreational Specialization Continuum ......................................................................... 16
Figure 5 Correlations between Risk Behavior and Risk Perception ............................................. 23
Figure 6 Locations of 82 Weather Stations................................................................................... 34
Figure 7 Michigan’s Five Regions ................................................................................................ 45
Figure 8 Number of Surveyed Snowmobilers by County............................................................. 53
Figure 9 Winter 2007-08 Snowfall Kriging Surface with Snowmobile Trail .............................. 54
Figure 10 Snowmobile Trail Network in Five Regions ................................................................ 55
Figure 11 Most Snowmobile Friendly Point in each Region ........................................................ 56
Figure 12 Michigan Snowmobiler Recreational Specialization – Summative Index Result ........ 68
Figure 13 Michigan Snowmobiler Recreational Specialization – Cluster Analysis Result.......... 73

xi

Chapter One
Introduction

Background Information
Snowmobiling is a flourishing outdoor activity that involves a motor-driven vehicle
traveling on snow and ice for transportation or recreational purposes. Since the early 1970s,
millions of Americans have participated in snowmobiling and manufacturers have developed
increasingly sophisticated machines, increasing speed, comfort, fuel economy, and safety. In
2012, the International Snowmobile Manufacturers Association (ISMA; 2013) reported that there
were a total of 1.4 million registered snowmobiles in the United States (U.S.). The economic
impact brought on by the activity is estimated at 22 billion dollars annually in the U.S. The state
of Michigan, with 216,144 registered snowmobiles in 2012, is the third largest snowmobile
market in the U.S. just trailing Minnesota and Wisconsin.
Michigan is one of the few states in the nation that offers a large system of interconnected
snowmobile trails. As of July 2013, there are 6,200 miles of designated snowmobile trail in
Michigan. Most of them are located in the Upper Peninsula and the northern and western Lower
Peninsula because of the availability of public land and deeper snowpack (See Figure 1 for
current Michigan snowmobile trail network). These trails are usually groomed regularly by local
snowmobile clubs to ensure riding safety and enjoyment. These extensive snowmobile trails
penetrate through multiple state forests, three national forests, and many acres of privately
owned lands, which provide abundant opportunity to mingle with nature (Michigan Department
of Natural Resource, 2013). The long-distance snowmobile trail network and abundant snow
precipitation has made Michigan an excellent environment for snowmobiling.

1

Figure 1 Michigan Snowmobile Trail Network
Data source: Michigan DNR
Snowmobiling is a unique outdoor recreation. It relies almost exclusively on natural
snowfall. Riding a snowmobile to view scenery, interact with nature, and experience the
excitement of rapid travel over a snow covered landscape, as well as the opportunity to socialize
with other snowmobilers are the main purposes of such winter outdoor activity. Many other
outdoor enthusiasts, such as anglers, hunters, and boaters, travel to a destination for the activity
and stay within a relatively smaller geographic area. Snowmobilers, because modern
snowmobile can travel long distance at a high speed, typically ride extensive distances on a
linear trail that may be more than 100 miles long and connects with other trails. For a day trip,
snowmobilers on average travel 30 to 75 miles within a favorite riding area or on a trail. For a
touring style overnight trip, the distance of travel normally ranges between 100 and 150 miles
per day (ISMA, 2013). Nelson and his colleagues (2009) studied Michigan snowmobiling trip
2

characteristics and found that 45% of the trips involved overnight stays, 12% were day trips 100
miles and more from home, and 43% were day trips traveling less than 100 miles from home.
According to ISMA (2013), more than half of the snowmobilers (53%) trailer their snowmobiles
to a destination region for riding, while the rest begin their trips from their primary residence or a
vacation home. Although the trail network is unevenly distributed across Michigan and mostly
distant from major metropolitan areas, passionate snowmobilers are still willing to travel long
distances or take an overnight trip to enjoy this unique winter recreation.

In Michigan, all snowmobiles riding on public lands, including the snowmobile trail
system and frozen water, must have a Michigan snowmobile trail permit permanently affixed to
the forward half of the snowmobile directly above or below the headlight, regardless of
registration in Michigan or another state. There were 199,592 snowmobile trail permits sold
during the winter of 2008-09. Those snowmobilers generated a total of 2,046,547 snowmobile
days in Michigan (Nelson, Wu, Stynes, & McCarty, 2009). It is also reported that these
snowmobilers spent an estimated 251 million dollars on trips to Michigan in the 2007-08 winter
season (Stynes, 2009). Snowmobiling is one of the most popular winter outdoor recreation
activities and an integral part of the state’s economy.

However, study of the social aspect of snowmobiling is limited (Lynch, 2000; Nelson et
al., 2009). The theoretical framework of recreational specialization has been applied to study the
within-group social world of many different outdoor recreation activities (Manning, 2011; Scott,
Ditton, Stoll, & Eubanks, 2005). Through a more profound analysis, recreational specialization
reveals how recreationists’ behavioral and psychological involvement changes over time.

3

Recreation planners and managers may be more informed in their managerial decisions through
an increased awareness of the diversity of snowmobilers.

Accompanying the maturity of the industry is concern about snowmobiling risks and the
rising number of accidents. It is reported that more than 200 people are killed and an additional
14,000 people are injured each year in North America due to snowmobile-related accidents
(Pierz, 2003). Because of its interaction with natural terrain and the sensation seeking nature,
snowmobiling has been included in the scope of adventure tourism (Buckley, 2010). Risk, the
major component of adventure tourism, has been well documented in the literature (Buckley,
2012; Ewert, 1989; Weber, 2001). Adventure tourism is therefore often referred to its broader
term, risk recreation. One of the most frequently cited definitions of adventure tourism
encapsulates the characteristics of snowmobile riding:
… often commercialized and involving an interaction with the natural environment away
from the participant’s home range and containing elements of risk; in which the outcome
is influenced by the participant, setting, and management of the touristic experience
(Weiler and Hall, 1992, p. 143)
While snowmobilers look for thrill and enjoyment, outdoor recreation managers aim to
reduce the number and severity of accidents through regulations, law enforcement, education,
and trail design and maintenance.

In Michigan, key laws and rules to reduce accidents and their severity relate to equipment
requirements, operator age limits, riding location, and the specific focuses on avoiding alcohol
use and speeding. Riding tips and regulation brochures are easily accessible online and in each
Department of Natural Resources (DNR) service center. In addition, snowmobile safety
4

education training is required for youth ages 12 to 17 and is recommended for all snowmobile
operators. Nevertheless, the sensation-seeking activity has taken 633 lives (30.1 fatalities
annually) in Michigan between 1992 and 2013 (Figure 2; Michigan DNR, 2008, 2009, 2010,
2011, 2012, and 2013). Although the annual number of fatalities has substantially decreased due
to the intensified enforcement efforts initiated in the late 1990s (Lynch, 2000), limited patrol
cannot fully cover the extensive landscape. The ultimate goal of preventing fatal snowmobile

Number of fatalities

accidents becomes challenging.

50
40
30
20
10
0

47
33

42 40

26 29

33

39 39

46
33

29 26
24

20

24 25 26
13 16

23

Snowmobile Season
Figure 2 Trends of Michigan Snowmobile Fatalities
Data source: Michigan DNR, 2008, 2009, 2010, 2011, 2012, and 2013

To move forward to effectively prevent snowmobile accidents, lessons from research in
motor vehicle accidents are useful. Snowmobile riding is very similar to automobile operation
given that both activities involve critical driving skills such as overtaking, time estimation, and
hazard detection (Deery, 1999; Leung & Starmer, 2005). Especially, when dealing with traffic
accidents, the ability of detecting and recognizing potential traffic hazards plays a key role in
preventing accidents (Deery, 1999). One paradigm of such research has focused on
drivers/riders’ risk perception.

5

Risk perception is defined as “the subjective experience of risk in potential traffic
hazards” (Deery, 1999, p. 226). It is “the process of identifying hazardous objects and events and
quantifying their dangerous potential.” Risk perception is considered a precursor of actual
driving behavior (Machin & Sankey, 2008). When a driver considers an object or event is
dangerous enough to jeopardize his/her wellbeing, the driver perceives a certain level of risk.
Once the driver perceives a high level of risk, he/she may take action to avoid the potential
hazard. Hatfield and Fernandes (2009) concluded that a high level of perceived risk is therefore
associated with a lower chance of having an accident. Conversely, if one does not believe a risky
behavior is harmful, he/she is likely to engage in the risky driving behavior, which increases the
probability of accident occurrence. With this in mind, exploring snowmobilers’ risk perceptions
will help further knowledge of how they make judgments when encountering potential hazards
and eventually achieve the purpose of accident reduction and prevention.

Problem Statement
Previous research viewed snowmobilers as one homogeneous group when studying their
personal characteristics, snowmobile behaviors, and attitudes toward management and law
enforcement initiatives (Lynch, 2000; Nelson et al., 2009; Stewart & Black, 2004). However,
these studies also revealed that snowmobilers are heterogeneous, given that a range of experience
and behavioral involvement were exhibited. An in-depth analysis using recreational
specialization is warranted to further not only the academic understanding of these outdoor
enthusiasts, but also to provide pragmatic knowledge to inform policy and managerial decisions.

Further, though previous studies have examined the relationships between recreational
specialization and various attitudinal variables (Dyck, Schneider, Thompson, & Virden, 2003;

6

Smith, Burr, & Reiter, 2010), the perception of risk has not been assessed using recreational
specialization as an independent variable. While prior snowmobile accident studies concluded
that young snowmobilers were the predominant victims, research also suggested that those cited
for snowmobile law violations snowmobiled more days and traveled more miles than those not
cited for violations (Lynch, 2000). Whether experienced or novice snowmobilers tend to
underestimate risk is unclear. Since the construct of recreational specialization reflects
recreationalists’ experience and skill level, it may provide a complex but solid measurement tool
in explaining different perceptions of risk.

Finally, while it may be true that recreational specialization provides rational explanation
for within-group diversity, what happens during the developing process of recreational
specialization remains unclear. In the last decade, recreational specialization research has shifted
its focus from how participants’ behaviors and attitudes vary among specialization levels to the
process and development of recreational specialization (Kuentzel & Heberlein, 2006; Oh, Sorice,
& Ditton, 2010; Scott & Lee, 2010). The focus on the behavioral and psychological change in
the specialization process ignores the external factor of how resource availability and
accessibility influence leisure behavior. The context within which recreational activity takes
place has not been taken into consideration (Giles-Corti & Donovan, 2002). This is especially
true when abundant snowfalls and available lands are critical elements to snowmobile riding.
Furthermore, since Michigan’s snowmobile trail network is unevenly distributed across the state,
each snowmobiler’s accessibility to snowmobiling opportunities differs depending on where they
live and how far they are willing to travel. Both resource availability and accessibility should be
integrated into one model for the purpose of investigating the effect of these external factors on
the development of snowmobilers’ recreational specialization. By measuring the spatial
7

interaction between origins and destination, thereby taking resource availability and accessibility
into account, a gravity model can be utilized to study how external environmental factors affect
leisure behavior. No previous research has examined the relationship between recreational
specialization and recreationists’ spatial interaction with recreation opportunity. To effectively
address this knowledge deficit, a spatial analytic method using Geographic Information System
(GIS) is applied. With this spatial technology, the process and development of recreational
specialization can be further investigated.

Research Objectives
There are three research objectives of this study.
1. Describe and compare recreational specialization among Michigan snowmobilers.
2. Examine the influence of recreational specialization on Michigan snowmobilers’
perception of risk.
3. Assess the relationships between Michigan snowmobilers’ recreational specialization and
their spatial interaction with Michigan’s snowmobile trail network.

Research Questions
To address the three research objectives, this study asks:
1. What are the behavior, skill/knowledge, and psychological attachment characteristics of
snowmobilers in different levels of specialization?
2. What are the relationships between levels of recreational specialization and
snowmobilers’ perception of risk?
3. What are the relationships between levels of recreational specialization and
snowmobilers’ spatial interaction with the snowmobile trail network?
8

Conceptual Framework and Null Hypotheses
Recreational specialization, perception of risk, and spatial interaction are three concepts
studied in this research. A conceptual framework of this research which utilized Michigan
snowmobiler as research subjects to examine the interaction between three concepts is illustrated
in Figure 3.

Figure 3 Conceptual Framework
Recreational specialization has not been used to empirically examine snowmobilers. The
use of this theoretical framework to represent different dimensions of participation provides
potential for an improved assessment of snowmobiling involvement. Previous specialization
research has demonstrated distinct behavioral, skill/knowledge, and psychological attachment
characteristics between each sub-group for various outdoor activities. Similar results are
conceivable for Michigan snowmobilers. Summative indices and cluster analysis are two
prevailing methods in classifying levels of specialization in a recreational activity. A limited
number of studies have applied both methods to examine a single group of recreationists (Scott
et al., 2005). Whether summative indices or cluster analysis is an appropriate segmentation

9

approach for Michigan snowmobilers is unclear. Therefore, to compare recreational
specialization among Michigan snowmobilers, this research proposes the null hypothesis:

H1: There is no difference in group composition between levels of specialization using
summative indices and cluster analysis.

Further, previous literature has confirmed the relationships between recreational
specialization and several attitudinal aspects of recreation participation, such as attitude toward
managerial actions and perception of pro-environmental programs. While the traffic literature
has concluded that young and novice drivers tend to underestimate risk and are most likely to be
in accidents, a Michigan snowmobile study found that those who violate snowmobile laws tend
to ride significantly more miles and spend more money on the activity than average
snowmobilers (Lynch, 2000). Lynch’s study also suggested that those snowmobile law violators
perceived lower risk on multiple potential hazardous stimuli such as speed of snowmobiles,
drinking while snowmobiling, snowmobiling on roadways, and the presence of other vehicles on
seasonal roads. Therefore, it is foreseeable that a certain relationship exists between
snowmobilers’ recreational specialization and their perception of risk. Lynch’s (2000) list of nine
snowmobile riding scenarios was adopted for measuring snowmobilers’ perception of risk. The
nine item list includes: operation of snowmobile by person who has been drinking but is not
legally intoxicated (0.01-0.07 blood alcohol); operation of snowmobile by a legally intoxicated
person (0.08 or higher blood alcohol); speed of snowmobile; driver lacking skill in operating
machine; cars/trucks on seasonal roads; public trail conditions; public trail design; other uses of
designated snowmobile trails (e.g., dog sledding, cross country skiing); and snowmobiling on

10

county/state road shoulders. Since the relationship between recreational specialization and the
perception of risk is unclear, nine null hypotheses are proposed:

H2: There is no relationship between snowmobilers’ recreational specialization and their
perception of operating a snowmobile by person who has been drinking but is not
legally intoxicated (0.01-0.07 blood alcohol).
H3: There is no relationship between snowmobilers ’recreational specialization and their
perception of operating a snowmobile by a legally intoxicated person (0.08 or higher
blood alcohol).
H4: There is no relationship between snowmobilers ’recreational specialization and their
perception of speed of snowmobile.
H5: There is no relationship between snowmobilers ’recreational specialization and their
perception of driver lacking skill in operating machine.
H6: There is no relationship between snowmobilers ’recreational specialization and their
perception of cars/trucks on seasonal roads.
H7: There is no relationship between snowmobilers ’recreational specialization and their
perception of public trail conditions.
H8: There is no relationship between snowmobilers ’recreational specialization and their
perception of public trail design.

11

H9: There is no relationship between snowmobilers ’recreational specialization and their
perception of other uses of designated snowmobile trails (e.g., dog sledding, cross
country skiing).
H10: There is no relationship between snowmobilers ’recreational specialization and their
perception of snowmobiling on county/state road shoulders.

Research has suggested that leisure participation is substantially affected by the
availability of recreation opportunities (Fesenmaier & Lieber, 1985). Improved access to a
recreation area results in increased visitation and an expansion of the market area. Distance
decay literature has also shown that the degree of spatial interaction (movement between
residential origin and destination) is inversely related to distance. Since the level of recreational
specialization measured is heavily based on recreationists’ leisure participation, the level of
specialization of each snowmobiler is assumed to be correlated to the spatial interaction with
snowmobile trail network. Consequently, to assess the relationships between snowmobile
specialization and their spatial interaction with the snowmobile trail network, this research
proposes the null hypothesis:

H11: There is no relationship between levels of recreational specialization and
snowmobilers’ spatial interaction with the snowmobile trail network.

Research Significance
In studying Michigan snowmobiling, the research significance is three-fold:
First, applying recreational specialization theory to categorize snowmobilers provides indepth insights into effective management practices from both social and behavioral aspects.
12

Through this study of behavior, skill/knowledge, and psychological attachment characteristics,
recreation program managers can discover new user trends, implement marketing plans, develop
policies and managerial strategies, and bring in potential revenue. From a research perspective,
comparing and contrasting two commonly used statistical grouping methods, summative indices
and cluster analysis, not only contributes to the advancement of recreational specialization theory
but also enhances the validity and robustness of analysis. Most previous research often employed
either summative indices or cluster analysis seemingly. This research tests both methods and
selects the appropriate approach to best represent the characteristics of Michigan snowmobilers,
and applies the method to answer research questions.
Second, from the lens of snowmobilers’ psychological recognition of potential traffic
hazards, this research aims to provide evidence to help reduce the incidence and severity of
snowmobile accidents. By identifying those snowmobilers who tend to underestimate
snowmobiling risks, Michigan DNR, regional managers, and law enforcement can more
effectively implement appropriate intervention programs such as rider improvement courses or
workshops, riding safety campaigns such as The Safe Rider!, and rules and regulations that target
aberrant riding behaviors. From the research standpoint, linking the theory of recreational
specialization to traffic research and safety science provides a meaningful taxonomic approach
aside from traditional demographic segmentation. Examination of the relationships between
recreational specialization and perception of risk expands the literature on how recreational
specialization influences characteristics and cognitions of recreationists.
Finally, by exploring the relationship between recreational specialization and
snowmobilers’ spatial interaction with the snowmobile trail system, this research potentially
explains how recreation opportunity affects leisure participation. By investigating the impact of
13

available and accessible snow and designated trails on snowmobilers’ level of specialization,
knowledge of how recreationists develop their specialization can be enhanced. In addition, using
geospatial technology as a research tool provides alternative view from a geographic dimension
in studying human leisure behavior. Building upon few other studies using GIS, this research not
only offers an example of applying geospatial statistics to leisure science, but also examines the
applicability of the gravity model in measuring spatial interaction between recreation demand
and supply. This model may shed light on developing spatial interaction models for other
recreation activities, as well as a more complex model in measuring spatial interaction.

14

Chapter Two
Literature Review

Recreational Specialization
Recreational specialization is the core theoretical framework for this research. A
literature review of its original concept, its relationships to multiple behavioral and attitudinal
constructs, and measurement is presented below.
Original Concept
Recreational specialization, a theory that was first introduced by Hobson Bryan in 1977,
describes the variation of recreationists from the inexperienced to the specialized within a
specific outdoor activity. The level of specialization of these recreationists is generally
symbolized by their participation history, equipment and skills used, and preferences for
recreation setting. As these recreationists move into a more specialized stage over time, their
behavior and attitudes change accordingly. Through on-site interviews and observations, Bryan
identified four subgroups on the continuum of trout fishermen and discovered that each subgroup
differs in attitudes and values about the activity, preferences toward resources and management
policy, as well as their social belongings, vacation patterns, and life styles (Figure 4).
Recreational specialization is thus suggested to have a ‘predictive utility’ for other attitudinal and
behavioral aspects of recreation participation and can be applied to a variety of other leisure
activities (Bryan, 1977; Bryan, 2000).
Recreational specialization was originally developed to account for the heterogeneity of
trout anglers within whom Bryan observed distinctly diverse beliefs and norms reflected by their
intensity of commitment to the activity (Bryan, 2000). Since its initiation, recreational

15

specialization has been utilized as a taxonomic approach to segment various outdoor recreation
participants. Multiple subgroups of specialization were successfully identified for anglers (Oh &
Ditton, 2006), boaters (Kuentzel & Heberlein, 2006), bird watchers (Lee & Scott, 2006), hikers
(Shafer & Hammitt, 1995), hunters (Needham, Vaske, Donnelly, & Manfredo, 2007), campers
(McFarlane, 2004), and other outdoor recreationists. For adventure recreation, most of the
specialization literature focused on whitewater rafting (Bricker & Kerstetter, 2000; Kuentzel &
McDonald, 1992), others studied rock climbing (Ewert & Hollenhorst, 1994), mountaineering
(Dyck, et al., 2003), surfing (Nourbakhsh, 2008), skiing and snowboarding (Needham & Little,
2013), off-highway vehicle (OHV) riding (Smith et al, 2010) and SCUBA diving (Thapa, Graefe,
& Meyer, 2006). None, to the author’s knowledge, has ever empirically applied recreational
specialization to study snowmobilers. Given that snowmobiling involves riding skill
development, significant economic investment, travel time, free time away from work, and club
membership, it is conceivable that subgroups of snowmobilers can be identified using the
recreational specialization construct. Results will aid in the further understanding of other
behavioral and attitudinal aspects of snowmobiling.

Figure 4 Recreational Specialization Continuum
Adapted from “Commitment and Leisure Behavior: A Theoretical Perspective,” by Buchanan,
T., 1985, Leisure Science, 7 (4), p.406.

16

Recreational Specialization and Behavior, Attitude, and Perception
Research has empirically proven that recreational specialization is related to various
behavioral aspects of recreation participation, which includes participants’ setting preferences,
flow experience, information source used, and knowledge learned. Virden and Schreyer (1988)
found that recreational specialization is significantly related to the preference of environmental
setting attributes among backcountry hikers. Similar conclusions were drawn from studies on
vehicle-based campers (McFarlane, 2004), bird watchers (Cole & Scott, 1999; Scott & Thigpen,
2003), and SCUBA divers (Thapa et al., 2006). Serious birders preferred natural bird habitats
rather than developed settings, while casual wildlife watchers favored various opportunities for
observing wildlife (Cole & Scott, 1999). Meanwhile, as the level of specialization increased,
SCUBA divers were less likely to engage in environmentally harmful behaviors and more likely
to participate in pro-environmental and educational activities (Thapa et al., 2006). Recreational
specialization is also proven to be an effective predictor of flow experience. More specialized
mountain hikers were more likely to experience flow than were less specialized mountain hikers
(Wöran & Arnberger, 2012). Additionally, a positive relationship was found between
recreational specialization and the information used for making trip decisions and the knowledge
learned about the activity involved. As level of specialization increased, the quantity of
information sources used to make trip decisions increased simultaneously (Cole & Scott, 1999;
Ditton, Loomis, & Choi, 1992). Highly specialized skiers and snowboarders were significantly
more knowledgeable about an environmental stewardship program at the ski resort than were
their counterparts (Needham & Little, 2013). Because of the evidence of behavioral changes as
recreationists move from general to specialized, managerial strategies and service were
suggested to meet the diverse needs of different subgroups.

17

Other than behavioral changes, recreational specialization was also found to be associated
with various attitudinal aspects of recreation participation. Despite few exceptions (Salz &
Loomis, 2005; Smith et al., 2010), findings from the substantial literature generally support
Bryan’s argument that recreational specialization is positively associated with recreationists’
attitude toward environmental restriction and resource protection policies. For example,
specialized mountaineers valued the importance of low-impact practices significantly more than
less specialized mountaineers (Dyck et al., 2003). Advanced anglers displayed greater
appreciation of managerial harvest restriction and were less willing to compromise, while casual
anglers preferred loose regulation so they were able to catch more fish (Oh & Ditton, 2006). The
pro-environmental attitude also reflected on recreationists’ expectation of rewards. As the level
of specialization increases, the experience itself becomes more important relative to a physical
trophy. More committed river users treasured guaranteed paddling opportunities rather than
physical, social and mental outcomes (Kuentzel & McDonald, 1992). Ditton, Loomis, and Choi
(1992) also found that it is more important for advanced anglers to enjoy the fishing experience
rather than catch lots of fish as the ultimate prize.
Besides environmental concerns, the variation of recreational specialization was found
crucially related to the perception of general management actions and other recreationists’
behavior. McIntyre and Pigram (1992) studied vehicle-based campers’ comments on camp site
character and quality of management intervention and distinguished between different
specialization levels. The most experienced campers were those with the most critical comments
toward facility maintenance and more rigorously resisted policy changes such as an introduction
of fees. Recreational specialization also explained how recreationists perceived others’
depreciative behaviors. As the level of specialization increased, hikers reported less tolerance for
18

seeing others, seeing motorized recreationists, and hearing loud recreationists (Virden &
Schreyer, 1988). Veteran goose hunters were also more likely to be distracted by other hunters
with bad manners than were novice hunters (Kuentzel & Heberlein, 1992). The association
between recreational specialization and paddlers’ perception of crowding was also tested but no
significant linkage found (Kuentzel & McDonald, 1992). Although the existing literature has
discovered certain levels of association between the level of recreational specialization and
recreationists’ various perceptions and attitudes, there has been no published work on the
relationship between recreational specialization and perception of risk in snowmobiling.
Recreational specialization was also found to be positively associated with place
attachment. Although the overall influence of the level of specialization on place attachment was
weak among whitewater river users, specialists were more likely to agree with the importance of
place identity and lifestyle component of place attachment than novice river users, while place
dependence was not influenced by the level of specialization (Bricker & Kerstetter, 2000). Oh,
Lyu, and Hammitt (2012) found that the skill-and-knowledge and commitment dimensions of
recreational specialization were positively associated with place identity among fishing license
holders. Needham and Little (2013) also found a significant positive correlation between
specialization and place attachment among skiers and snowboarders.
Recreational Specialization Measurements
In the past 35 years, studies have emphasized developing a recreational specialization
index to represent within group variation. Although little consensus exists on the specific
indicators and their dimensions, researchers generally agreed a multi-dimensional model of
behavior, skill/knowledge, and psychological attachment to the activity to measure recreational
specialization (Lee & Scott, 2004; McIntyre & Pigram, 1992; Scott & Shafer, 2001). Early
19

literature focused on using behavioral and attitudinal aspects of participation to measure
specialization. Two dimensions appeared in Bryan’s (1977) original construct, behavior reflected
by equipment and skills used, and attitude governed by beliefs and preferences. Donnelly, Vaske,
and Graefe (1986) followed Bryan’s construct, developing a specialization index using four
dimensions to study paddlers’ behaviors: previous participatory experience, equipment owned,
skill perceived, and boating-related interests. While critics considered Donnelly and his
colleague’s measurement “solely in terms of behavior” (Scott & Shafer, 2001, p. 325), a close
review of the variables revealed evidence of measurement from skill, knowledge, and
psychological attachment dimensions. McIntyre and Pigram (1992) later adapted Little’s (1976)
three-dimensional specialization loop and proposed a recreational specialization model that
includes behavioral, affective, and cognitive measurements, which coincided with the threedimensional approach proposed by Scott and Shafer (2001). The three-dimensional measurement
has been widely applied for measuring recreational specialization. The behavioral dimension
usually includes, but is not limited to, indicators such as frequency of participation, equipment
used/owned, and financial investment (Bryan, 1977; Jett, Thapa, & Ko, 2009; Needham et al.,
2007). The skill/knowledge dimension consists of self-reported skill level and knowledge about
the activity (Lee & Scott, 2006; Oh et al., 2012). Centrality-to-life and membership were
commonly seen indicators of the psychological attachment dimension (McIntyre & Pigram,
1992; Scott, et al., 2005; Thapa et al., 2006).
Disparities on how to classify levels of specialization exist. Summative indices and
cluster analysis are two major approaches for segmenting recreationists (Scott et al., 2005).
Bryan defined recreational specialization as “a continuum of behavior from the general to the
particular” (Bryan, 1977, p. 175); therefore, early specialization studies tended to treat each
20

measurement as one factor of the entire index (Donnelly et al., 1986; Dyck, et al., 2003;
Kuentzel & McDonald, 1992). All measurements in the index were standardized if multiple
levels of measurement (nominal, ordinal, interval and ratio) were involved. Then the aggregated
score was deemed continuous and utilized to define the location of a respondent on the
specialization continuum for the given activity. Research samples were often evenly divided into
two, three, four or more sub-groups thereafter as representing different levels of specialization.
The disadvantage of this approach is that it assumes all measurements progress simultaneously
and the most specialized recreationists demonstrate the most commitment in all aspects of
behavior, skill/knowledge, and psychological attachment. However, research has since found that
measurements in different dimensions did not always increase linearly in the same direction
(Kuentzel & McDonald, 1992, Needham et al., 2007; Scott, Baker, & Kim, 1999; Scott &
Thigpen, 2003). Scott and Shafer (2001) argued that some recreationists “continue to participate
in activities on a regular basis and accrue commitments but exhibit little evidence of skill
development… other individuals may participate in leisure activities infrequently but
demonstrate a high level of skill development and personal commitment” (p. 338). Other
researchers turned to cluster analysis for identifying subgroups in the activity (McIntyre &
Pigram, 1992; Needham et al., 2007; Scott & Thigpen, 2003), given that cluster analysis
provides a multivariate classification with more objective measurement then summative indices.
Cluster analysis is a statistical technique which allows the analyst to search for
conceptually meaningful sub-groups where each member in the sub-group shares common
characteristics (Pang-Ning, Steinbach, & Kumar, 2006). Cluster analysis does not assume each
dimension of specialization co-varies and therefore was suggested to be less biased and more
appropriate for segmenting recreationists (McFarlane, 1996; Needham et al., 2007; Scott et al.,
21

2005). However, one drawback of this method is that those identified sub-groups are treated as
categorical data and do not necessarily follow the “continuum” proposed by Bryan (1977). In
addition, the number of sub-groups is often arbitrarily decided either by the researcher’s
subjective judgment or following previous studies for the purposes of comparison. Oh and Ditton
(2006) suggested that there is simply not a “sacred” number of sub-groups that would fit all
recreational specialization studies.
Despite support from both summative indices and cluster, the literature has shown no
superiority between the two aforementioned approaches. Scott and his colleagues (2005)
examined the efficacy of recreational specialization in predicting birders’ generic motivation
using both approaches and found little difference between them. Since both summative indices
and cluster analysis have been applied to investigate the relationship between recreational
specialization and multiple attitudinal preferences (Needham et al., 2007; Smith et al., 2010), this
research will use both methods to segment snowmobiling specialization and compare the results
for further analyses.

Perception of Risk
Two pieces of literature were reviewed in this section: the concept and measurement of
risk perception, and snowmobile fatalities in Michigan and other areas.
Concept and Measurement
Risk perception, or perceived risk, is the cognitive process of identifying and quantifying
potentially harmful objects or events. This concept has been applied mostly in traffic and
accident prevention studies to understand risky driving behaviors, defined as “the subjective
experience of risk in potential traffic hazards” (Deery, 1999, p. 226). As shown in Figure 5,

22

previous research concluded that a higher perception of risk (i.e., considering the behavior is
highly dangerous) is strongly associated with a lower tendency of engaging in risky behaviors
(Curry & Youngblade, 2006; Hatfield & Fernandes, 2009; Machin & Sankey 2008). To prevent
and reduce the incidence and severity of snowmobiling accidents, studying operators’
perceptions of riding risks provides valuable insights in developing intervention programs,
education, and management actions.

Figure 5 Correlations between Risk Behavior and Risk Perception
Adapted from “Negative affect, risk perception, and adolescent risk behavior,” by Curry, L. A.
and Youngblade, L. M., 2006, Journal of Applied Developmental Psychology, 27, p.479.
Perception of risk is the attitudinal interpretation of potential hazardous situations or
stimulus while snowmobiling. To measure perception of risk, many studies measured the
perceived danger of several selected driving actions. For instance, eight risky driving actions
(e.g., running a red light, riding a moped without wearing a helmet, etc.) were incorporated by
Machin and Sankey (2008) to assess perception of risk. Another eight driving related events such
as the likelihood of ‘be fined for speeding’ and ‘have a crash due to drink-driving’ were used by
Hatfield and Fernandes (2009) to measure the perceived risks. Curry and Youngblade (2006)
23

measured adolescents’ perceptions of risk utilizing a list of six hazardous behaviors such as
‘riding in a car with a drunk driver’ and ‘drinking alcohol.’ In the above studies, Likert scales
(five or seven points) were adapted in the questionnaires to have respondents rate their
perceptions toward the potential hazardous behaviors or events encountered while operating a
vehicle. To measure snowmobilers’ perception of riding risks, events or conditions that could
threaten a snowmobiler’s safety should be investigated.
Snowmobile Accidents and Fatalities
For a better understanding of snowmobile accidents and in searching for measurement of
snowmobilers’ perception of risks, six years of Michigan snowmobile fatality summaries from
2007-08 to 2012-13 were reviewed (Michigan DNR, 2008; 2009; 2010; 2011; 2012; 2013). The
fatality summary includes key information on accidents such as victim gender and age, location,
cause, possible factors, and a brief description of the accidents. This information provides a
profile in identifying potentially hazardous situations or stimulus while snowmobiling.
There were a total of 122 snowmobile accidents resulting in 128 fatalities (21 annually)
in the previous six years. This is a decline compared to a total of 319 fatalities (40 annually) for
the snow seasons of 1995-96 to 2002-03, due to intensified law enforcement activities such as
increased visibility and active inspection (Michigan DNR, 2008; Nelson et al., 2009). Of those
128 fatalities, 102 (80%) were males and 26 (20%) were female. By age, 6% were under 20
years old, 34% were 20 to 39 years old, 49% were 40 to 59 years old, and 11% were 60 years
and older. Male snowmobilers between 40 and 59 were the most common snowmobiling fatality
victims in Michigan. This is in contrast to several retrospective studies in Minnesota (Beilman et
al., 1999), Sweden (Öström & Eriksson, 2002), and Canada (Rowe, Milner, Johnson, & Bota,
1992; Stewart & Black, 2004). These studies concluded that young males in their thirties were
24

the predominate victims of snowmobile accidents. While the age and gender profile provided an
important indication of recognizing those who tend to be involved in snowmobile accidents, it is
important to note that there are snowmobilers who do not start snowmobiling until a later stage
of their lives. The probability of having a snowmobile accident should be more directly related to
the psychological recognition of risk behavior, as well as the familiarity with and the skill of
operating and controlling a snowmobile.
The importance of skills and the control of a snowmobile were demonstrated in the
fatality reports. Operators losing control of their snowmobiles accounted for at least 77% of the
fatal accidents in the past six years, posing major concerns for drivers lacking skills of operating
snowmobiles. Failure to negotiate a curve in the road or trail, collision with a snow bank,
ejection from the snowmobile, and impact with fixed objects such as trees and steel cables were
common scenarios illustrated in the description of accidents (Michigan DNR, 2008; 2009; 2010;
2011; 2012; 2013). Previous studies also suggest that inexperienced riders and unfamiliarity with
the environment were important risk factors contributing to snowmobiling accidents (Pierz,
2003). Michigan addresses this concern by providing a safety training course. This management
effort mandates that adolescents aged 16 or younger complete the training and earn a safety
certification. In addition, this program is recommended but not required for those over 16.
The location of snowmobile fatalities offers additional information in understanding and
preventing fatalities. Michigan snowmobilers are allowed to ride on the designated trail system
and the right-of-way or shoulders of a public highway in a state/county road system. However,
the majority of fatal accidents (43%) occurred on state, county, or seasonal roads; followed by
38% that were on a designated snowmobile trail. Riding on the road shoulders clearly poses
challenges since there are other vehicles using the road system at the same time. There were a
25

number of snowmobiling fatalities caused by the impact of snowmobiles with vehicles on the
road system (Michigan DNR, 2011; 2012; 2013). The design of snowmobile trails and trail
grooming conditions also deserve attention when searching for factors contributing to
snowmobiling fatalities.
Alcohol use and excessive speed have been the major factors directly resulting in
snowmobiling accidents. In the past six years, alcohol use was a factor in 45% of fatalities, while
excessive speed was involved in 32% of the fatal accidents (Michigan DNR, 2008; 2009; 2010;
2011; 2012; 2013). Another 10 year Michigan snowmobile fatality trend report showed that
alcohol/drug use was a factor in 60% of fatalities, while “speed too fast” was involved in 72% of
crashes (Michigan DNR, 2003, p.6). Alcohol consumption was also reported to be associated
with 44% of snowmobile injuries in Minnesota (Beilman et al., 1999), 70% in Canada (Rowe et
al., 1992; Stewart & Black, 2004), and 64% in Sweden (Öström & Eriksson, 2002). With the
development of more powerful engines, excessive speed was associated with more than half of
snowmobile accidents in Swedan and Canada (Öström & Eriksson, 2002; Stewart & Black,
2004). Local snowmobile clubs and organizations have promoted a “Zero Tolerance” campaign
to prevent alcohol use while operating a snowmobile since the early 2000s. In terms of speed,
modern snowmobiles are capable of reaching a speed of 90 miles per hour or greater, much
faster than the speed limit of 50 miles per hour regulated within the states of Minnesota and
Wisconsin. Michigan requires snowmobilers to follow various speed limits posted along the
designated trail system. Alcohol use and the speed of snowmobile certainly are two key foci of
prevention and intervention programs.
Snowmobilers’ riding skill, location of accidents, speeding, and alcohol use were the risk
factors directly associated with Michigan snowmobile fatalities, according to the review of
26

fatality summaries. A list of nine snowmobiling riding scenarios that was used by Lynch (2000)
for assessing snowmobilers’ perception of risks covered the factors identified in the summaries.
The nine item list includes: operation of snowmobile by person who has been drinking but is not
legally intoxicated (0.01-0.07 blood alcohol); operation of snowmobile by a legally intoxicated
person (0.08 or higher blood alcohol); speed of snowmobile; driver lacking skill in operating
machine; cars/trucks on seasonal roads; public trail conditions; public trail design; other uses of
designated snowmobile trails (e.g., dog sledding, cross country skiing); and snowmobiling on
county/state road shoulders. Given its comprehensive nature, this list is considered appropriate
for measuring Michigan snowmobilers’ perceptions of risk.

Spatial Interaction
Three pieces of the spatial interaction literature were reviewed in this section: the general
construct of spatial interaction, its application to different fields, and the calibration and
measurement of the spatial interaction between Michigan snowmobilers and the trail network.
General Construct
Spatial interaction is a generalized representation of flows of activity, such as the
movement of people, capital, freight, or information, between locations. Three basic conditions
are necessary for the realization of a spatial interaction to occur: complementarity, no intervening
opportunity, and transferability (Rodrigue, Comtois, & Slack, 2013). Complementarity describes
the supply and demand relationships between origin and destination. Spatial interaction occurs
when snowmobilers (demand) travel from their origins to a destination for services provided by
DNR’s snowmobile trail system (supply). No intervening opportunity means there are no
alternative locations of supply or demand to hinder the spatial interaction between current origin
and destination. If Michigan snowmobilers chose to travel to Wisconsin instead of Michigan for
27

snowmobiling, there will not be a spatial interaction between snowmobilers and Michigan’s
snowmobile trail network. Finally, transferability is the linkage between origin and destination
where the flows of activity actually occur. For Michigan snowmobilers, transferability is
represented by the transportation infrastructure between their residential origin and destination.
The highway or road system represents the transferability of spatial interaction. To quantify
spatial interaction, research has been done seeking approaches to measure and model spatial
interaction.
The theoretical basis for modeling spatial interaction in geographical spaces is Newton’s
laws of motion. Newton’s second law of motion provides an explanation for the force existing
between two bodies and how the attributes of two bodies contribute to the force. Essentially, the
force between two bodies is calculated as the product of two masses divided by the square of the
distance between them (Batty, 2007). Early social physicists believed that this physical law can
be analogized to spatial interaction models for explaining social phenomena (Witt & Witt, 1995).
The gravity model is one such widely accepted model for quantifying spatial interaction.
One of the early attempts in developing the gravity model was by Isard and Bramhall in
the 1960s. The basic assumption of the gravity model is that ‘the interaction between any two
populations can be expected to be directly related to their size; and … inversely related to
distance’ (as cited in Sen & Smith, 1995, p.1). Following this intuitive assumption, many
researchers proposed their own versions of the gravity model by replacing population size with
other parameters of the location, or replacing distance with other means of friction such as travel
cost or travel time between origin and destination (e.g., Anderson, 1979; Baxton, 1979; Cheung,
1972). To keep the most flexibility and include different factors that affect the measure of the

28

force between locations, a more general expression of this gravity model was proposed by many
authors (e.g., Hua & Porell, 1979; Sen and Sööt, 1981):
𝑇𝑖𝑗 = 𝐴(𝑖)𝐵(𝑗)𝐹(𝑑𝑖𝑗 )

Where 𝑇𝑖𝑗 is the spatial interaction between origin 𝑖 and destination 𝑗, 𝐴(𝑖) is the

unspecified parameter associated with the origin 𝑖, 𝐵(𝑗) is the unspecified parameter associated

with the destination 𝑗, and 𝐹(𝑑𝑖𝑗 ) is the unspecified distance deterrence function between origin 𝑖

and destination 𝑗. The magnitude of spatial interaction is essentially based on three factors: the
attributes of the origin, the attributes of the destination, and the friction of separation between

origin and destination (Rodrigue et al., 2013). To assess the spatial interaction between Michigan
snowmobilers and the snowmobile trail network, this research adapted the gravity assumption
and incorporated the attributes of and the spatial friction between origins and destinations that
were associated with Michigan snowmobiling.
Application of the Gravity Model
Given its intuitiveness and flexibility, the gravity model has been utilized to model spatial
accessibility to locations of supply, migration, and international trade in various disciplines. The
gravity model has also been applied in the field of leisure studies. For instance, in assisting
tourism planning, Lee, Choi, Yoo, and Oh (2013) integrated a gravity model into the evaluation
of the spatial tourism interaction between Korean rural villages for assisting managerial decision
making and marketing strategies. Zhang, Lu, and Holt (2011) integrated a gravity model in the
development of a more intuitive population-weighted distance to model the potential spatial
accessibility to parks. Khadaroo and Seetanah (2008) investigated the role of tourism
infrastructure in tourism flows among 28 countries and found that transport infrastructure is a

29

significant determinant of tourism inflows into a destination. Utilizing a more sophisticated
gravity model, Kim and Fesenmaier (1990) examined the effects of spatial structure of recreation
opportunity on recreational travel. A competitive and agglomerative relationship was found
between nearby recreation opportunities. Parks within a 25 mile travel distance appeared to
compete for visitor preference, while the number of nearby parks, when considered a cohesive
unit, positively related to visitation with the increase of distance. All these studies altered the
gravity assumption based on the purposes and nature of their studies and incorporated attributes
and friction associated with the origins and destinations focused on.
Spatial Interaction Measurements
In the review of previous studies, the gravity model was generally utilized for either
modeling or measuring spatial interaction between locations. While the regularity and flexibility
of the gravity model enables wide application in different domains, criticisms of and debates
about the appropriate calibration of destination/origin attributes and friction parameters seldom
agree, even more so when estimating the algebraic forms between parameters (Sen & Smith,
1995). The majority of the gravity model literature therefore focused on modeling spatial
interaction, in other words, the emphasis was on estimating the parameters and the mathematic
functions of different attributes and the friction parameters in searching for the model that fit the
best (Sugiyama, Francis, Middleton, Owen, & Giles-Corti, 2010; Witt & Witt, 1995). Using the
observed flow measurement (e.g., actual visitation or tourism arrivals), the function of
destination/origin and friction parameters, and the algebraic functions were calculated using
linear modeling techniques (e.g., Baxter & Ewing, 1986; Keum, 2010; Khadaroo & Seetanah,
2008). A series of more sophisticated gravity models other than the generalized model such as
the singly constrained model and doubly constrained entropy-maximizing model were suggested
30

(Baxter & Ewing, 1986; Wilson, 1967). Other calculation processes such as the two-step floating
catchment area (2SFCA) method or enhanced 2SFCA (E2SFCA) were proposed to improve the
accuracy of modeling spatial accessibility to health care supplies (Luo & Qi, 2009; Radke & Mu,
2000). These researchers sought a gravity model that could accurately predict the spatial
interaction between locations.
The generalized gravity model was incorporated in many studies for the purpose of
measuring spatial interaction. Despite the uncertainty of calibration and algebraic functions
between parameters, the gravity model was still regularly utilized because it provides a partial
explanation and a starting point for more complex models (Witt & Witt, 1995). Previous studies
using a gravity model to evaluate spatial accessibility provide valuable calibration and algebraic
functions in measuring spatial interaction between Michigan snowmobilers and the snowmobile
trail network. Destination park size and Euclidean distance between origin census block and
destination park were used in the gravity model by Zhang, Lu, and Holt (2011) to assess spatial
accessibility to neighborhood parks. Attractiveness of the destination and shortest road network
distance were utilized by Giles-Corti and Donovan (2002) to assess the environmental
determinants of physical activity in Perth, Australia. In assessing spatial accessibility to health
services, availability of the health care provider and travel impedance were incorporated in the
gravity-based accessibility model (Delamater, Messina, Shortridge, & Grady, 2012; Khan, 1992;
Guagliardo, 2004; Penchansky & Thomas, 1981). Based on the generalized gravity model and its
applications in measuring spatial accessibility, snowmobile trail network availability and
accessibility are important parameters that potentially affect the spatial interaction between
Michigan snowmobilers and the snowmobile trail network. Trail mileage and daily snowfall are
therefore used to represent the availability of the snowmobile trail network. Distance between
31

each snowmobiler’s residential origin and the destination snowmobiler trail network is defined as
the accessibility to snowmobile trail network.
In addition to calibration, the mathematic function of the friction between locations was
also estimated and discussed in the literature. Since the spatial interaction between two locations
is inversely proportional to the square of the distance between them according to Newton’s laws
of motion, the friction parameter is usually placed as the denominator with an exponent of two in
a gravity model. The gravity model is therefore expressed as:

𝑇𝑖𝑗 =

𝐴(𝑖)𝐵(𝑗)
𝛽

𝑑𝑖𝑗

Researchers generally had few arguments on the calibration of friction between two
locations, whether it was defined as distance, travel cost, travel time or other friction means.
However, research suggested that the friction exponent (or decay exponent) β varies depending
on the geographic area (e.g., urban, suburb, and rural) or human activity (e.g., physical activity
versus tourism) (Luo & Wang, 2003; Zhang et al., 2011). A large exponent indicates that the
spatial interaction between two locations is sensitive to distance and the likelihood of movement
between locations declines quickly. A small exponent generates a smoother travel curve so
distance has less effect on spatial interaction (Schuurman, Berube, & Crooks, 2010). Empirical
study of the accessibility to recreational facilities provided evidence that the friction exponent
differed among types of recreational service provider. Using a linear regression model, GilesCorti and Donovan (2002) regressed the log of distance on the log of percentage of opportunities
to examine the friction exponents for nine different recreational facilities within and around the
Perth metropolitan area, Australia. These were: public open space (β =1.91), river (β =1.71),
tennis court (β =1.64), beach (β =1.48), gym/health club (β =1.39), swimming pool (β =1.27),
32

sporting complex (β =1.16), golf course (β =1.06), and other facilities (β =1.03). Results showed
that publically accessible facilities or services had larger friction exponents compared to
membership or fee-based facilities or services. Although riding on the Michigan snowmobile
trail network requires a permit, the yearly cost of $45 is incomparable to that of golf course
membership or a gym pass. This research considered the wide distribution of snowmobile trail
network a public open space, therefore the friction exponent β =1.91was applied to measure the
spatial interaction between Michigan snowmobilers and the snowmobile trail network.

33

Chapter Three
Research Methods

Research Data
Three sets of data were used to address the research questions: daily snow precipitation
for the winter of 2007-08, Michigan snowmobile trail GIS shapefiles, and survey data from the
2008-09 Michigan Snowmobile Use and User Survey.
Snow Precipitation Data

Figure 6 Locations of 82 Weather Stations
Snow precipitation is also known as snowfall. Daily snowfall data for winter 2007-08
from the National Climatic Data Center Summary of the Day Series were purchased from the
Office of the Michigan State Climatologist ( http://mrcc.isws.illinois.edu/CLIMATE/). The
Michigan State Climatologist and the author came to the agreement that daily snowfall data from
34

October 1st, 2007 to April 30th, 2008 (213 days) best represent the winter days of Michigan.
Daily snowfall data were derived from 82 weather stations including three stations in Ohio, five
stations in Wisconsin, 10 stations in Indiana, and 64 stations in Michigan. Figure 6 illustrates the
location of these weather stations.
Michigan Snowmobile Trail Shapefiles
The Michigan snowmobile trail shapefiles were provided by the Michigan DNR Forest
Resources Division. The Michigan DNR constantly seeks to extend the snowmobile trail system,
therefore the routes and the total mileage of snowmobile trail may change overtime. Data
shapefiles used in this research were received on July 1st, 2013.
Survey Data
Survey research data came from a statewide survey of Michigan snowmobile trail permit
purchasers in the winter of 2007-08 (Nelson et al., 2009). The Michigan snowmobile trail permit
is required to ride on all public lands (including the designated trail system even if it is on private
properties or federal lands) and frozen waters, by both residents and non-residents. There are two
ways to obtain the snowmobile trail permit; snowmobilers can purchase the permit from the
DNR authorized retail sales system (RSS) by swiping a driver’s license at a license terminal or
on-line from the DNR, or purchase a permit from vendors (such as the Michigan Snowmobile
Association or snowmobile equipment dealers), who are not part of the RSS system, by filling
out a paper record. Both permit purchase approaches generate a list of names with their
addresses. Since Michigan requires all snowmobiles operated on public lands or trails as well as
frozen waters to have a snowmobile trail permit affixed to the forward half of the snowmobile,
this research considers the combination of these two lists as the population of legal snowmobilers
using public resources in Michigan in the winter of 2007-08.
35

A total sample of 3,000 names/addresses was selected from the 199,592 Michigan
snowmobile trail permit holders in the winter of 2007-08. To create the sample names, 1,905
(63.5%) were randomly selected from the electronic records and 1,095 (36.5%) were systematically
selected from the paper records with a random start. This is roughly the proportion of trail permits
sold by each means (Stephen Kubisiak, Michigan Department of Natural Resources, pers.
comm.).
The sample of 3,000 was evenly divided into three groups for questionnaire mailing, in
order to compare snowmobiling travel behaviors during different segments of the snowmobiling
season. Questionnaires to the first 1,000 were mailed on January 15th, 2009, the second group was
on February 15, 2009, and the third group was on March 15, 2009. Each subject received the
questionnaire with a cover letter and a postage paid business reply envelope. The cover letter
described the research, how the subject was selected, provided a guarantee of confidentiality,
stressed the voluntary nature of participation, and provided contact information for the principal
investigator. A reminder postcard was delivered if the survey was not returned in two weeks. After
another two weeks, an additional questionnaire, the business reply envelope, and a revised cover
letter that encouraged participation were sent to non-respondents. Any survey returned after May
15th, 2009 was not included in the result. Refer to Appendix for the questionnaire, initial cover
letter, and second mailing cover letter.
Survey Instrument
The survey instrument was designed in part based on a previous Michigan snowmobile
study in 1998 for the purpose of comparison (Nelson, Lynch, & Stynes, 1998). The questionnaire
emphasizes snowmobiling behavior and attitudes toward Michigan snowmobile programs and
contains 34 questions with 202 variables in six pages.
36

The behavioral dimension of the instrument includes items relating to household member
involvement in snowmobiling in the winters of 2007-2008 and 2008-2009; information about each
snowmobile the household owned; previous snowmobile travel behavior to the five different
Michigan regions (western Upper Peninsula, eastern Upper Peninsula, northwestern Lower
Peninsula, northeaster Lower Peninsula, and southern Lower Peninsula) in both winters; type of
trips; information about the most recent Michigan snowmobiling trip; economic involvement in the
previous 12 months; and experience of encountering a law enforcement officer.
Attitudinal dimension items includes overall satisfaction and level of satisfaction toward
nine selective services provided by the Michigan snowmobile program; perception of crowding;
improvement suggestions; support of pertinent policies, willingness to pay for snowmobiling
opportunities; and, perceptions of risk for nine selected snowmobiling behaviors.
In addition to behavioral and attitudinal questions, the survey asked general questions about
each snowmobiler’s lifestyle. These questions included participation in 16 other outdoor activities;
five digit home zip code; current age and age began snowmobiling; and snowmobiling organization
membership.
To address this study’s research questions, only a portion of the survey results were used for
further statistical analysis. The total number of snowmobile machines owned by the household;
economic involvement in the previous 12 months; current age and age began snowmobiling; total
miles driven; total days spent snowmobiling; and snowmobiling organization membership were
used to assess respondents’ recreational specialization. Perceptions of risk for nine selected
snowmobiling behaviors were used to assess respondents’ perceptions of snowmobiling risks. Total
days spent snowmobiling and snowmobile days in five different Michigan regions were used to

37

assess respondents’ spatial interaction with the snowmobile trail network. The full survey can be
found in the Appendix.

Data Analysis
There were 1,128 snowmobilers who completed and returned the survey. Based on their
reported residential zip codes, 853 (76%) were Michigan residents and 275 (24%) were nonresidents. Due to differential state snowmobile registration fees in other states, this study only uses
Michigan respondents, as all of them paid the same Michigan snowmobile registration fee as well as
the same Michigan trail permit fee.
Although the survey asked snowmobilers to report their riding experience in the winters of
both 2007-08 and 2008-09, data for winter 2008-09 was not complete for most respondents as they
responded before the 2008-09 snowmobiling season was completed. Hence, this research used only
the snowmobiling experience data of winter 2007-08 for further analysis.
Measuring Recreational Specialization
Recreational specialization serves as the independent variable of this research. The threedimensional measurement of behavior, skill/knowledge, and psychological attachment, proposed
by Scott and Shafer (2001) was followed to measure snowmobilers’ recreational specialization.
Table 1 summarizes the scale and data type of the three specialization dimensions.

38

Table 1 Dimension, Data Type, and Scale of Snowmobile Specialization Measurement
Variable Name
Data Type
Scale
Behavior
Numbers of snowmobiles owned
Ratio
Number
Money spent on snowmobiling in the last 12 months ($)
Ratio
Number
Snowmobiling experience
Ratio
Percentage
(Current Age – Begin Age) / Current Age
Skill/Knowledge
Total snowmobile days in winter 07-08
Ratio
Number
Total miles driven in winter 07-08
Ratio
Number
Psychological Attachment
Snowmobile organization membership
Ratio
Number
The behavioral dimension includes three variables: numbers of snowmobiles owned,
money spent on snowmobiling in the previous 12 months, and snowmobiling experience.
Amount of equipment owned has proven to be effective in measuring the level of specialization
(Bryan, 1977; Dyck, et al., 2003). The more snowmobiles owned indicates a higher level of
specialization. Further, financial investment in equipment purchase and operation represents
intensity of participation. Although expenditure may be due to conspicuous consumption (Scott
& Shafer, 2001), financial investment has proven to be a valid measure of recreational
specialization (Cole & Scott, 1999; Wöran & Arnberger, 2012). Finally, snowmobile experience
was measured using the proportion of one’s life he/she has snowmobiled. Survey respondents
were asked their current age and the age they began snowmobiling; to avoid age bias,
snowmobiling experience was stated as a percentage and calculated as:
(Current Age – Age Begin Snowmobiling) / Current Age = Snowmobiling Experience
This method has been used to assess specialization among skiers and snowboarders (Needham &
Little, 2013), OHV riders (Smith et al., 2010), and deer hunters (Needham et al., 2007).
The skill/knowledge dimension was measured through snowmobilers’ recent experience
and consisted of two variables: total snowmobile days in winter 2007-08 and total miles driven
39

in the winter of 2007-08. Past experience has been a reliable indicator of specialization (Dyck, et
al., 2003; McFarlane, 1996; Virden and Schreyer, 1988). Lynch (2000) found that convicted
snowmobile law violators rode significantly more days and more miles than other snowmobilers.
Prevailing recreational specialization studies used self-reported skill level (Oh, Ditton, Anderson,
Scott, & Stoll, 2005; Thapa et al., 2006) and knowledge of the activity (i.e., number of bird
species observed over the lifetime for birders; Lee & Scott, 2006; Won, Bang, & Shonk, 2008) to
measure this dimension. Since the survey did not ask respondents to rate their skill level, this
research considered the riding skills cumulated in the past snowmobile season a valid
measurement for skill level. The proficiency of riding snowmobiles is hypothesized to increase
along with riding days and miles. Riding a snowmobile more frequently indicates a higher skill
level.
Snowmobiling organization membership was used to measure the psychological
attachment dimension of snowmobiling specialization. Two questions were asked in the survey:
a dichotomous “Are you a member of a snowmobiling association or club” and an open-ended
question “Please list each organization.” The responses to both questions were recoded into a
new variable “snowmobiling organization membership.” Respondents who reported “No” in the
former question were coded “0” in the new variable. For those who reported “Yes,” the
researcher counted the number of organizations or clubs listed in the open-ended question and
used the resulting number for the new variable. If respondents reported “Yes” for the first
question but did not list any organization in the second question, a “1” was assigned to the
variable. Membership has been an essential indicator of the psychological attachment dimension
of recreational specialization (Kuentzel & McDonald, 1992; Oh et al., 2005; Smith et al., 2010),

40

while more snowmobile club or association memberships indicates an even higher level of
snowmobile specialization.
Identifying Snowmobile Specialization
Both a summative index and cluster analysis were applied to segment snowmobilers’
levels of recreational specialization. The results of these two approaches were compared to
decide which method was more meaningful and appropriate for investigating the social world of
Michigan snowmobilers. Because the recreational specialization variables were in different
measurement ranges and units, each variable was standardized to a mean of zero and a standard
deviation of one in preparation for the analyses that followed. The descriptive statistics,
percentages of the identified subgroups, and between-group differences were used as criteria to
justify which method and subgroups to be applied for examining the research hypotheses.
For the summative index method, this research adopted procedures executed in previous
recreational specialization research (e.g., Dyck, et al., 2003; Smith et al., 2010). All six
standardized scores were summed for each survey respondent to represent his/her specialization
index score. The index score for all research samples formed a continuum of snowmobile
recreational specialization. These samples were then evenly divided into three specialization
subgroups (Expert, Intermediate, and Novice) based on their location on the specialization
continuum. The least specialized 33.3% of the sample was defined as novice snowmobilers, the
middle 33.3 % of the sample was defined as intermediate snowmobilers, and the highest 33.4 %
of the sample was defined as expert snowmobilers. Once the specialization subgroups were
identified, they were compared using the original (non-standardized) six indicators of behavior,
skill/knowledge, and psychological attachment. Descriptive statistics were used to understand
distinct characteristics of the three levels of snowmobile specialization. A between-group
41

analysis of variance (ANOVA) was performed to test the behavior, skill/knowledge, and
psychological attachment characteristic differences between levels of snowmobile specialization
(Hypothesis 1) using IBM SPSS Statistics computer program.
For cluster analysis, a non-hierarchical K-mean cluster analysis was used to determine
specialization clusters with homogeneous behavioral, skill/knowledge, and psychological
attachment characteristics. Hair and his colleagues (1998) suggested that K-mean cluster analysis
is the most efficient approach to segment clusters. K-mean cluster analysis has also proven
effective in generating meaningful subgroups in the recreational specialization literature
(McFarlane, 1994; Needham et al., 2007; Scott et al., 2005). To enable comparison with the
summative approach, this study chose three clusters. Three cases were therefore randomly
selected as the initial, temporary cluster centers, and then updated in an interactive process to
minimize the within-cluster sum of squares (Hartigan & Wong, 1979). The maximum number of
iterations for updating of cluster centers was set to 20. The F test within ANOVA was performed
to test cluster validity and the characteristic differences between clusters of snowmobiling
specialization. Once the three clusters were identified, they were compared using the original
(non-standardized) six indicators of behavior, skill/knowledge, and psychological attachment.
Descriptive statistics were used to understand distinct characteristics between the three clusters
and address hypothesis one.
Measuring Perception of Risk
This research defines the perception of risk as the attitudinal interpretation a snowmobiler
holds when confronted with a potential hazardous situation or stimulus while snowmobiling. A
list of snowmobiling scenarios that was used by Lynch (2000) for assessing snowmobilers’
perception of risky behavior was adopted for this research. The nine item list includes: operation
42

of snowmobile by person who has been drinking but is not legally intoxicated (0.01-0.07 blood
alcohol); operation of snowmobile by a legally intoxicated person (0.08 or higher blood alcohol);
speed of snowmobile; driver lacking skill in operating machine; cars/trucks on seasonal roads;
public trail conditions; public trail design; other uses of designated snowmobile trails (e.g., dog
sledding, cross country skiing); and snowmobiling on county/state road shoulders. A five-point
rating scale was used to measure snowmobilers’ attitudinal interpretation, ranging from one (not
dangerous) to five (extremely dangerous). Higher scores indicate higher perception of risk and
danger. Since the respondents saw both rating numbers (from one to five) and descriptions (from
not dangerous to extremely dangerous) while reporting their perceptions of risk on the survey,
this research treated the rating scale as interval data, meaning the difference between rating one
and two was the same as the difference between rating four and five. Cronbach's alpha was
computed as an internal consistency estimate of reliability (Vaske, 2008). Table 2 displays the
scale and data type of measures of perceptions of risk.
Table 2 Scale and Data Type of the Perception of risks Measurement
Variable Name
Data Type Scale
Operation of snowmobile by person who has been
Interval
1) Not Dangerous
drinking but is not legally intoxicated (0.01-0.07
2) Slightly Dangerous
blood alcohol)
3) Moderately Dangerous
Operation of snowmobile by a legally intoxicated
Interval
4) Highly Dangerous
person (0.08 or higher blood alcohol)
5) Extremely Dangerous
Speed of snowmobile
Interval
Driver lacking skill in operating machine
Interval
Cars/trucks on seasonal roads
Interval
Public trail conditions
Interval
Public trail design
Interval
Other uses of designated snowmobile trails (e.g., dog Interval
sledding, cross country skiing)
Snowmobiling on county/state road shoulders
Interval
Examining the Influence of Specialization on the Perception of Risk

43

The relationship between levels of snowmobiling specialization and nine perceptions of
risks was tested with an effect size index Eta (𝜂). Eta measures the association of a continuous
level dependent variable and a categorical independent variable. An Eta value between .10 and
.242 represents a weak relationship between two variables, a value between .243 and 370
represents a moderate relationship, and a value of .371 or greater represents a strong relationship
(Vaske, 2008). In addition, Eta squared (𝜂2 ) was assessed to interpret the amount of variance in
the dependent variable that was explained by the independent variable. Correlations between
each of nine selected perception of risks and recreational specialization were used to address
hypotheses two to ten.
In addition to the effect size index Eta, the analysis of variance was applied to address the
perception of risk difference across three specialization sub-groups. The multivariate analysis of
variance (MANOVA) was applied to address the overall sub-group difference on nine perception
of risk. Instead of performing a one-way ANOVA nine times, MANOVA is known for
examining the main and interaction effects of several correlated dependent variables (Vaske,
2008). MANOVA is sensitive not only to mean differences but also to the direction and size of
correlations among the dependent variables. Therefore, MANOVA increases the chance of
finding a group difference compared to ANOVA. A multivariate F value (Wilks’ Λ) was
obtained to test the significance of group difference. Similar to ANOVA, a Post Hoc Tukey’s
Honesty Significant Difference test was performed for testing individual differences between
specific subgroups (i.e., Expert and Novice, Novice and Intermediate, and Intermediate and
Expert).

44

Measuring Spatial Interaction
Since the literature related to snowmobiling or recreational specialization lacked studies
focused on spatial interaction, this research borrowed the measuring concept from the health
geography literature (Delamater et al., 2012; Khan, 1992; Zhang et al., 2011). According to the
generalized gravity model,

𝑇𝑖𝑗 =

𝐴(𝑖)𝐵(𝑗)
𝛽

𝑑𝑖𝑗

, the calibration of the spatial interaction between Michigan snowmobilers and the trail
network should include origin and destination attributes, and the friction parameter. Because this
research aimed at assessing each snowmobiler’s spatial interaction with the Michigan
snowmobile trail network, the origin attribute for each snowmobiler was set at 1.

Figure 7 Michigan’s Five Regions
To meet resource management and marketing strategy development needs, the survey
divided Michigan into five regions (See Figure 7 for Five Regions). Each survey respondent was

45

asked to report their days spent snowmobiling in each region. The spatial interaction between
snowmobilers and the snowmobile trail network therefore was measured based on these
managerial regions.
The destination attribute is defined as the availability of the snowmobile trail network.
Availability refers to the volume and type of existing services (Penchansky & Thomas, 1981).
Therefore, total snowmobile trail mileage and snow precipitation in winter 2007-08 in each
region were defined as the availability of Michigan snowmobiling opportunity.
The friction parameter, or accessibility, on the other hand, referred to the travel cost
between residential origin and destination (Widener, Farber, Neutens, & Horner, 2013). Travel
cost can be calculated using Euclidean distance, actual travel distance, travel time, or cognitive
distance. This research used Euclidean distance to measure accessibility to Michigan’s
snowmobile trail network for the following reasons. First, the road network in Michigan is well
constructed in both north-south and east-west directions. Using Euclidean distance measurement
can represent the magnitude and variation of travel distance in such a large scale as statewide
travel. Second, Michigan is divided into Upper and Lower Peninsulas by the Mackinac Bridge.
This research measured Euclidean distance using the middle point of the Bridge as the starting or
ending point if cross-peninsula travel was calculated. This approach aided in measurement
accuracy by avoiding non-existent travel paths crossing over the Great Lakes. Lastly, Euclidean
distance measurement is less time consuming because of fewer computational inputs compared
to road network analysis and is likely to yield a more generalizable result. While incorporating
real-world traffic network may provide a more accurate measurement of travel distance or travel
cost than Euclidean distance measurement, the survey didn’t ask if respondents took the shortest
distance between their origin and their snowmobiling destination; using the shortest road
46

network distance may therefore not be a fully accurate representation of snowmobilers’ travel
behavior.
To calculate the distance between origin and each region, a destination point needed to be
located in order to best represent the trail network a snowmobiler visited in the winter 2007-08.
The most snowmobile friendly point in each region, defined as the centroid point of the county in
each region with the largest product of average snowfall and trail mileage, was located in order
to calculate distance between origin and destination regions.
While availability and accessibility of snowmobiling opportunity provide objective
measurements of spatial interaction between Michigan snowmobilers and the trail network,
actual participation is the realization of subjective input. The percentage of actual participation
reflects the preference each snowmobiler has for different destinations. Therefore, total
snowmobile days and snowmobile days in each Michigan managerial region in the winter of
2007-08 were incorporated into the measure of spatial interaction.
In sum, the measure of the spatial interaction between Michigan snowmobilers and the
snowmobile trail network included each snowmobiler’s snowmobile days within the region,
Euclidean distance to each region’s most snowmobile friendly point, snowmobile trail mileage
within the region, and winter snow precipitation within the region. The gravity model was
adapted to account for the availability and accessibility of Michigan snowmobiling opportunity,
while the percentage of actual participation spent in each region was applied to weight the spatial
interaction.
Each respondent’s spatial interaction with the Michigan snowmobile trail network in each
region was therefore be calculated by:
47

𝑇𝑖𝑗 =

𝑎𝑗 𝑏𝑗 𝑐𝑗
×
𝐷 𝑑𝛽
𝑖𝑗

Where 𝑇𝑖𝑗 : Spatial interaction between origin 𝑖 and region 𝑗
𝑎𝑗 : Snowmobile days in region 𝑗
𝑏𝑗 : Trail mileage in region 𝑗

𝑐𝑗 : Winter snowfall in region 𝑗

𝑑𝑖𝑗 : Distance from residential origin 𝑖 to region 𝑗
𝐷 : ∑5𝑖=1 𝑎𝑖

(Total snowmobile days in Michigan)

Snowmobile days in region 𝑗 (𝑎𝑗 ) were derived from the survey data. Respondents were

asked to report their snowmobile days in winter 2007-08 in each of the five regions. Snowmobile
days in five regions were totaled to compute 𝐷 (Total snowmobile days in Michigan). Trail

mileage in region 𝑗 (𝑏𝑗 ) was calculated using the Michigan Snowmobile Trail shapefile and the
overlay function “identity” in ESRI’s ArcGIS 10.1. Winter snowfall in region 𝑗 (𝑐𝑗 ) was

calculated by using snow precipitation data and ESRI’s ArcGIS 10.1, described in detail in the
next section.
In the gravity model, the distance (𝑑𝑖𝑗 ) is placed as the denominator to account for

distance decay. The distance decay parameter 𝛽 was set to 1.91 for the spatial accessibility

measures when using public open spaces (Giles-Corti & Donovan, 2002). The distance (𝑑𝑖𝑗 ) was

measured from each respondent’s residential origin to the most snowmobile friendly point of

each destination region. The mailing address of each survey respondent served as the residential
origin. For those who use a Post Office Box (P.O. Box) as their mailing address, the centroid of
their reported zip-code was used as the residential origin to calculate distance. Addresses of

48

survey respondents were geocoded into coordinates with latitude and longitude data using an online geocoding source (http://www.findlatitudeandlongitude.com/). As mentioned earlier, the
middle point of the Mackinac Bridge was used as the starting and ending point if cross-peninsula
travel was calculated. Distance between the residential origin and the destination point in each
region was measured by using a proximity function “Point Distance” in ESRI’s ArcGIS 10.1.
Calculating Winter Snowfall in Region j

Winter snowfall in each region was derived from a continuous raster surface representing

Michigan snow precipitation, which was created by a spatial interpolation method called kriging.
All procedures utilized snow precipitation data in ESRI’s ArcGIS 10.1. Daily snow precipitation
for winter 2007-08 (213 days) was summed to create the Z value for each weather station for the
procedures that followed. The X and Y coordinates for all 82 weather stations and their Z values
(accumulated snowfall) were plotted as the input point feature for kriging. The ordinary kriging
method was used, with an exponential semivariogram model, 12 neighbor points, and the raster
resolution of one inch by one inch grid, to visualize the continuity and variability of the snowfall
across Michigan. Once the snowfall surface was created, the five region shapefile was used as
the mask layer and a zonal function “zonal statistics as table” was used to calculate the winter
snowfall for each region and present the results in a table.
Assessing the Relationships between Spatial Interaction and Specialization
Spatial interaction for each survey respondent was treated as continuous data. Therefore,
the relationship between spatial interaction with Michigan’s snowmobile trail network and
snowmobile specialization was tested with an effect size index Eta (𝜂). The value of 𝜂 was used
to indicate the strength and the direction of the relationship (Vaske, 2008). Correlation between
spatial interaction and recreational specialization was utilized to examine hypothesis 11.
49

Additionally, the one-way ANOVA was run to test how different specialization subgroup
affected snowmobilers’ spatial interaction with the snowmobile trail network. The post-hoc
comparison used Tukey’s Honesty Significant Difference test to examine spatial interaction
differences among snowmobile specialization groups. A significance level of .05 was used to
determine whether to reject the null hypothesis. Setting the level of significance at .05 reduces
the probability of committing a Type II error while slightly increasing the probability of
committing a Type I error.

Definition and Treatment of Outliers
Extreme values may substantially distort the results of survey data, making analysis
problematic. To address research question one and to ensure the generic characteristics of
Michigan snowmobilers were well represented, outliers were identified and treated as a special
group. This identified group was excluded from the process of identifying recreational
specialization subgroups and was termed extreme cases hereafter.
A standardized Z-score was calculated for each of the six recreational specialization
variables to identify outliers. Traditionally, any value beyond three standard deviations of the
mean was suggested as an outlier (Stevens, 2009). To preserve the variability of the data, this
research used six standard deviations to help identify outliers. Those cases with any recreational
specialization variable Z value six or larger (≥ 6) or minus six or smaller (≤ -6) were set aside as
extreme cases.

50

Treatment of Missing Data
Missing data existed in both the survey and snow precipitation data. To reduce bias in the
analysis and findings when missing data occurred, two approaches were used: the delete
respondent solution and the sample means solution (Vaske, 2008).
For survey data, the delete respondent solution was first used to deal with missing values.
The entire case was omitted from the sample when any of these four criteria applied: no
residential address was available; snowmobile days in five regions was missing; three or more
(out of six) recreational specialization variables were missing; and four or more (out of nine)
perception of risk variables were missing. After omitting outliers, the sample mean of the
variable was then applied to replace missing values for the rest of the cases. Although using
sample means solution reduces variability, the solution provides conservative correlation
coefficients when comparing to replacing missing data with group means or random assignments
within groups.
For the snow precipitation data, there were missing values randomly distributed across
the dataset. To make sure all weather stations provided precipitation data for the entire winter of
2007-08 for spatial analysis, the missing values of each weather station were replaced by the
station mean precipitation.

51

Chapter Four
Research Findings
This chapter presents the research findings and addresses the research questions and
hypotheses. There are two sections in this chapter: descriptive statistics and inferential statistics.

Descriptive Statistics
This section provides descriptive results derived from Michigan snowmobiler survey
data, Michigan snow precipitation data, and Michigan snowmobile trail network GIS shapefile
data. Major components include general descriptive statistics of the research data with a focus on
the three themes of this research: recreational specialization, perception of risk, and spatial
interaction.
Surveyed Michigan Snowmobilers
Of those 853 Michigan residents who completed and returned the survey, 39 cases were
omitted from the analysis because no residential address was available, data on snowmobile days
in the five regions were missing, three or more (out of six) recreational specialization variables
were missing, or four or more (out of nine) perception of risk variables were missing. An
additional seven cases were defined as extreme cases and were excluded from the process of
identifying recreational specialization subgroups. A total of 807 cases with the necessary data for
analysis were retained. Among them, 573 (71.0%) lived in region five (southern Lower
Peninsula), 91 (11.3%) lived in region four (northeastern Lower Peninsula), 82 (10.2%) lived in
region three (northwestern Lower Peninsula), 31 (3.8%) lived in region one (western Upper
Peninsula), and 30 (3.7%) lived in region two (eastern Upper Peninsula). All valid addresses
were aggregated to the county level and plotted in Figure 8.

52

Figure 8 Number of Surveyed Snowmobilers by County
The Year of 2007-08 Winter Snowfall
The continuous raster surface representing Michigan snow precipitation was created by
ordinary kriging. The zonal function “zonal statistics as table” was used to calculate the winter
snowfall for each region. On average, region two (eastern Upper Peninsula) had the highest
winter snowfall of 138.4 inches, followed by 136.9 inches in region one (western Upper
Peninsula), 102.6 inches in region three (northwestern Lower Peninsula), 83.7 inches in region
four (northeastern Lower Peninsula), and 74.2 inches in region five (southern Lower Peninsula).
Table 3 shows the descriptive statistics of winter snowfall while figure 9 demonstrates the
continuous surface of winter 2007-08 snowfall and the allocation of the Michigan snowmobile
trail network.
Table 3 Winter Snowfall by Region (2007-08)
Region
Region 2
Region 1
Region 3
Region 4
Region 5
S.D = Standard Deviation

Mean
138.4
136.9
102.6
83.7
74.2

53

S.D.
32.1
34.4
12.7
16.8
19.0

Figure 9 Winter 2007-08 Snowfall Kriging Surface with Snowmobile Trail

Michigan Snowmobile Trail
The overlay function “identity” was used to calculate trail mileage for each region. The
result showed that region one (western Upper Peninsula) had the most snowmobile trail mileage,
followed by region two (eastern Upper Peninsula), region four (northwestern Lower Peninsula,
region three (northeastern Lower Peninsula), and region five (south Lower Peninsula). Region
one had 1,819 miles of snowmobile trail, region two had 1370.5 miles, region four had 1194.6
miles, region three had 994.6 miles, and region five had 809.4 miles. Figure 10 illustrates the
distribution of the snowmobile trail network in the five regions of Michigan.

54

Figure 10 Snowmobile Trail Network in Five Regions
Most Snowmobile Friendly Point in Each Region
As the research divided Michigan into five managerial regions, the most snowmobile
friendly point in each region was defined as the centroid point of the county with the greatest
product of average snowfall and trail mileage. The centroid point of Ontonagon County was
identified as the most snowmobile friendly point for region one (western Upper Peninsula),
Chippewa County for region two (eastern Upper Peninsula), Lake County for region three
(northwestern Lower Peninsula), Oscoda County for region four (northeastern Lower Peninsula),
and Cass County for region five (southern Lower Peninsula). These most snowmobiler friendly
points were utilized to calculate the distance between each snowmobiler’s residential origin and
their snowmobile destination in each region. Figure 11 illustrates the location of the most
snowmobile friendly point for all five regions.

55

Figure 11 Most Snowmobile Friendly Point in each Region
General Characteristics of Michigan Snowmobilers
Six variables were selected to assess Michigan snowmobilers’ recreational specialization.
On average, Michigan snowmobilers owned two snowmobiles, spent 1,954 dollars on
snowmobile related items such as maintenance and insurance in the previous 12 months, had
snowmobiled for 58% of their life, and rode their snowmobiles 21 days traveling 1,544 miles in
the winter of 2007-08. Only 25% of them belonged to one or more snowmobile associations or
clubs. Table 4 presents the descriptive statistics of Michigan snowmobilers’ general
characteristics.
Table 4 Michigan Snowmobilers’ General Characteristics
Variable
Mean
Minimum
Numbers of snowmobiles owned
2.4
1
Money spent in the last 12 months ($)
1953.6
0
Snowmobiling experience
.58
.02
Snowmobile days in winter 07-08
20.9
0
Miles driven in winter 07-08
1544.0
0
Snowmobile membership
.3
0
S.D = Standard Deviation
56

Maximum
5
23900
.98
128
8250
3

S.D.
1.1
3424.6
.23
18.3
1345.5
.5

Michigan Snowmobilers’ General Perception of Risk and Internal Consistency
Nine snowmobile riding hazards were utilized to assess Michigan snowmobilers’
perception of risk. A five point Likert Scale, ranging from one being not dangerous to five being
extremely dangerous, was provided for describing snowmobilers’ perception of risk. The result
showed that Michigan snowmobilers considered “Operation of snowmobile by a legally
intoxicated person” (4.4), “Driver lacking skill in operating machine” (4.1), and “Speed of
snowmobile” (3.8) highly dangerous; “Operation of snowmobile by person who has been
drinking but is not legally intoxicated” (3.3), “Cars/trucks on seasonal roads” (3.1), and “Other
uses of designated snowmobile trails” (2.8) moderately dangerous; and “Public trail conditions”
(2.4), “Snowmobiling on county/state road shoulders” (2.2), and “Public trail design” (2.1)
slightly dangerous.
Table 5 Descriptive Statistics and Reliability of Snowmobiling Perception of Risk
Variable Name
Mean*
Operation of snowmobile by a legally intoxicated person (0.08 or higher
4.4
blood alcohol)
Driver lacking skill in operating machine
4.1
Speed of snowmobile
3.8
Operation of snowmobile by person who has been drinking but is not
3.3
legally intoxicated (0.01-0.07 blood alcohol)
Cars/trucks on seasonal roads
3.1
Other uses of designated snowmobile trails (e.g., dog sledding, cross
2.8
country skiing)
Public trail conditions
2.4
Snowmobiling on county/state road shoulders
2.2
Public trail design
2.1
* Based on a five point Likert scale with 1being Not Dangerous, 2 Slightly Dangerous, 3
Moderately Dangerous, 4 Highly Dangerous, and 5 Extremely Dangerous
Internal consistency score: Cronbach’s α = 0.76

S.D.
0.9
1.0
1.1
1.4
1.2
1.3
1.1
1.0
1.0

Overall, the reliability analysis demonstrated that the nine items in the Michigan
snowmobiler perception of risk measurement had a fairly adequate level of internal consistency

57

(Cronbach’s α = 0.76). Table 5 shows the descriptive statistics for the nine items of Michigan
snowmobiling perception of risk.

Michigan Snowmobilers’ Actual Participation
The results were grouped based on respondents’ residential region as snowmobilers from
the same region were more likely to face similar situations with regard to the spatial distribution
and structure of the snowmobile trail network. The survey asked snowmobilers to report their
snowmobiling days in each of the five Michigan regions. The results showed that survey
respondents who did not live in region five (southern Lower Peninsula) were more likely to
snowmobile in their own residential region, while the majority of the respondents (71%) who
lived in region five tended to snowmobile in regions three and four. On average, respondents
who lived in region one (western Upper Peninsula) spent 20 days in the winter of 2007-08
snowmobiling in the western Upper Peninsula and about 3 days in region two (eastern Upper
Peninsula). Respondents who lived in region two spent 19 days snowmobiling in their residential
region and about 2 days in the western Upper Peninsula. According to the results, those who
lived in the Upper Peninsula did not go to the lower half of Michigan for snowmobiling.
On the other hand, for those who lived in the northern half of the Lower Peninsula, region
two, three (Northwestern Lower Peninsula), and four (Northeastern Lower Peninsula) were their
preferred snowmobile destinations. Survey respondents who lived in region three spent 21 days
in their own region, 3 days in region two, and about a day in region one and region four. They
seldom went to region five for snowmobiling. Those who lived in region four spent 20 days
snowmobiling in their own residential region, 6 days in region three, 2 days in region two, and
less than a day in region one. No snowmobilers from region four went to region five for
snowmobiling. The majority of the survey respondents lived in region five but tended to spend
58

their winter snowmobiling in the northern half of the Lower Peninsula of Michigan. On average,
they spent 7 days in region three, 4 days in region four, 4 days in region five, 3 days in region
two and 1 day in region one.
Table 6 Michigan Snowmobilers’ Actual Participation in Five Michigan Regions
Residential Region
N
Destination Region
Mean
Region 1
Region 1
20.1
31
(western Upper Peninsula)
Region 2
2.6
(4%)
Region 3
.0
Region 4
.0
Region 5
.0
Region 2
Region 1
1.9
30
(eastern Upper Peninsula)
Region 2
19.4
(4%)
Region 3
.0
Region 4
.0
Region 5
.0
Region 3
Region 1
.8
82
(northwestern Lower Peninsula)
Region 2
2.7
(10%)
Region 3
20.7
Region 4
.5
Region 5
> .0
Region 4
Region 1
.4
91
(northeastern Lower Peninsula)
Region 2
2.1
(11%)
Region 3
6.0
Region 4
20.0
Region 5
.0
Region 5
Region 1
1.0
573
(southern Lower Peninsula)
Region 2
2.9
(71%)
Region 3
6.9
Region 4
4.1
Region 5
4.0
Overall
Region 1
1.7
807
Region 2
3.4
(100%)
Region 3
7.7
Region 4
5.2
Region 5
2.9

S.D.
17.6
7.7
.0
.0
.0
5.0
21.3
.0
.0
.0
2.4
7.6
21.6
2.4
.2
2.2
4.9
12.6
22.6
.0
2.9
5.8
11.3
7.3
8.3
5.8
7.8
13.4
11.2
7.3

Overall, the northern half of the Lower Peninsula was the most common destination for
all Michigan snowmobilers as they spent 8 days snowmobiling in region three and 5 days in
region four. Region two was the preferred destination in the Upper Peninsula with 3
snowmobiling days. Additionally, Michigan snowmobilers spent a little less than 3 days in
59

region 5 and about 2 days in region one. Table 6 presents the descriptive statistics for Michigan
snowmobilers’ actual participation overall and in each region.
Travel Distance to Michigan Snowmobile Trail Network
Spatial accessibility of Michigan snowmobile trail network was measured by Euclidean
distance from all respondents’ mailing addresses to the most snowmobile friendly point in each
region, regardless if the respondent actually traveled to the region for snowmobiling or not. The
longest mean distance Michiganders traveled to snowmobile was 393 miles to region one
(western Upper Peninsula), followed by 214 miles to region two (eastern Upper Peninsula), 159
miles to region five (southern Lower Peninsula), 129 miles to region four (northeastern Lower
Peninsula), and 125 miles to region three (northwestern Lower Peninsula). Table 7 presents the
descriptive statistics of travel distance to snowmobile trail network in each region.
Table 7 Average Travel Distance for All Snowmobilers to Trail Network in each Region
Destination Region
Mean
S.D.
Distance to Region 1
392.5
101.2
Distance to Region 2
214.2
69.8
Distance to Region 5
158.9
88.5
Distance to Region 4
128.7
55.1
Distance to Region 3
124.8
61.4
Spatial Interaction between Snowmobiler’ Residential Region and Snowmobile Trail Network
Spatial interaction between Michigan snowmobilers’ residential region and the
snowmobile trail network was defined as the function of snowmobile trail network availability,
accessibility, and snowmobilers’ actual participation in each region. Because the distribution of
all spatial interaction was highly skewed (skewness range from 15.3 to 28.7; kurtosis range from
267.7 to 823.4), this research ranked spatial interaction. It ranked all 807 respondents with rank 1
meaning the least spatial interaction and rank 807 meaning the most spatial interaction. A mean
rank was reported instead of mean score. The descriptive spatial interaction results showed that
60

region three (northwestern Lower Peninsula) received the most spatial interaction among five
Michigan regions with a mean rank of 333.6. This is followed by region four (northeastern
Lower Peninsula) with a mean rank of 263, region two (eastern Upper Peninsula) with 258,
region five (southern Lower Peninsula) with 184.3, and region one (western Upper Peninsula)
with 146.2. Table 8 shows the descriptive statistics of spatial interaction between all Michigan
snowmobilers’ residential region and the snowmobile trail network in each region.
Table 8 Total Spatial Interaction Received by Snowmobile Trail Network Destination Region
Destination Region
Mean Rank
Sum
S.D.
Region 3
333.6
25049.9
309.5
Region 4
263.0
65747.2
1843.0
Region 2
258.0
32191.1
757.1
Region 5
184.3
6711.7
2017.6
Region 1
146.2
9329.2
146.2
Due to proximity, Michigan snowmobilers had the most spatial interaction within their
own residential region. For regions other than their residential region, snowmobilers who lived in
region one had the most spatial interaction with the eastern Upper Peninsula; snowmobilers in
region two had the most spatial interaction with the western Upper Peninsula; snowmobilers in
region three tended to have more spatial interaction with the eastern Upper Peninsula than other
regions; snowmobilers in region four had the most spatial interaction with the northwestern
Lower Peninsula; and, those who lived in region five also had the most spatial interaction with
the northwestern Lower Peninsula. Additionally, snowmobilers from the Upper Peninsula
(region one and two) had zero spatial interaction with the Lower Peninsula (region three, four,
and five). Those who lived in the northern Lower Peninsula had minimal spatial interaction with
the southern Lower Peninsula.

61

Table 9 Spatial Interaction between Snowmobilers and Trail Network by Region
Residential Region
Destination Region
Sum
Mean Rank
Region 1
Region 1
9175.1
812.4
(western Upper Peninsula)
Region 2
29.3
173.8
Region 3
.0
1.0
Region 4
.0
1.0
Region 5
.0
1.0
Region 2
Region 2
31151.5
780.8
(eastern Upper Peninsula)
Region 1
66.0
211.4
Region 3
.0
1.0
Region 4
.0
1.0
Region 5
.0
1.0
Region 3
Region 3
20057.0
675.5
(northwestern Lower Peninsula)
Region 2
319.0
210.2
Region 4
103.9
73.9
Region 1
15.8
102.1
Region 5
.2
8.7
Region 4
Region 4
64252.1
648.0
(northeastern Lower Peninsula)
Region 3
633.0
244.4
Region 2
191.1
206.7
Region 1
3.3
41.2
Region 5
.0
1.0
Region 5
Region 5
6711.5
258.0
(southern Lower Peninsula)
Region 3
4359.9
334.2***
Region 4
391.2
256.8**
Region 2
500.2
250.2*
Region 1
69.0
129.7
*** Highest spatial interaction
** Second highest spatial interaction
*Third highest spatial interaction.

S.D.
697.9
2.2
.0
.0
.0
3854.8
5.8
.0
.0
.0
948.5
9.5
2.0
.7
> .0
5474.8
13.0
5.2
.2
.0
72.6
17.6
4.8
1.5
.3

For cross-regional travel, as table 9 shows, the highest spatial interaction occurred when
snowmobilers from region five traveled to the northwestern Lower Peninsula for snowmobiling,
and the second highest spatial interaction happened when snowmobilers from region five
traveled to the northeastern Lower Peninsula for snowmobiling. When snowmobilers from
region five traveled to the eastern Upper Peninsula for snowmobiling, they generated the third
highest spatial interaction during the winter of 2007-08.

62

Inferential Statistics
This section presents the results of the inferential statistics which were used to address
the research questions and hypotheses. The section first addresses results of using summative
indices and cluster analysis for identifying Michigan snowmobilers’ recreational specialization.
Secondly it focuses on the influence of specialization on the perception of snowmobiling risks
and spatial interaction.
Michigan Snowmobilers’ Recreational Specialization
The theoretical framework of recreational specialization was utilized to study the withingroup social world of Michigan snowmobilers. Using a summative index and cluster analysis are
two common classification approaches for identifying different levels of recreational
specialization. The statistical results of both approaches were used to address research question
one: What are the behavior, skill/knowledge, and psychological attachment characteristics of
snowmobilers in different levels of specialization using summative indices and cluster analysis?
Summative Index for Recreational Specialization
After the specialization continuum was formed, three subgroups: expert, intermediate,
and novice, with 269 Michigan snowmobilers in each, were identified. Overall, novice
snowmobilers owned little less than 2 snowmobiles, spent 539 dollars on snowmobile related
items such as maintenance and insurance in the previous 12 months, had snowmobiled for 45%
of their life, and rode 11 days traveling 648 miles in the winter of 2007-08. Only 6% of them
belonged to a snowmobile association or club.
Intermediate snowmobilers on average owned slightly more than 2 snowmobiles, spent
1,130 dollars for snowmobiling related items, had snowmobiled for 62% of their life, and rode

63

18 days traveling 1,277 miles in the winter of 2007-08. About 24% of them participated in
snowmobile associations or clubs.
The expert snowmobilers generally owned 3 sleds, spent 4,192 dollars on snowmobiling
maintenance, insurance, and other related items, had snowmobiled for 69% of their life, and rode
33 days traveling 2,707 miles in the winter of 2007-08. About 46% of them were a member of
snowmobile associations or clubs. Table 10 provides the results using the summative index to
identify Michigan snowmobilers’ subgroups.
Table 10 Summative Index Result for Snowmobilers’ Recreational Specialization
Subgroup
Variable
Mean
Novice
Numbers of snowmobiles owned
1.7
(N = 269)
Money spent in the last 12 months ($)
539.1
Snowmobiling experience
0.45
Snowmobile days in winter 07-08
11.1
Miles driven in winter 07-08
647.9
Snowmobile membership
0.1
Intermediate Numbers of snowmobiles owned
2.4
(N = 269)
Money spent in the last 12 months ($)
1129.9
Snowmobiling experience
0.62
Snowmobile days in winter 07-08
18.2
Miles driven in winter 07-08
1277.0
Snowmobile membership
0.3
Expert
Numbers of snowmobiles owned
3.3
(N = 269)
Money spent in the last 12 months ($)
4192.0
Snowmobiling experience
0.69
Snowmobile days in winter 07-08
33.3
Miles driven in winter 07-08
2707.2
Snowmobile membership
0.5

S.D.
0.7
1162.2
0.24
8.0
509.8
0.2
0.87
1739.1
0.19
12.3
805.0
0.5
1.1
4814.4
0.18
23.1
1516.9
0.6

To test research hypothesis one: There is no difference in group composition across
levels of specialization using summative indices and cluster analysis, the examination of the
differences among three snowmobile subgroups is warranted. A boxplot and Shapiro-Wilk’s
normality test were performed to assess if the underlying assumptions of the one-way ANOVA
were met. The boxplot result showed that there were outliers in each of the three subgroups for
64

money spent in the past 12 months, snowmobile days in winter of 2007-08, and miles driven in
the winter of 2007-08. The normality test also revealed that data of all six specialization
variables in each subgroup were not normally distributed (p < .05). Therefore, the nonparametric equivalent of Kruskal-Wallis test was run to determine if there were differences in
each of the specialization variables among the novice, intermediate, and expert snowmobiler
subgroups. Statistical significance was accepted at the p < .05 level for the omnibus test and p <
.0167 level for the multiple comparisons (.05 divided by 3). Pairwise comparisons were
performed using Dunn's (1964) procedure with a Bonferroni correction for multiple
comparisons.
The results showed that the number of snowmobiles owned was significantly different
among the three subgroups of recreational specialization: χ2 (2) = 275.235, p < .0005. Post-hoc
analysis revealed statistically significant differences in numbers of snowmobiles owned between
the novice (Mean = 1.7) and intermediate snowmobilers (Mean = 2.4; p < .0005), intermediate
(Mean = 2.4) and expert snowmobilers (Mean = 3.3; p < .0005), and novice and expert
snowmobilers (p < .0005).
Money spent on snowmobiling in the last 12 months was significantly different among
the three subgroups of recreational specialization: χ2 (2) = 209.313, p < .0005. Post-hoc analysis
revealed statistically significant differences in money spent in the last 12 months between the
novice (Mean = 539.1) and intermediate snowmobilers (Mean = 1129.9; p < .0005), intermediate
(Mean = 1129.9) and expert snowmobilers (Mean = 4192.0; p < .0005), and novice and expert
snowmobilers (p < .0005).

65

Snowmobile experience was significantly different among the three subgroups of
recreational specialization: χ2 (2) = 141.595, p < .0005. Post-hoc analysis revealed statistically
significant differences in snowmobile experience between novice (Mean = .45) and intermediate
snowmobilers (Mean = .62; p < .0005), intermediate (Mean = .62) and expert snowmobilers
(Mean = .69; p < .0005), and novice and expert snowmobilers (p < .0005).
Snowmobile days in winter 2007-08 were significantly different among the three
subgroups of recreational specialization: χ2 (2) = 253.814, p < .0005. Post-hoc analysis revealed
statistically significant differences in snowmobile days in winter 2007-08 between the novice
(Mean = 11.1) and intermediate snowmobilers (Mean = 18.2; p < .0005), intermediate (Mean =
18.2) and expert snowmobilers (Mean = 33.3; p < .0005), and novice and expert snowmobilers (p
< .0005).
Miles driven in winter 2007-08 were significantly different among the three subgroups of
recreational specialization: χ2 (2) = 387.970, p < .0005. Post-hoc analysis revealed statistically
significant differences in miles driven in winter 2007-08 between the novice (Mean = 647.9) and
intermediate snowmobilers (Mean = 1277.0; p < .0005), intermediate (Mean = 1277.0) and
expert snowmobilers (Mean = 2707.2; p < .0005), and novice and expert snowmobilers (p <
.0005).
Snowmobile membership was also significantly different among the three subgroups of
recreational specialization: χ2 (2) = 117.472, p < .0005. Post-hoc analysis revealed statistically
significant differences in snowmobile membership between the novice (Mean = .1) and
intermediate snowmobilers (Mean = .3; p < .0005), intermediate (Mean = .3) and expert
snowmobilers (Mean = .5; p < .0005), and novice and expert snowmobilers (p < .0005). Table 11

66

presents the result of the post-hoc analysis for summative index method of identifying
recreational specialization subgroups.
Table 11 Pairwise Comparison Result of Summative Index for Snowmobilers’ Specialization
Specialization Subgroups
Variable
Novice
Intermediate
Expert
χ2 (sig.)
Numbers of snowmobiles owned
ac
ab
bc
275.235**
Money spent in the last 12 months
ac
ab
bc
209.313**
Snowmobiling experience
ac
ab
bc
141.595**
Snowmobile days in winter 07-08
ac
ab
bc
253.814**
Miles driven in winter 07-08
ac
ab
bc
387.970**
Snowmobile membership
ac
ab
bc
117.472**
ac,ab,bc Groups with a same letter were significantly different at the .05 level of confidence
** = Significant at the .001 level
In total, the summative index method displayed a logical recreational specialization
continuum for Michigan snowmobilers as expert snowmobilers showed the highest commitment
in all six specialization measures, while novice snowmobilers committed the least (Figure 12).
The Kruskal-Wallis test and post-hoc results showed that there were statistically significant
differences on behavior, skill/knowledge, and psychological attachment characteristics among
novice, intermediate, and expert snowmobilers. For each of the six recreational specialization
variables, the three subgroups also revealed distinct differences from the other subgroups.

67

z score
1

2707
3.3
33
$4,192
46%
69%

0.8
0.6
0.4
62%

0.2

-0.2
-0.4
-0.6
-0.8

No. of Snowmobile

24%

0

2.4
18
1277
$1,130

$539
6%
11
45%
648
1.7

Novice
N = 269

Money spent
Experience
Snowmobile days
Miles driven
Membership

Intermediate
N = 269

Expert
N = 269

Figure 12 Michigan Snowmobiler Recreational Specialization – Summative Index Result
Cluster Analysis for Recreational Specialization
K-mean cluster analysis was performed to identify subgroups with similar characteristics.
Three subgroups, moderates, big spenders, and trail warriors, were identified and named based
on their within-group features. The results of cluster analysis showed that there were 194 trail
warriors (24%), 89 big spenders (11%), and 524 moderate snowmobilers (65%). In general,
moderate snowmobilers owned 2 snowmobilers, spent 686 dollars on snowmobile related items
such as maintenance and insurance in the previous 12 months, had snowmobiled for 55% of their
life, and rode about 14 days traveling 890 miles in the winter of 2007-08. About 19% of them
belonged to one or more snowmobile associations or clubs.

68

Big spenders on average owned 3 snowmobiles, had snowmobiled for 63% of their life,
and rode 27 days traveling 2,058 miles in the winter of 2007-08. About 31% of them participated
in one or more snowmobile associations or clubs. Particularly, these big spenders spent on
average 10,398 dollars for snowmobiling related items that may have included purchase of a
snowmobile, snowmobile equipment, snowmobile repair/maintenance, insurance on the
snowmobile(s), and off-season storage costs. Of these 89 big spenders, 84 (94%) purchased a
snowmobile in the past 12 months, spending 8,398 dollars for the machine.
Table 12 Cluster Analysis Result for Snowmobilers’ Recreational Specialization
Subgroup
Variable
Mean
Moderate
Numbers of snowmobiles owned
2.0
(N = 524)
Money spent in the last 12 months ($)
685.9
Snowmobiling experience
.55
Snowmobile days in winter 07-08
13.9
Miles driven in winter 07-08
899.6
Snowmobile membership
.2
Big Spender
Numbers of snowmobiles owned
3.0
(N = 89)
Money spent in the last 12 months ($)
10398.2
Snowmobiling experience
.63
Snowmobile days in winter 07-08
27.0
Miles driven in winter 07-08
2058.3
Snowmobile membership
.4
Trail Warrior Numbers of snowmobiles owned
3.3
(N = 194)
Money spent in the last 12 months ($)
1503.7
Snowmobiling experience
.66
Snowmobile days in winter 07-08
37.0
Miles driven in winter 07-08
3048.7
Snowmobile membership
.4

S.D.
.8
1062.7
.24
9.7
634.8
.4
1.1
3560.9
.23
19.4
1213.4
.6
1.1
1638.6
.18
23.7
1507.3
.6

The trail warriors generally owned slightly more than 3 sleds, spent 1,504 dollars on
snowmobiling maintenance, insurance, and other related items, had snowmobiled for 66% of
their life, and rode 37 days traveling 3,049 miles in the winter of 2007-08. About 40% of them
belonged to one or more snowmobile associations or clubs. Table 12 shows the result for using
K-mean cluster analysis to identify Michigan snowmobilers’ subgroups.

69

To test research hypothesis one: There is no difference in group composition across
levels of specialization using summative indices and cluster analysis, examination of the
differences among the three snowmobile subgroups is warranted. A boxplot and Shapiro-Wilk’s
normality test were performed to assess if the underlying assumptions of the one-way ANOVA
were met. The boxplot result showed that there were outliers in money spent in the past 12
months and snowmobile memberships. The normality test also revealed that data of all six
specialization variables in each subgroup were not normally distributed (p < .05). Therefore, the
non-parametric equivalent of the Kruskal-Wallis test was again performed to decide if there were
differences in each of the specialization variables among moderate, big spender, and trail warrior
snowmobilers. Statistical significance was accepted at the p < .05 level for the omnibus test and
p < .0167 level for the multiple comparisons (.05 divided by 3).
Pairwise comparisons result showed that numbers of snowmobiles owned was
significantly different among the three subgroups: χ2 (2) = 212.894, p < .0005. Post-hoc analysis
revealed statistically significant differences in numbers of snowmobiles owned between the
moderates (Mean = 2.0) and big spenders (Mean = 3.0; p < .0005) and moderates (Mean = 2.0)
and trail warriors (Mean = 3.3; p < .0005). There was no significant difference between trail
warrior and big spender snowmobilers.
Money spent in the last 12 months was statistically different among the three subgroups:
χ2 (2) = 300.272, p < .0005. Post-hoc analysis revealed statistically significant differences in
money spent in the last 12 months between the moderates (Mean = 685.9) and big spenders
(Mean = 10398.2; p < .0005), big spenders (Mean = 10398.2) and trail warriors (Mean = 1503.7;
p = .000), and moderate and trail warrior snowmobilers (p < .0005).

70

Snowmobile experience was significantly different among the subgroups: χ2 (2) = 39.845,
p < .0005. Post-hoc analysis revealed statistically significant differences in numbers of
snowmobiles owned between the moderates (Mean = .55) and big spenders (Mean = .63; p <
.0005) and moderates (Mean = .55) and trail warriors (Mean = .66; p < .0005). There was no
significant difference between trail warrior and big spender snowmobilers.
Snowmobile days in winter 2007-08 was statistically significantly different among the
three subgroups, χ2 (2) = 254.678, p < .0005. Post-hoc analysis revealed statistically significant
differences in snowmobile days in winter 2007-08 between the moderates (Mean = 13.9) and big
spenders (Mean = 27.0; p < .0005), big spenders (Mean = 27.0) and trail warriors (Mean = 37.0;
p = .001), and moderate and trail warrior snowmobilers (p < .0005).
Miles driven in winter 2007-08 was significantly different among the three subgroups: χ2
(2) = 374.389, p < .0005. Post-hoc analysis revealed statistically significant differences in miles
driven in winter 2007-08 between the moderates (Mean = 899.6) and big spenders (Mean =
2058.3; p < .0005), big spenders (Mean = 2058.3) and trail warriors (Mean = 3048.7; p = .001),
and moderate and trail warrior snowmobilers (p < .0005).
Snowmobile membership was significantly different among the subgroups: χ2 (2) =
38.476, p < .0005. Post-hoc analysis revealed significant differences between the moderates
(Mean = .2) and big spenders (Mean = .4; p = .029) and moderates (Mean = .2) and trail warriors
(Mean = .4; p < .0005). There was no significant difference between trail warrior and big spender
snowmobilers. Table 13 presents the result of the post-hoc analysis for K-mean cluster analysis
method of identifying recreational specialization subgroups.

71

Table 13 Pairwise Comparison Result of Cluster Analysis for Snowmobilers’ Specialization
Specialization Subgroups
Variable
Moderate Big Spender Trail Warrior
χ2 (sig.)
Numbers of snowmobiles owned
ab
a
b
212.894**
Money spent in the last 12 months ($)
ac
ab
bc
300.272**
Snowmobiling experience
ab
a
b
39.845**
Snowmobile days in winter 07-08
ac
ab
bc
254.678**
Miles driven in winter 07-08
ac
ab
bc
374.389**
Snowmobile membership
ab
a
b
38.476**
a,b,ab,ac Groups with a same letter were significantly different at the .05 level of confidence
** = Significant at the .001 level
In summary, the cluster analysis approach did not present as smooth a recreational
specialization continuum along all dimensions as the summative index method did for Michigan
snowmobilers (Figure 13). However, the group composition of moderate, big spender, and trail
warrior snowmobilers illustrated different aspects of Michigan snowmobilers with unique and
statistically significant different behavior, skill/knowledge, and psychological attachment
characteristics. While the differences among moderates, big spenders, and trail warriors were
statistically significant, the post-hoc analysis results showed no apparent differences between big
spenders and trail warriors on numbers of snowmobiles owned, snowmobiling experience, and
snowmobile organization memberships. However, 94% of the big spenders purchased a
snowmobile in the snowmobile season 2007-08, which differentiated big spenders from trail
warriors. Other than the money spent in the past 12 months, the post-hoc analysis results also
showed significant differences between big spenders and trail warriors on snowmobile days in
winter 2007-08 and miles driven in winter 2007-08. The distinct behavior, skill/knowledge, and
psychological attachment characteristics were evident among moderate, trail warrior, and big
spender snowmobilers.

72

z score
3
No. of Snowmobile

2.5
2
1.5

$ 10,398

Money spent
Experience
Snowmobile days
Miles driven

3.3

Membership

37

1
0.5
0
-0.5

19%
55%
$686
900
14 2.0

3049
40%
66%
$1,504

3.0
2058
27
31%
63%

-1
Moderates
N = 524

Trail Warriors
N = 194

Big Spenders
N = 89

Figure 13 Michigan Snowmobiler Recreational Specialization – Cluster Analysis Result
Summary of Comparison of Summative Index and Cluster Analysis
Both the summative index method and the cluster analysis approach generated
significantly different levels of specialized Michigan snowmobilers. However, group
composition across levels of specialization between the two methods appeared to be dissimilar.
Novice, intermediate, and expert snowmobilers presented different behavior, skill/knowledge,
and psychological attachment characteristics than did moderate, big spender, and trail warrior
snowmobilers. For this reason, the research rejected the null hypothesis one and concluded that
there were differences in group composition across levels of specialization between the
summative index method and the cluster analysis.

73

To address research question one, the results of the summative index and cluster analysis
showed unique features of Michigan snowmobilers who purchased a trail permit in the winter of
2007-08. The summative index results described three distinct subgroups in the snowmobiler
social world: novice, intermediate, and expert. They revealed significant differences in their
behavioral, skill/knowledge, and psychological attachment characteristics. While novice
snowmobilers owned the least numbers of snowmobiles, spent just 539 dollars for their 11 days
and 648 miles of snowmobiling experience, expert snowmobilers on average owned 3 sleds,
spent 4,192 dollars for their 33 days of snowmobiling experience traveling 2,707 miles on their
sled. It is clear that more specialized snowmobilers presented more intensive behavioral
commitment (snowmobiles owned, money spent, and snowmobiling experience),
knowledge/skill (snowmobile days and mileage), and psychological attachment (snowmobile
organization membership) to the activity compared to less specialized snowmobilers. Expert
snowmobilers were more specialized than intermediate and novice snowmobilers, just as
intermediate snowmobilers showed a higher level of snowmobiling specialization than novice
snowmobilers.
On the other hand, the cluster analysis described different aspects of the within group
social world of Michigan snowmobilers. Three subgroups of recreational specialization,
moderates (65%), big spenders (11%), and trail warriors (24%), revealed significantly different
behavior, skill/knowledge, and psychological attachment characteristics. Unlike the three
subgroups identified by the summative index method, moderates, big spenders, and trail warriors
did not appear to be as hierarchical as novice, intermediate, and expert snowmobilers. The
majority of the Michigan snowmobilers were moderates, who casually participated in the activity
with the least number of snowmobiles owned (2 sleds), money spent on snowmobiling related
74

items ($686), snowmobile experience (55%), snowmobile days (14), miles driven (900), and
participation in snowmobile organization membership (19%). The rest of the snowmobilers were
divided into two groups with unique features in money spent, snowmobile days, and miles
driven. Big spenders spent a significant amount of money ($10,398) on snowmobiling related
items such as snowmobile purchase, insurance, repair/maintenance, and storage. They rode
significantly less days and fewer miles in the winter of 2007-08 compared to trail warriors, who
rode 37 days traveling 3,049 miles and spent 1,504 dollars on snowmobiling related items. There
were no significant differences between big spenders and trail warriors on number of
snowmobiles owned, snowmobiling experience, and organization membership.
Based on these results, it is considered that the combination of novice, intermediate, and
expert snowmobiling subgroups best represents the continuous nature of Michigan snowmobiling
recreational specialization. Not only because all aspects of the six specialization variables were
significantly different among the three subgroups, but the distribution of the three subgroups also
fits the original assumptions of the recreational specialization continuum proposed by Bryan
(1977). The influences of specialization on the perception of risks and spatial interaction were
then assessed by the results identified by the summative index method. The following research
findings were based on the difference among novice, intermediate, and expert snowmobilers.
Influence of Specialization on the Perception of Risk
Previous studies showed mixed results on recreationalists’ characteristics and their
perception of risks. To address research question two: what are the relationships between levels
of recreational specialization and snowmobilers’ perception of risk, MANOVA and an effect size
index Eta (𝜂) were run to test null hypotheses two to ten:

75

H2: There is no relationship between snowmobilers’ recreational specialization and their
perception of operating a snowmobile by person who has been drinking but is not
legally intoxicated (0.01-0.07 blood alcohol).
H3: There is no relationship between snowmobilers ’recreational specialization and their
perception of operating a snowmobile by a legally intoxicated person (0.08 or higher
blood alcohol).
H4: There is no relationship between snowmobilers ’recreational specialization and their
perception of speed of snowmobile.
H5: There is no relationship between snowmobilers ’recreational specialization and their
perception of driver lacking skill in operating machine.
H6: There is no relationship between snowmobilers ’recreational specialization and their
perception of cars/trucks on seasonal roads.
H7: There is no relationship between snowmobilers ’recreational specialization and their
perception of public trail conditions.
H8: There is no relationship between snowmobilers ’recreational specialization and their
perception of public trail design.
H9: There is no relationship between snowmobilers ’recreational specialization and their
perception of other uses of designated snowmobile trails (e.g., dog sledding, cross
country skiing).

76

H10: There is no relationship between snowmobilers ’recreational specialization and their
perception of snowmobiling on county/state road shoulders.

The MANOVA was run to determine the effect of the level of recreational specialization
on Michigan snowmobilers’ perception of risks. Nine measures of perceived risk were assessed:
operation of snowmobile by person who has been drinking but is not legally intoxicated (0.010.07 blood alcohol); operation of snowmobile by a legally intoxicated person (0.08 or higher
blood alcohol); speed of snowmobile; driver lacking skill in operating machine; cars/trucks on
seasonal roads; public trail conditions; public trail design; other uses of designated snowmobile
trails (e.g., dog sledding, cross country skiing); and snowmobiling on county/state road
shoulders. Snowmobilers were categorized into three subgroups based on their recreational
specialization: Novice, Intermediate, and Expert.
Multiple preliminary tests were run to make sure the underlying assumptions of one-way
MANOVA were met. To test the multivariate normality assumption, Shapiro-Wilk's test (p >
.05) was performed on all nine perceptions of risk. Despite the fact that the normality assumption
was violated, analysis proceeded because: 1) One-way MANOVA is fairly robust to deviations
from normality, particularly if the sample size in each subgroup is equal, or nearly equal, as is
the case here (Lix, Keselman & Keselman, 1996). 2) The sample size of each recreational
specialization subgroup (n = 269) is fairly large, thus the skewed distributions are not
problematic (Sawilowsky & Blair, 1992). Further, the result of detecting outliers showed that
there were no extreme univariate outliers in the data, as assessed by inspection of a boxplot.
However, six cases were defined as multivariate outliers as assessed by Mahalanobis distance (p
> .001). To meet the assumption, those six cases were omitted from the analyses that followed.

77

For the assumption of multicollinearity, Pearson’s correlation showed that the assumption was
not violated. Finally, there was homogeneity of variance-covariance matrices, as assessed by
Box's M test of equality of covariance matrices (p = .363).
Overall, novice, intermediate, and expert snowmobilers shared very similar perceptions
toward nine snowmobile riding risks. They considered “Operation of snowmobile by a legally
intoxicated person” (4.4, 4.4, and 4.4), “Driver lacking skill in operating machine” (4.0, 4.1, and
4.1), and “Speed of snowmobile” (3.9, 3.8, and 3.6) highly dangerous. A further review showed
that novice snowmobilers perceived slightly lower danger on “Driver lacking skill in operating
machine” compared to intermediate and expert snowmobilers. Conversely, when asked their
perceptions on “Speed of snowmobile,” less specialized snowmobilers generally perceived
higher risk than those more specialized snowmobilers.
Michigan snowmobilers considered “Operation of snowmobile by person who has been
drinking but is not legally intoxicated” (3.2, 3.3, and 3.3), (2.9, 3.3, and 3.3), and “Other uses of
designated snowmobile trails” (2.7, 2.9, and 2.8) moderately dangerous. Novice snowmobilers
perceived less risk when “Operation of snowmobile by person who has been drinking but is not
legally intoxicated” and “Cars/trucks on seasonal roads” compared to intermediate and expert
snowmobilers. A mixed perception of risk was found among snowmobilers when asked to rate
“Other uses of designated snowmobile trails.” Intermediate snowmobilers perceived the highest
risk, while novice snowmobilers perceived the least.

78

Table 14 Michigan Snowmobilers’ Perception of risks by Specialization Subgroup
Perception of Risk
Specialization Subgroups
Mean
Novice
4.4
Operation of snowmobile by a legally
intoxicated person (0.08 or higher blood
Intermediate
4.4
alcohol)
Expert
4.4
Novice
3.2
Operation of snowmobile by person who has
been drinking but is not legally intoxicated
Intermediate
3.3
(0.01-0.07 blood alcohol)
Expert
3.3
Novice
3.9
Speed of snowmobile
Intermediate
3.8
Expert
3.6
Novice
4.0
Driver lacking skill in operating machine
Intermediate
4.1
Expert
4.1
Novice
2.9
Cars/trucks on seasonal roads
Intermediate
3.2
Expert
3.2
Novice
2.4
Public trail conditions
Intermediate
2.4
Expert
2.3
Novice
2.2
Public trail design
Intermediate
2.2
Expert
2.0
Novice
2.7
Other uses of designated snowmobile trails
Intermediate
2.9
(e.g., dog sledding, cross country skiing)
Expert
2.8
Novice
2.2
Snowmobiling on county/state road shoulders
Intermediate
2.2
Expert
2.2

S.D.
1.0
1.0
.9
1.4
1.4
1.4
1.1
1.1
1.1
1.1
1.0
1.0
1.2
1.1
1.2
1.0
1.1
1.0
1.0
1.0
1.0
1.2
1.3
1.4
1.0
1.1
1.0

Lastly, novice, intermediate, and expert snowmobilers considered “Public trail
conditions” (2.4, 2.4, and 2.3), “Snowmobiling on county/state road shoulders” (2.2, 2.2, and
2.2), and “Public trail design” (2.2, 2.2, 2.0) only slightly dangerous. Table 14 presents the
descriptive statistics of nine perceptions of risk among the three snowmobile recreational
specialization subgroups.
The MANOVA results showed that there was an overall statistically significant
difference among the snowmobile recreational specialization subgroups, F (9, 791) = 2.136, p =

79

.0004; Wilks' Λ = .953; partial 𝜂2 = .024. Follow-up univariate ANOVAs found that “Speed of
snowmobile” (F (2, 798) = 5.024, p = .007) and “Cars/trucks on seasonal roads” (F (2, 798) =
5.109, p = .006) were significantly different among the snowmobilers from different
specialization subgroups, using a Bonferroni test with an adjusted α level of .025.
The Tukey post-hoc tests showed that novice snowmobilers perceived significantly
higher risk on “Speed of snowmobile” than expert snowmobilers (p = .005). No statistical
significant differences were found between novice and intermediate snowmobilers (p = .426) or
intermediate and expert snowmobilers (p = .139). For “Cars/trucks on seasonal roads,” post-hoc
tests indicated that novice snowmobilers perceived significantly lower risk than either
intermediate (p = .017) or expert snowmobilers (p = .015). However, no statistically significant
difference was found between intermediate and expert snowmobilers (p = .999).
The effect size index Eta (𝜂) was run to assess the relationship between recreational
specialization and nine perceptions of risk. Overall, the effects of recreational specialization on
perception of risk were very weak. A negative weak relationship was found between recreational
specialization and the perceptions of “Speed of snowmobile” (𝜂 = .112), while a positive weak
relationship was found between recreational specialization and “Cars/trucks on seasonal roads”
(𝜂 = .112). Partial eta squared (𝜂𝑝2 ) indicated that recreational specialization only explained 1.2%

of the variance and the associated errors of “Speed of snowmobile” (𝜂𝑝2 = .012) and 1.3% of the

variance and the associated errors of “Cars/trucks on seasonal roads” (𝜂𝑝2 = .013). For the other
seven risk items, recreational specialization explained 0.2% of the variance of “Operation of

snowmobile by person who has been drinking but is not legally intoxicated,” 0.1% of “Operation
of snowmobile by a legally intoxicated person,” 0.6% of “Driver lacking skill in operating

80

machine,” 0.1% of “Public trail conditions,” 0.3% of “Public trail design,” 0.3% of “Other uses
of designated snowmobile trails,” and less than 0.01% of “Snowmobiling on county/state road
shoulders.” Table 15 presents the comparison result for nine perceptions of risks across three
levels of recreational specialization.
Table 15 Comparison of Perception of Risk Across Specialization Subgroups
Perception of Risks
Specialization Subgroups
𝜂𝑝2
Variable
Novice Intermediate Expert F (sig.) 𝜂
Operation of snowmobile by person who
has been drinking but is not legally
.852 .046 .002
intoxicated
Operation of snowmobile by a legally
.359 .030 .001
intoxicated person
Speed of snowmobile
a
a
5.024* .112 .012
Driver lacking skill in operating machine
2.292 .076 .006
Cars/trucks on seasonal roads
ab
a
b
5.109* .112 .013
Public trail conditions
.327 .029 .001
Public trail design
1.121 .053 .003
Other uses of designated snowmobile trails
1.146 .054 .003
Snowmobiling on county/state road
.138 .019 .000
shoulders
a,b,ab Groups with a same letter were significantly different at the .05 level of confidence
* = Significant at the .05 level
The MONOVA and effect size index Eta results showed no evidence of a significant
relationship existing between levels of recreational specialization and the risk perception of
operating a snowmobile by a person who has been drinking but is not legally intoxicated.
Therefore, the research failed to reject the null hypothesis two and concluded that there is no
relationship between level of recreational specialization and the risk perceptions of “operating a
snowmobile by person who has been drinking but is not legally intoxicated (0.01-0.07 blood
alcohol).”
The MONOVA and effect size index Eta results showed no evidence of a significant
relationship exist between levels of recreational specialization and the risk perception of
81

operating a snowmobile by a legally intoxicated person. Therefore, the research failed to reject
the null hypothesis three and concluded that there is no relationship between level of recreational
specialization and the risk perceptions of “operating a snowmobile by a legally intoxicated
person (0.08 or higher blood alcohol).”
The MONOVA and effect size index Eta results showed a significant relationship exists
between levels of recreational specialization and the risk perception of speed of the snowmobile.
Levels of recreational specialization explained 1.2% of the variance and the associated errors of
“speed of snowmobile.” Therefore, the research rejected the null hypothesis four and concluded
that there is a negative weak relationship between level of recreational specialization and the risk
perceptions of “speed of snowmobile.”
The results from MONOVA and effect size index Eta tests showed no evidence of a
significant relationship exist between levels of recreational specialization and the risk perception
of a driver lacking skill in operating his/her machine. Therefore, the research failed to reject the
null hypothesis five and concluded that there is no relationship between level of recreational
specialization and the risk perceptions of “driver lacking skill in operating machine.”
The MONOVA and effect size index Eta results showed a significant relationship exist
between levels of recreational specialization and the risk perception of cars/truck on seasonal
roads. Levels of recreational specialization explained 1.3% of the variance and the associated
errors of “cars/truck on seasonal roads.” Therefore, the research rejected the null hypothesis six
and concluded that there is a positive weak relationship between level of recreational
specialization and the risk perceptions of “cars/truck on seasonal roads.”

82

The results from MONOVA and effect size index Eta tests showed no evidence of a
significant relationship exist between levels of recreational specialization and the risk perception
of public trail conditions. Therefore, the research failed to reject the null hypothesis seven and
concluded that there is no relationship between level of recreational specialization and the risk
perceptions of “public trail conditions.”
The MONOVA and effect size index Eta results showed no evidence of a significant
relationship exist between levels of recreational specialization and the risk perception of public
trail design. Therefore, the research failed to reject the null hypothesis eight and concluded that
there is no relationship between level of recreational specialization and the risk perceptions of
“public trail design.”
The results from MONOVA and effect size index Eta tests showed no evidence of a
significant relationship exist between levels of recreational specialization and the risk perception
of other uses of designated snowmobile trails. Therefore, the research failed to reject the null
hypothesis nine and concluded that there is no relationship between level of recreational
specialization and the risk perceptions of “other uses of designated snowmobile trails (e.g., dog
sledding, cross country skiing).”
The MONOVA and effect size index Eta results showed no evidence of a significant
relationship between levels of recreational specialization and the risk perception of
snowmobiling on county/state road shoulders. Therefore, the research failed to reject the null
hypothesis ten and concluded that there is no relationship between level of recreational
specialization and the risk perceptions of “snowmobiling on county/state road shoulders.”

83

In sum, null hypotheses two, three, five, seven, eight, and nine were rejected. However,
there was no meaningful influence of recreational specialization on snowmobilers’ perception of
risks. Despite the characteristic differences among novice, intermediate, and expert
snowmobilers, there were minimal differences in key perceived snowmobiling risks. The only
statistically significant differences among snowmobiler subgroups were found on perceptions of
“Speed of snowmobile” and “Cars/trucks on seasonal roads.” In general, those categorized as
novice perceived the highest risk on “Speed of the snowmobile” compared to expert
snowmobilers. When asked about the perceived risk on “Cars/trucks on seasonal roads” while
snowmobiling, novice snowmobilers perceived significantly lower risk compared to intermediate
or expert snowmobilers. No significant difference was found between intermediate and expert
snowmobilers. Moreover, a negative weak relationship was found between levels of recreational
specialization and the risk perception “Speed of snowmobile” and a positive weak relationship
was found between levels of specialization and the risk perception “Cars/trucks on seasonal
roads.”
Influence of Specialization on Spatial Interaction
The literature suggests an association between leisure participation and spatial
interaction. Since the level of recreational specialization measured in this research is heavily
based on recreationists’ snowmobiling participation, level of specialization is assumed to be
correlated with spatial interaction between snowmobilers’ residential origin and destination, the
snowmobile trail network. To address research question three: What are the relationships
between levels of recreational specialization and snowmobilers’ spatial interaction with
snowmobile trail network, ANOVA and an effect size index Eta (𝜂) were run to test null
hypothesis 11:
84

H11: There is no relationship between levels of recreational specialization and
snowmobilers’ spatial interaction with the snowmobile trail network.
Spatial Interaction Result by Destination Regions
One-way ANOVA was used to test differences across levels of specialization in
snowmobilers’ spatial interaction with the snowmobile trail network. A boxplot and ShapiroWilk’s normality test were performed to assess if the underlying assumptions of one-way
ANOVA were met. The boxplot result showed that there were extreme outliers in all three
Michigan snowmobile recreational specialization subgroups. The normality test also revealed
that spatial interaction data in each subgroup were not normally distributed (p < .05). Therefore,
the non-parametric equivalent of the Kruskal-Wallis test was run to decide if there were
differences among novice, intermediate, and expert snowmobilers. Statistical significance was
accepted at the p < .05 level for the omnibus test and p < .0167 level for the multiple
comparisons (.05 divided by 3). A post-hoc analysis was performed using Dunn's (1964)
procedure with a Bonferroni correction for multiple comparisons.
Spatial interaction measures the flow between locations; therefore the results of spatial
interaction between Michigan snowmobilers and their five destination regions were presented
using the summation value instead of the mean in the following paragraphs. Additionally, since
the spatial interaction results were highly skewed and not normally distributed, the mean rank
scores were utilized instead of mean score to report the strength of spatial interaction. The SPSS
program did not generate the mean rank scores for those destination regions with no significant
differences among specialization subgroups. A higher ranking score indicates a stronger spatial
interaction between origin and destination.

85

Table 16 Spatial Interaction between Michigan Snowmobilers and Five Regions
Destination Region Specialization Subgroups Mean Rank Median
Sum
Novice
388.67
.0
2821.9
Region 1*
Intermediate
388.65
.0
2455.4
Expert
434.68
.0
4051.9
Novice
366.84
.0
2552.2
Region 2*
Intermediate
377.55
.0
24173.1
Expert
467.62
.5
5465.7
Novice
N/A
.5
6205.3
Region 3
Intermediate
N/A
2.6
12380.0
Expert
N/A
2.0
6464.5
Novice
375.88
.0
3692.7
Region 4*
Intermediate
423.99
.0
4720.9
Expert
412.12
.0
57333.5
Novice
N/A
.0
2423.2
Region 5
Intermediate
N/A
.0
2737.2
Expert
N/A
.0
1551.4
* = Significant at the .05 level

S.D.
112.6
93.8
206.7
89.5
1299.2
155.4
88.0
512.2
133.2
100.2
71.6
3189.7
68.4
76.5
28.5

The Kruskal-Wallis result showed that spatial interaction between Michigan
snowmobilers and region one (χ2 (2) = 14.363, p = .001), two (χ2 (2) = 39.510, p < .0005), and
four (χ2 (2) = 7.970, p = .019) were significantly different among novice, intermediate, and
expert snowmobilers. Table 16 shows the descriptive result for spatial interaction between
Michigan snowmobilers and five destination regions.
For interaction between all snowmobilers and region one, post-hoc analysis revealed
statistically significant differences between expert snowmobilers and novice snowmobilers (p =
.003), and expert and intermediate snowmobilers (p = .003). The mean rank scores showed that
expert snowmobilers had significant higher spatial interaction with the snowmobile trail network
in region one compared to novice and intermediate snowmobilers. There was no statistically
significant difference between novice and intermediate snowmobilers (p = 1.000).
For spatial interaction between snowmobilers and region two, post-hoc analysis showed
that there were significant differences between expert and novice snowmobilers (p < .0005), and
86

expert and intermediate snowmobilers (p < .0005). The mean rank scores showed that expert
snowmobilers had significant higher spatial interaction with the snowmobile trail network in
region two compared to intermediate and novice snowmobilers. No significant differences were
found between novice and intermediate snowmobilers (p = 1.000).
Finally, for spatial interaction between snowmobilers and region four, post-hoc analysis
showed that there were statistically significant differences between novice and intermediate
snowmobilers (p = .020). The mean rank scores showed that intermediate snowmobilers had the
highest spatial interaction with the snowmobile trail network in region four and the interaction
was significantly higher than that of novice snowmobilers. No significant differences were found
between expert and novice (p = .124) or expert and intermediate snowmobilers (p = 1.000).
Table 17 presents the comparison result of spatial interaction among novice, intermediate, and
expert snowmobilers.
Table 17 Comparison Result of Spatial Interaction Among Specialization Subgroups
Specialization Subgroups
Destination Region
Novice
Intermediate
Expert
χ2 (sig.)
Spatial Interaction with Region 1
a
b
ab
14.636*
Spatial Interaction with Region 2
a
b
ab
39.510*
Spatial Interaction with Region 3
2.382
Spatial Interaction with Region 4
a
a
7.970*
Spatial Interaction with Region 5
.338
a,b,ab Groups with a same letter were significantly different at the .05 level of confidence.
* = Significant at the .05 level
Spatial Interaction Result by Snowmobilers’ Residential Regions
Research has indicated that the spatial structure of locations varies from one origin to
another (Fesenmaier & Lieber, 1985). The spatial interaction between Michigan snowmobilers
and the trail network in each destination region was examined by dividing snowmobilers based
on their residential regions. The non-parametric equivalent of Kruskal-Wallis test as well as post87

hoc analysis were again run for each residential region to decide if there were differences among
novice, intermediate, and expert snowmobilers.
Table 18 Spatial Interaction between Region One and Two Snowmobilers and Trail Network
Residential
Destination
Recreational
Median
Sum
S.D.
Region
Region
Specialization
Region 1
Region 1
Novice
66.4
2768.9
486.5
(western Upper
Intermediate
77.0
2425.5
446.4
Peninsula)
Expert
67.2
3980.8
1167.1
Region 2
Novice
.0
10.6
2.6
Intermediate
.0
7.2
2.0
Expert
.3
11.5
2.0
Region 3
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Region 4
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Region 5
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Region 2
Region 1
Novice
.0
23.6
6.8
(eastern Upper
Intermediate
.0
8.5
3.0
Peninsula)
Expert
.0
34.0
6.7
Region 2
Novice
87.2
2236.2
448.6
Intermediate
34.9
23926.2
7396.6
Expert
70.2
4989.1
623.2
Region 3
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Region 4
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Region 5
Novice
.0
.0
.0
Intermediate
.0
.0
.0
Expert
.0
.0
.0
Based on the Kruskal-Wallis test results, there were no statistically significant differences
among specialization subgroups on the spatial interaction between either region one or region
two snowmobilers and their destination regions. However, statistically significant differences
were found for region three, four, and five snowmobilers’ spatial interaction with different
88

destinations. Table 18 shows the descriptive result of spatial interaction between region one and
region two Michigan snowmobilers and their five destination regions.
For the relationship between recreational specialization and spatial interaction, the result
of the effect size index Eta (𝜂) indicated that there were weak relationships between the two
concepts when snowmobilers who lived in region one travel to region two or stayed in their
residential region. For snowmobilers who lived in region two, there was a weak relationship
when they traveled to region one for snowmobiling. There was a moderate relationship between
recreational specialization and spatial interaction when region two snowmobilers staying at their
own residential region for the activity. Overall, recreational specialization accounted for 10% or
less of the spatial interaction variance for snowmobilers lived in the Upper Peninsula. Table 19
shows the effect size index results of the relationship between region one and region two
snowmobile recreational specialization and their spatial interaction.
Table 19 Relationship between Region One and Two Specialization and Spatial Interaction
Residential Region
Destination Region
𝜂
𝜂2
Region 1
Region 1
.174
.03
(western Upper Peninsula)
Region 2
.134
.01
Region 2
Region 1
.121
.01
(eastern Upper Peninsula)
Region 2
.311
.10
For snowmobilers lived in region three, Kruskal-Wallis results showed that there were
statistically significant differences on spatial interaction among specialization subgroups when
they travel to region two (χ2 (2) = 10.718, p = .005) and region four (χ2 (2) = 9.060, p = .011) for
snowmobiling. For region three snowmobilers who went to region two to snowmobile, post-hoc
analysis showed that there were statistically significant differences between expert snowmobilers
and novice snowmobilers (p = .033), and expert and intermediate snowmobilers (p = .007). Since
the spatial interaction results were highly skewed and not normally distributed, the mean rank
89

scores were utilized instead of mean score to report the strength of spatial interaction. The SPSS
program did not generate the mean rank scores for those destination regions with no significant
differences among specialization subgroups. A higher ranking score indicates a stronger spatial
interaction between origin and destination.
Table 20 Spatial Interaction between Region Three Snowmobilers and Trail Network
Residential
Destination Recreational
Mean
Median
Sum
S.D.
Region
Region
Specialization
Rank
Region 3
Region 1
Novice
N/A
.0
6.1
.9
(northwestern
Intermediate
N/A
.0
2.9
.4
Lower Peninsula)
Expert
N/A
.0
6.8
.6
Region 2* Novice
37.56
.0
114.8
12.5
Intermediate
35.54
.0
45.9
6.8
Expert
50.66
.7
158.4
8.6
Region 3
Novice
N/A
122.5
4687.0
231.2
Intermediate
N/A
45.7 10654.8
1572.0
Expert
N/A
22.8
4715.2
381.2
Region 4* Novice
37.00
.0
.0
.0
Intermediate
47.25
.0
84.8
7.8
Expert
39.83
.0
19.1
3.0
Region 5
Novice
N/A
.0
.0
.0
Intermediate
N/A
.0
.0
.1
Expert
N/A
.0
.0
.0
* = Significant at the .05 level
Results showed that expert snowmobilers had significantly higher spatial interaction
compared to novice and intermediate snowmobilers. No apparent spatial interaction difference
was found between intermediate and novice snowmobilers. For region three snowmobilers who
went to region four to snowmobile, post-hoc analysis showed that there was a statistically
significant difference between novice and intermediate snowmobilers (p = .012). The mean rank
scores showed that intermediate snowmobilers from region three had the highest spatial
interaction with the trail system in region four compared to novice and expert snowmobilers. No
statistically significant differences were found among other groups. Table 20 presents the

90

descriptive result for spatial interaction among three specialization subgroups from region three
and snowmobile trail networks by five regions.
Additionally, the effect size index Eta results showed that there were weak relationships
between region three snowmobilers’ recreational specialization and their spatial interaction with
regions two, three, and five. A moderate relationship was found between recreational
specialization and spatial interaction with region four. Eta squared showed that less than 7% of
the spatial interaction variance was explained by recreational specialization. Table 21 presents
the comparison result of spatial interaction between region three specialization subgroups and the
five destination regions.
Table 21 Comparison Result of Spatial Interaction for Region Three Specialization Subgroups
Region 3 Specialization Subgroups
Destination Region
Novice Intermediate Expert
χ2 (sig.)
𝜂
𝜂2
Spatial Interaction with Region 1
2.090
.098
.01
Spatial Interaction with Region 2
a
b
ab
10.718* .176
.03
Spatial Interaction with Region 3
3.945
.104
.01
Spatial Interaction with Region 4
a
a
9.060* .260
.07
Spatial Interaction with Region 5
1.929
.154
.02
a,b,ab Groups with a same letter were significantly different at the .05 level of confidence
* = Significant at the .05 level
For snowmobilers who resided in region four, Kruskal-Wallis results showed that there
were statistically significant differences on spatial interaction among recreational specialization
subgroups when they travel to region one (χ2 (2) = 9.638, p = .008) and region four (χ2 (2) =
7.065, p = .029). For region four snowmobilers who went to region one to snowmobile, post-hoc
analysis showed that there were statistically significant differences between expert snowmobilers
and novice snowmobilers (p = .025), and expert and intermediate snowmobilers (p = .021). Since
the spatial interaction results were highly skewed and not normally distributed, the mean rank
scores were utilized instead of mean score to report the strength of spatial interaction. The SPSS
91

program did not generate the mean rank scores for those destination regions with no significant
differences among specialization subgroups. A higher ranking score indicates a stronger spatial
interaction between origin and destination. Table 22 presents the descriptive statistics for spatial
interaction between three specialization subgroups from region four and snowmobile trail
networks by five regions.
Table 22 Spatial Interaction between Region Four Snowmobilers and Trail Network
Residential
Destination Recreational
Mean
Median
Sum
S.D.
Region
Region
Specialization
Rank
Region 4
Region 1* Novice
43.50
.0
.0
.0
(Northeastern
Intermediate
43.50
.0
.0
.0
Lower Peninsula)
Expert
50.61
.0
3.3
.3
Region 2
Novice
N/A
.0
33.1
3.7
Intermediate
N/A
.0
59.9
5.6
Expert
N/A
.0
98.1
5.8
Region 3
Novice
N/A
.0
302.4
17.1
Intermediate
N/A
.0
197.8
11.7
Expert
N/A
.0
132.7
8.9
Region 4* Novice
35.09
32.8
3182.0
296.3
Intermediate
52.05
71.5
1113.7
173.4
Expert
49.69
68.3 56956.4
9223.8
Region 5
Novice
N/A
.0
.0
.0
Intermediate
N/A
.0
.0
.0
Expert
N/A
.0
.0
.0
* = Significant at the .05 level
The mean rank scores showed that expert snowmobilers had significantly higher spatial
interaction than either novice or intermediate snowmobilers when traveling to region one to
snowmobile. For region four snowmobilers riding in their own residential region, post-hoc
analysis showed that there was a statistically significant difference between novice and
intermediate snowmobilers (p = .041). Intermediate snowmobilers had significantly higher
spatial interaction than novice snowmobilers when snowmobiling in region four. No other
statistically significant differences were found among other groups.

92

The effect size index Eta result showed that there were weak relationships between
region four snowmobilers’ recreational specialization and their spatial interaction with region
two, three, and four. A moderate relationship was found between recreational specialization and
spatial interaction with region one. Eta squared showed that less than 9% of the spatial
interaction variance was explained by recreational specialization. Table 23 presents the
comparison result of spatial interaction between region four specialization subgroups and five
destination regions.
Table 23 Comparison Result of Spatial Interaction for Region Four Specialization Subgroups
Region 4 Specialization Subgroups
Destination Region
Novice Intermediate Expert χ2 (sig.)
𝜂
𝜂2
Spatial Interaction with Region 1
a
b
ab
9.638* .302
.09
Spatial Interaction with Region 2
5.700
.150
.02
Spatial Interaction with Region 3
1.087
.211
.04
Spatial Interaction with Region 4
a
a
7.065* .145
.02
Spatial Interaction with Region 5
< 0.0005
a,b,ab Groups with a same letter were significantly different at the .05 level of confidence
* = Significant at the .05 level
For snowmobilers resided in region five, Kruskal-Wallis result showed that there were
statistically significant differences on spatial interaction among recreational specialization
subgroups when they travel to region one (χ2 (2) = 15.363, p < .0005) and region two (χ2 (2) =
22.751, p < .0005) for snowmobiling. For region five snowmobilers who went to region one for
snowmobiling, post-hoc analysis showed that there were statistically significant differences
between expert and novice (p = .001), and expert and intermediate snowmobilers (p = .010). For
region four snowmobilers who went to region one for snowmobiling, post-hoc analysis showed
that there were statistically significant differences between expert and novice (p = .025), and
expert and intermediate snowmobilers (p = .021). Since the spatial interaction results were highly
skewed and not normally distributed, the mean rank scores were utilized instead of mean score to

93

report the strength of spatial interaction. The SPSS program did not generate the mean rank
scores for those destination regions with no significant differences among specialization
subgroups. A higher ranking score indicates a stronger spatial interaction between origin and
destination. Table 24 presents the descriptive statistics for spatial interaction between three
specialization subgroups from region four and snowmobile trail networks by five regions.
Table 24 Spatial Interaction between Region Five Snowmobilers and Trail Network
Residential
Destination Recreational
Mean
Median
Sum
Region
Region
Specialization
Rank
Region 5
Region 1* Novice
270.21
.0
23.3
(Northeastern
Intermediate
279.20
.0
18.5
Lower Peninsula)
Expert
312.42
.0
27.2
Region 2* Novice
263.28
.0
18.5
Intermediate
270.52
.0
133.9
Expert
328.53
.5
208.7
Region 3
Novice
N/A
1.2
1215.9
Intermediate
N/A
2.8
1527.4
Expert
N/A
2.1
1616.7
Region 4
Novice
N/A
.0
510.8
Intermediate
N/A
.0
522.4
Expert
N/A
.0
358.0
Region 5
Novice
N/A
.0
2423.2
Intermediate
N/A
.0
2736.9
Expert
N/A
.0
1551.4
* = Significant at the .05 level

S.D.
.4
.3
.3
.3
1.3
1.6
14.3
20.4
17.6
5.6
5.1
3.3
80.3
90.3
33.9

The mean rank scores indicated that expert snowmobilers had statistically significantly
higher spatial interaction than either intermediate or novice snowmobilers when traveling to
region one for snowmobiling. For region five snowmobilers who went to region two for
snowmobiling, post-hoc analysis also showed that there were statistically significant differences
between expert and novice (p < .0005), and expert and intermediate snowmobilers (p < .0005).
The mean rank scores indicated that the expert snowmobilers had statistically significantly
higher spatial interaction than either intermediate or novice snowmobilers when traveling to
region two for snowmobiling.
94

The effect size index Eta result showed that there were weak relationships between
region five snowmobilers’ recreational specialization and their spatial interaction with region
two. Eta squared showed that less than 1% of the spatial interaction variance was explained by
recreational specialization. Table 25 presents the comparison result of spatial interaction between
region four specialization subgroups and five destination regions.
Table 25 Comparison Result of Spatial Interaction for Region Five Specialization Subgroups
Region 5 Specialization Subgroups
Destination Region
Novice Intermediate Expert χ2 (sig.) 𝜂
𝜂2
Spatial Interaction with Region 1
a
b
ab
15.363* .057 > .00
Spatial Interaction with Region 2
a
b
ab
22.751* .114
.01
Spatial Interaction with Region 3
4.034 .057 > .00
Spatial Interaction with Region 4
1.467 .075
.01
Spatial Interaction with Region 5
.302 .034 > .00
a,b,ab Groups with a same letter were significantly different at the .05 level of confidence
* = Significant at the .05 level

Section Summary
According to the results of the Kruskal-Wallis and effect size Eta tests, evidence of
statistical differences in spatial interaction was found across levels of recreational specialization.
Therefore, the research rejected null hypothesis 11 and concluded that there were some minimal
to moderate relationships between different levels of specialization and snowmobilers’ spatial
interaction with the snowmobile trail network.
Statistically significant differences were found across novice, intermediate, and expert
snowmobilers on the spatial interaction between all Michigan snowmobilers and the snowmobile
trail network in regions one, two, and four. When segmenting snowmobilers by their residential
regions, statistically significant differences among levels of specialization were found on spatial
interaction between region three snowmobilers and the snowmobile trail network in region two
and four; region four snowmobilers and the snowmobile trail in regions one and four; and region
95

five snowmobilers and the snowmobile trail in regions one and two. The relationship between
specialized snowmobilers from different regions and their spatial interaction with the
snowmobile trail network, indicated by Eta, ranged from 0.034 to 0.311, showing a minimal
relationship to a moderate relationship (0.10 – 0.242 = weak relationship; 0.243 – 0.370 =
moderate relationship; 0.371 or greater = strong relationships).
When segmenting snowmobilers by their residential regions, no spatial interaction was
found between the Upper and Lower Peninsulas for those who lived in the Upper Peninsula.
However, moderate relationships were found for region two snowmobilers riding sleds in their
own region; region three snowmobilers riding in region four; and region four snowmobilers
traveling to region one. Levels of recreational specialization explained at least 7% of the
variance of spatial interaction. Weak relationships were found among various travel patterns, for
instance, when region five snowmobilers traveled to region two and when region four
snowmobilers traveled to region three for snowmobiling. In summary, levels of recreational
specialization explained 1% to 4% of the variance of spatial interaction, making for a minimal
meaningful difference.

96

Chapter Five
Discussion, Implications, and Conclusions

Research Background
Snowmobiling is a winter recreation activity that contributes substantially to Michigan’s
economy through travel and equipment expenditures by snowmobilers. Snowmobiles, originally
developed for over-snow transportation purposes, are now used by snowmobilers mostly for
enjoyment, including the thrill of high speed travel over the snow. This research explores the
within-group social world of Michigan snowmobilers using the well-established theoretical
framework of recreational specialization, as well as two topics of interest associated with
snowmobiling. The first is snowmobile fatalities. Law enforcement and recreation program
managers seek to provide a safe riding environment and quality recreation experiences for all
snowmobiling participants. To help reduce the incidence and severity of snowmobile accidents,
snowmobilers’ perceptions of key riding risks were assessed and potential links with recreational
specialization were examined, in the hopes of discovering recognizable patterns among
segments to best target safe riding education and training intervention programs, as well as law
enforcement management actions. Secondly, for the recreational specialization construct, the
research explored the determinants that facilitate or constrain the advancement of recreational
specialization in terms of the spatial construct of resource availability and accessibility. This
included the proximity and quantity of the recreation resource, both in terms of trail mileage and
quantity of snow. In sum, the objectives of this research were to describe and compare different
levels of recreational specialization among Michigan snowmobilers; examine the influence of
recreational specialization on snowmobilers’ perception of riding risks; and assess the

97

relationships between snowmobilers’ recreational specialization and spatial interaction with the
Michigan snowmobile trail network.

Discussion
Three research objectives are addressed and discussed in this section: Michigan
snowmobilers’ within-group social world, the influence of recreational specialization on
perceptions of risk, and the relationships between recreational specialization and snowmobilers’
spatial interaction with the Michigan snowmobile trail network.
Table 26 provides a summary of the tests and outcomes for 11 research hypotheses. This
research found a significant difference in group composition between levels of specialization
using summative indices and cluster analysis. Further, the only significant differences among
snowmobiler subgroups in terms of risk perception were found on perceptions of “Speed of
snowmobile” and “Cars/trucks on seasonal roads.” Recreational specialization not appeared to be
an effective construct in predicting snowmobilers’ perception of risks. Lastly, Michigan
snowmobilers’ spatial interaction with the snowmobile trail system showed that different levels
of specialized snowmobilers had their preferred destinations. There is a weak to moderate
statistically significant relationship between recreational specialization and spatial interaction
with Michigan snowmobile trail system.

98

Table 26 Research Hypotheses, Tests, and Outcome
Null Hypotheses
H1: There is no difference in group composition between
levels of specialization using summative indices and cluster
analysis.
H2: There is no relationship between snowmobilers’
recreational specialization and their perception of operating a
snowmobile by person who has been drinking but is not
legally intoxicated (0.01-0.07 blood alcohol).
H3: There is no relationship between snowmobilers
’recreational specialization and their perception of operating
a snowmobile by a legally intoxicated person (0.08 or higher
blood alcohol).
H4: There is no relationship between snowmobilers
’recreational specialization and their perception of speed of
snowmobile.
H5: There is no relationship between snowmobilers
’recreational specialization and their perception of driver
lacking skill in operating machine.
H6: There is no relationship between snowmobilers
’recreational specialization and their perception of
cars/trucks on seasonal roads.
H7: There is no relationship between snowmobilers
’recreational specialization and their perception of public
trail conditions.
H8: There is no relationship between snowmobilers
’recreational specialization and their perception of public
trail design.
H9: There is no relationship between snowmobilers
’recreational specialization and their perception of other uses
of designated snowmobile trails (e.g., dog sledding, cross
country skiing).
H10: There is no relationship between snowmobilers
’recreational specialization and their perception of
snowmobiling on county/state road shoulders.
H11: There is no relationship between levels of recreational
specialization and snowmobilers’ spatial interaction with the
snowmobile trail network.

99

Tests
ANOVA

Outcome
Rejected

MANOVA
ANOVA
Effect Size Eta (𝜂)

Failed to
reject

MANOVA
ANOVA
Effect Size Eta (𝜂)

Failed to
reject

MANOVA
ANOVA
Effect Size Eta (𝜂)
MANOVA
ANOVA
Effect Size Eta (𝜂)
MANOVA
ANOVA
Effect Size Eta (𝜂)
MANOVA
ANOVA
Effect Size Eta (𝜂)

Failed to
reject
Rejected

Failed to
reject

MANOVA
ANOVA
Effect Size Eta (𝜂)
MANOVA
ANOVA
Effect Size Eta (𝜂)

Failed to
reject

MANOVA
ANOVA
Effect Size Eta (𝜂)

Failed to
reject

Kruskal-Wallis
Effect Size Eta (𝜂)

Rejected

Failed to
reject

Michigan Snowmobilers’ Within-group Social World
The theoretical construct of recreational specialization was applied to study the withingroup social world of Michigan snowmobilers. Two common classification approaches,
summative index and cluster analysis, were utilized to identify different levels of specialization
in snowmobilers. Consistent with previous studies on anglers (Oh & Ditton, 2006), skiers and
snowboarders (Needham & Little, 2013), off-highway vehicle (OHV) riders (Smith, Burr, &
Reiter, 2010), and many other outdoor recreationalists, different levels of recreational
specialization were successfully identified with snowmobilers. The summative index method
generated three snowmobile subgroups, novice, intermediate, and expert, while cluster analysis
provided three different levels of specialization: moderate, big spender, and trail warrior.
Through the analysis, this research showed that Michigan snowmobilers are heterogeneous
differing in their levels of recreational specialization. More specifically, they differ in their
behavioral commitment, knowledge/skill about snowmobiling, and psychological attachment to
snowmobile clubs or organizations.
Unlike many studies, which only used either the summative index or cluster analysis
approach to segment recreationalists without clear justification (e.g., Donnelly et al., 1986; Dyck
et al., 2013; Needham et al., 2007), this research adds to the literature on recreational
specialization by comparing the composition and characteristics of subgroups identified by both
methods. This meets one of the goals of this research. In contrast to the only published research
that compared both methods and found no difference between them in terms of results (Scott et
al., 2005), this research suggested that the summative index method and K-mean cluster analysis
provide dissimilar perspectives of snowmobilers’ within-group social worlds.

100

Using the summative index method, three subgroups – novice, intermediate, and expert –
formed a continuum “from general to particular” that was consistent with what Bryan originally
hypothesized (Bryan, 1977, p. 175). In contrast to Tennessee River users (Kuentzel &
McDonald, 1992) and rock climbers (Ewert & Hollenhorst, 1994), all six variables used to
measure snowmobilers’ recreational specialization appeared to increase in the same direction as
the level of specialization increased. Expert snowmobilers presented more intensive behavioral
commitment (snowmobiles owned, money spent, and snowmobiling experience),
knowledge/skill (snowmobile days and mileage), and psychological attachment (snowmobile
organization membership) to the activity when compared to intermediate and novice
snowmobilers. Intermediate snowmobilers also showed significantly deeper involvement in the
activity compared to novices. The progression of recreational specialization is evident as
commitment and involvement in the activity increased.
Using an alternative K-mean cluster analysis approach, three subgroups, moderates
(65%), big spenders (11%), and trail warriors (24%), were identified. Although there were
statistically significant differences among three subgroups on behavior, skill/knowledge, and
psychological attachment characteristics, disparities between big spender and trail warrior
snowmobilers were less apparent compared to the differences between moderate and both big
spender and trail warrior snowmobilers. Big spenders had a much greater economic commitment
than other groups, while trail warriors rode more days traveling more miles. Similar to the group
composition results in studies of birders (McFarlane, 1996; Scott & Thigpen, 2003), the majority
(77%) of Michigan snowmobilers casually participated in the activity and presented relatively
weak behavioral commitment, skill/knowledge, and psychological attachment to a snowmobiling
club or organization. Two relatively more specialized subgroups, big spenders and trail warriors,
101

had not necessarily progressed along the specialization continuum. Rather they were so named
because they displayed their own particular interests. In a further review of the data, 94% of the
big spenders purchased a snowmobile in the snowmobile season 2007-08. This brought up the
question that those big spenders may have been the trail warriors who simply purchased a
snowmobile in the past 12 months. However, the results also showed that big spenders and trail
warriors were not only significantly different in money spent ($10,398 vs. $1,504), but also
snowmobile days (27 vs. 37) and miles driven (2058 vs. 3048) in 2007-08 winter. A very similar
result was found in a study of birders in Thailand (Hvenegaard, 2002). There the more
specialized birders were further divided into two subgroups and the two advanced subgroups
significantly differed in their economic commitment. These finding are similar to other studies
(McIntyre & Pigram, 1992; Scott & Thigpen, 2003) that suggested recreationalists may place
their priority on one aspect of specialization over another. The trajectories and progression
within each of the dimensions are not necessarily identical and may increase or decrease in
different directions (Needham et al., 2007). Similar comments were addressed by Scott and
Shafer (2001) that some recreationists “continue to participate in activities on a regular basis and
accrue commitments but exhibit little evidence of skill development… other individuals may
participate in leisure activities infrequently but demonstrate a high level of skill development and
personal commitment” (p. 338). The cluster analysis result, different from that found using the
summative method, supported the above studies in that a portion of Michigan snowmobilers
cannot be placed neatly along a continuum from low to high. Big spenders and trail warriors did
not necessarily progress along the continuum in the linear fashion on all six specialization
variables as they displayed distinct specialization characteristics on behavioral involvement in
and skill/knowledge of snowmobiling.

102

The summative index method and K-mean cluster analysis are two commonly used but
fundamentally and mathematically different approaches for categorizing recreationalists. The
subgroups generated from the summative index approach were selected to assess the influence of
recreational specialization on snowmobilers’ perceptions of riding risks and their spatial
interaction with the snowmobile trail network for the following reasons. First, novice,
intermediate, and expert snowmobilers formed a specialization continuum from low to high
which fit the original hypothesis of Bryan (1977). The linear relationships between concepts can
be better explained when the independent variable presented a hierarchical status in terms of
snowmobile participation. Second, the summative index approach generated much more distinct
subgroups compared to the cluster analysis. All six measures of specialization were significantly
different among the subgroups identified by the summative index approach. The between group
differences among the cluster analysis subgroups were less apparent, especially between the two
advanced subgroups, big spenders and trail warriors. This research did not select the
recreationalist classification method arbitrarily. The decision was made based on the comparison
of two commonly used methods - summative index and cluster analysis as well as the purposes
of the study.
It is important to note that the classification method and the six measures of recreational
specialization may potentially constrain the representation of Michigan snowmobilers’ withingroup social world. Both summative index and cluster analysis methods are straightforward and
easily employed, however, there are other less commonly used methods of measuring
recreational specialization and its effect on other behavioral or cognitive constructs. While a
considerable literature suggests that recreational specialization is a multi-dimension
measurement with behavioral, knowledge/skill, and psychological attachment variables, it also
103

notes that self-rated skill level and centrality-to-life were suggested to be the more reliable
indicators of recreational specialization (McFarlane, 2004; McIntyre & Pigram, 1992; Smith et
al., 2010). A simplified, self-classification measure with descriptive statements that reflect
specialization features was also suggested to provide a valid grouping result (Scott et al., 2005).
More sophisticated analysis procedures such as confirmatory factor analysis, structural equation
modeling and discriminant analysis were also suggested to provide meaningful results (Jett et al.,
2009; Lee & Scott, 2004; Needham et al., 2007; Oh et al., 2012). Since this research utilized a
data set that was originally designed to obtain information on behavioral involvement in
snowmobiling and economic impact of snowmobile spending, using different measures or
indicators of behavioral, knowledge/skill, and psychological attachment variables in the future
could provide a richer picture of other aspects of Michigan snowmobilers’ within-group social
world.
Influence of Specialization on the Perception of Risk
Transportation and accident research shows evidence of a strong relationship between
drivers’ personality and perception of risks. Since no known published research studied
perception of risk utilizing the recreational specialization construct, this research turned to
literatures that studied the influence of recreational specialization on recreationalist’ attitudinal
change. Unlike previous studies which showed that increasing recreational specialization
predicts perceived flow experiences and generally supported the positive association of
recreational specialization with recreationists’ attitudes toward environmental restrictions and
resource protection policies (Dyck et al., 2003; Oh & Ditton, 2006; Wöran & Arnberger, 2012),
this research found recreational specialization did not effectively predict snowmobilers’
perception of riding risks. The only statistically significant differences among snowmobiler
104

subgroups were found on perceptions of “Speed of snowmobile” and “Cars/trucks on seasonal
roads,” but with weak relationships between recreational specialization and perception of risk. A
similar result was found in a boater study in “Manatee Zones” in Florida, where recreational
specialization performed poorly in predicting self-reported compliance behavior and restricted
vessel speed (Jett et al., 2009). Possible explanations regarding the weak association between
specialization and perception of risk in this study may be related to how participants interpreted
the risks of different hazardous riding scenarios. While the dangers of driving drunk are well
known, information about almost half of snowmobile fatalities occurring on county roads and
state highways may not be well known across the snowmobiling community (Nelson et al. 1998).
It is also possible that snowmobilers rated their responses to the various underlying components
of risk in a relatively neutral fashion. Given a five-point Likert scale for measuring perceptions,
Bertram (2007) indicated that participants may avoid selecting an extreme value or provide a
more socially acceptable answer rather than being honest.
Despite the weak association, the research found two statistically significant differences
among snowmobiler subgroups on the perceptions of “Speed of snowmobile” and “Cars/trucks
on seasonal roads.” For the purpose of identifying those who tend to underestimate
snowmobiling risks, this research found that expert snowmobilers tended to underestimate the
risk of speeding, while novice snowmobilers perceived low risk of cars/trucks on seasonal roads.
Speeding has been one of the key factors directly associated with Michigan snowmobile
fatalities (Michigan DNR, 2008; 2009; 2010; 2011; 2012; 2013). It is reasonable to assume that
snowmobilers who underestimate this relationship tend to engage in risky, high speed behavior.
Those who were categorized as expert snowmobilers perceived significantly less risk from
speeding compared to novice snowmobilers. Confidence in operating and controlling a
105

snowmobile may result in a reduced perception of risk from speeding. This finding is consistent
with Morgan and Stevens’ (2008) study on SCUBA divers, in that perceived risk decreased
significantly as experience increased. A study on mountaineers’ perception of risk showed
similar results in that individuals with the most experience perceived the lowest level of risk
(Demirhan, 2005). Although his research did not examine the relationship between recreational
specialization and risky riding behaviors, Lynch’s (2000) study on convicted Michigan
snowmobile law violators found those who were convicted of violating snowmobile laws owned
a higher number of snowmobiles, rode more days traveling more miles and perceived speeding
as less dangerous compared to those who were not convicted of violating snowmobile laws.
This finding is different than vehicular traffic research, which suggests that young and
inexperienced drivers (typically teenagers) tend to underestimate the potential risk of excessive
speeding (Deery, 1999; Machin & Sankey, 2008). Studies of snowmobile fatalities in the United
States and Scandinavia concluded that young males in their thirties were the predominate victims
of snowmobile accidents (Öström & Eriksson, 2002; Rowe et al., 1992; Stewart & Black, 2004).
This difference may be explained by the different age distribution between licensed drivers and
Michigan snowmobilers. While licensed drivers were relatively evenly distributed across age 20
to 65 (National Safety Council, 2014), 70% of the Michigan snowmobilers were between the age
of 35 to 60, with mean age of mid-40s (Nelson et al., 2009). It is also likely that individuals
operated a snowmobile as a sensation-seeking recreational activity, rather than one with a clear
transportation component that characterizes vehicular travel on roads. Therefore, recreationists
may be more likely to neglect rules for snowmobile operation and underestimate potential risks
from thrill seeking behaviors such as excessive speed and operation on roads where conditions
often favor reaching high speeds.
106

Those categorized as novice snowmobilers were found to perceive significantly less
danger from “Cars/trucks on seasonal roads” compared to intermediate and expert snowmobilers.
Michigan snowmobile fatality accident data shows that 10% of the snowmobile fatalities from
2008-13 resulted from snowmobilers colliding with cars or trucks (Michigan DNR, 2008; 2009;
2010; 2011; 2012; 2013). The finding is different from Lynch’s (2000) study, which found that
those who snowmobiled more days and miles perceived cars/trucks on seasonal roads as less
dangerous than those who did not violate snowmobile laws. That novice snowmobilers in the
current study perceived significantly lower risk from cars or trucks on seasonal roads may due to
a lack of experience and awareness. The results from this research indicate that once a novice’s
riding days and miles increase, they become marginally more aware of the threats posed by
vehicles on the seasonal roads often used for snowmobiling.
Although less apparent, the influence of recreational specialization on other perceptions
of riding risk varied in direction. Recreational specialization seemed to positively affect
perception of risks when intermediate and expert snowmobilers perceived higher risks on
“Operation of snowmobile by person who has been drinking but is not legally intoxicated” and
“Driver lacking skill in operating machine,” when compared to novice snowmobilers.
Recreational specialization appeared to negatively affect perception of risks when novice and
intermediate perceived higher risk on “Public trail condition” and “Public trails design”
compared to expert snowmobilers. Although this finding is consistent with the previous
Michigan snowmobile study that found snowmobiler subgroups differed in their perception of
risks (Lynch, 2000), intensified enforcement efforts on both the designated trail system and
Michigan road shoulders initiated in the late 90s could have changed Michigan snowmobiling
dynamics. As shown in two recent Michigan snowmobiling reports (Nelson et al., 1998; Nelson
107

et al., 2009), the reported percentage of snowmobilers seeing law enforcement personnel while
on their snowmobile or being contacted by a law enforcement official while snowmobiling has
increased 10%. This may be linked to the decline in snowmobile fatalities, which have fallen
from an average of 36 annually in the early 1990s to 20 fatalities annually in the past 10 years.
The presence of law enforcement may also have a positive influence on reducing risky
snowmobiling behavior, even as machines have gotten faster and the trail system has expanded.
While increases in snowmobile speed, handling and comfort have made it easier to ride longer
and faster, the increased visible presence of officers may be providing an effective deterrent to
illegal or risky behaviors. This still suggests that further improvements in safety and perception
of risk can be made and it is likely that snowmobile program managers and policy makers will
need to specifically develop intervention programs, regulations or even greater enforcement
presence to better prevent snowmobile accidents.
It is worthy of note that novice, intermediate, and expert snowmobilers saw evenly
minimal risk on “snowmobiling on county/state road shoulders,” while 43% of snowmobiling
fatalities occurred on county/state or seasonal roads (Michigan DNR, 2008; 2009; 2010; 2011;
2012; 2013) and during the early 1990s, 48% occurred in those same situations (Nelson et al.
1998). With this continued clear risk of riding on county/state road shoulders, it is important to
develop strategies to reduce fatalities in these locations as described in the administrative
recommendation section. In summary, based on the lack of distinction along the continuum of
recreational specialization in relationship to perception of risk, massages, programs, regulations
and enforcement would appear to be effective if targeted to snowmobilers as a while without
tailoring efforts to certain segments of the population. This is borne out by the significant

108

reduction in snowmobiling fatalities since more aggressive and visible enforcement efforts were
initiated across the snowmobiling population following the Nelson et al. (1998) study.
Relationships between Recreational Specialization and Spatial Interaction
The results of the effect size Eta showed a weak relationship between Michigan
snowmobilers’ recreational specialization and their spatial interaction with snowmobiler trail
network. However, a spatial interaction difference was found among novice, intermediate, and
expert snowmobilers. For Michigan snowmobilers, the destination trail networks in regions one
(western UP), two (eastern UP), and four (northeastern LP) appeared to attract different levels of
specialization among snowmobilers. Expert snowmobilers from region four and five (southern
LP) had a significantly stronger spatial interaction with region one compared to less specialized
snowmobilers. Similar attractiveness was found for region two as expert snowmobilers from
region three (northwestern LP) and region five particularly went to the region for snowmobiling.
Interestingly, region four specifically attracted intermediate snowmobilers from region three and
four; they had significantly stronger spatial interaction with the snowmobile trail network in
region four compared to expert and novice snowmobilers.
The above results are closely in line with the studies of recreationalists’ site preferences
as different levels of specialization differ in setting preferences and site choices. Bryan (1977)
hypothesized that highly specialized individuals seek settings to test their skill and exert control.
While this aspect would need future investigation with snowmobilers, research concerning
skiers and snowboarders’ consumption behaviors confirmed that those with higher
skill/knowledge preferred a variety of new and challenging downhill opportunities and that
snow condition became more important as the frequency of skiing and snowboarding increased
(Won et al., 2008). McFarlane’s (2004) vehicle-based camper study also concluded that
109

specialized campers seek settings that require a higher degree of self-reliance. It is likely that
snowmobile trails in the Upper Peninsula have special meaning to expert snowmobilers. The
scenery along the trail, quantity and quality of available trail loops, trail grooming quality,
remoteness, and lodges and restaurants available could all be contributing factors. More
specifically, the predictive mechanism of spatial interaction could benefit from a close
examination of place attachment and recreational specialization, as Oh and his colleagues (2012)
suggest in their study of anglers’ skill levels and their emotional attachment to a recreational
place. Those with high commitment tend to develop functional meanings and emotional
attachment to their favorite sites.
This study showed the spatial availability and accessibility of the snowmobile trail
network in regions one, two, and four are likely to explain the variation between snowmobilers.
There were 1,819 miles of snowmobile trail in region one with 136.9 inches of snowfall in winter
2007-08, which produced the best spatial availability among Michigan regions. The significant
snowmobiling opportunity attracted expert snowmobilers from region four and five who were
willing to travel a long distance for the winter activity. Region four snowmobilers on average
traveled 312 miles while region five snowmobilers on average traveled 445 miles in order to
access to region one for snowmobiling. This result confirmed Bryan’s (1977) assumption that
more specialized recreationalists are willing to travel farther in order to access what they
consider exceptional opportunity, compared to their less specialized counterparts. Similarly, the
snowmobiling opportunity in region two had 1,371 miles of snowmobile trail with 138.4 inches
of snowfall in winter 2007-08, providing the second best environment for snowmobilers,
particularly expert snowmobilers from regions three and five. The mean travel distance from
regions three and five to region two were 123 miles and 252 miles respectively. Region four, on
110

the other hand, was the favorite destination for intermediate snowmobilers from region three and
four. The 1,194 miles of trail and 83.7 inches of snowfall in the region attracted snowmobilers on
average traveling only 73.6 miles and 36.8 miles, respectively, for snowmobiling. As the
literature states, accessibility to recreational opportunity can be one of the major constraints for
regional skiers and snowboarders. Less specialized skiers and snowboarders tend to be more
concerned about cost barriers than those who are more advanced (Won et al., 2008). Novice
snowmobilers presented similar characteristics in this research as they were likely to travel
shorter distances to snowmobile than expert snowmobilers.
The study found no evidence of spatial interaction between Upper and Lower Peninsula
for those who lived in Upper Peninsula. This suggests that those snowmobilers who resided
proximate to relatively better snowmobile opportunity in the Upper Peninsula did not travel
south across the Mackinac Bridge to access comparatively less miles of snowmobile trail and
less snowfall in Lower Peninsula of Michigan. It is unlikely that they would access other
adjacent states such as Minnesota and Wisconsin, where while there is a substantial trail
network, snowfall/snow cover is less dependable.
The gravity model applied in this research provided a useful approach to measuring
spatial interaction between snowmobilers and recreation opportunity using the destination
attributes of winter snowfall, snowmobile trail mileage, and Euclidean travel distance as a
friction function. The percentage of visitation in each managerial region was also incorporated to
account for the attractiveness of the destination. Baxter (1979) argued that the attractiveness of
trip destinations should be specified exogenously instead of being constant. However, the
debates on appropriate calibration, algebraic forms between parameters, and friction exponential
as raised by Sen and Smith (1995) are not likely to end with this current study. Nevertheless, the
111

model used in this study does provide meaningful measures and a partial explanation of spatial
interaction, furthering understanding of the relationship between the spatial structures of
snowmobile trails and users’ specialization. The findings in this snowmobiling study confirmed
findings regarding Oklahoma state parks using the gravity model. Fesenmaier and Lieber (1985)
found that the spatial distribution of state parks and their recreation facilities had a substantial
effect on visitation at one park over another. Future research is certainly warranted to further
improve functioning and accuracy of the gravity model.

Managerial Implications and Recommendations
General Managerial Implications
Three key managerial implications were revealed by this research. First, the results from
both summative index and cluster analysis demonstrated that Michigan snowmobilers are a
heterogeneous group with unique behavior, skill/knowledge, and psychological attachment
characteristics. Some of them progressed through a linear trajectory such that they became more
specialized when they spent more time and money on the activity. Others focused on different
aspects of snowmobiling such as trail riding or conspicuous consumption, once they acquired
enough snowmobiling experience, without advancing on a linear path of specialization.
Second, overall perceptions of risky snowmobile riding behaviors were not well predicted
by level of recreational specialization. Universally and most critically, snowmobilers of all levels
of specialization perceived relatively little risk form the riskiest place to ride a snowmobile, a
state or county road shoulder. However, the dangers of operating a snowmobile while intoxicated
were known across the continuum of specialization. Further research could perhaps most
profitably focus on the influence of the increased visible presence of officers as actual fatalities
have fallen in Michigan since such additional enforcement efforts have been put in place.
112

Finally, different regions of the Michigan snowmobile trail system appeared to attract
different levels of specialization in Michigan snowmobilers. Expert snowmobilers, who resided
in the Lower Peninsula, preferred snowmobile trails in the Upper Peninsula, with its extensive
trail mileage and snowfall. Intermediate snowmobilers favored snowmobile trails in northern
Lower Peninsula, which provided somewhat less snowmobiling opportunity due to less snowfall,
but facilitated snowmobiling with less commitment as drive time/distance to trails was reduced.
Additional snowmobile trail development in the northern Lower Peninsula appears more likely to
provide a larger economic return in regards to Michigan residents considering the proximity to
Michigan’s population centers.
Administrative Recommendations
Several administrative recommendations regarding Michigan snowmobilers and their
perception of risks of and spatial interaction with the snowmobile trail network are presented in
this section. Program directors and managers may be more informed in terms of diverse
characteristics displayed by Michigan snowmobilers and how this profile can aid in the
execution of marketing strategies and management actions.
A mandatory snowmobile training course for all first time Michigan snowmobile trail permit
purchasers for those born on or after January 1st of 1998
The current snowmobile education program targets snowmobilers 16 years or younger,
however, with the aging baby boomers and the incoming tourist searching for experience,
individuals may start snowmobiling at a later stage of the life. A mandatory education or training
only for those under 16 neglects those novice snowmobilers. Therefore, for those first time
Michigan snowmobile trail permit purchasers born on or after January 1st of 1998 (17 years or

113

younger), should take and pass a mandatory snowmobile training course. If this birth year is held
constant, annually this will increasingly involve new snowmobilers of more ages, just as has
been done with hunter safety education in Michigan and many other states (Michigan DNR,
2014). Because this has happened gradually with hunter safety, there has been minimal
complaint and hunting fatalities related to firearm and archery safety have declined by more than
half in Michigan since this hunter safety education requirement was implemented coupled with
the requirement to wear hunter orange when hunting in most situations.
Education should be focused on riding behaviors that are critical in safeguarding
snowmobilers as well as snowmobile rules and laws. The training program should help not only
raise awareness of several key riding risks underestimated by snowmobilers (i.e., drivers lacking
skill in operating machine; snowmobile on county/state road shoulders), but also familiarize all
snowmobilers with important operating skills and threats posted by other snowmobilers or other
users of the snowmobile trail system. As a result, the incidence and severity of snowmobile
accidents can be reduced by properly designed training and education.
Disseminate snowmobile regulations, riding tips, the consequence of violating regulations, and
fatality details to identified target segments of snowmobilers in key snowmobile information
nodes such as snowmobile rental outlets, Secretary of State Offices, and snowmobile dealers, as
well as on the DNR website
While Michigan snowmobilers are not a homogeneous group, the minimal variation of
perceived risk on nine hazardous snowmobile riding behaviors disclosed by this research
demonstrates that there is little need to craft detailed strategies for less than the full population of
snowmobilers regarding appropriate assessment of risk. Relevant information such as booklets,

114

brochures, accident details, and safe riding guides should be widely distributed particularly to
snowmobile dealers, snowmobile rentals, travel agencies, Secretary of State Offices, and tourism
service providers in snowmobiling areas linked to the designated trail network. Internet
information and social media such as the DNR’s official website, Facebook and Twitter accounts
are other vehicles to disseminate this crucial information. In addition, regarding the danger of
speeding, programs and campaigns hosted by snowmobile clubs and organizations, such as the
Michigan Snowmobile Association, may be especially effective. Messages that specifically
target excessive speeding could utilize “peer pressure” to help reduce accidents and fatalities
because of the speed of snowmobiles.
Develop additional rail-trails without car/truck traffic that directly connect community to
community
Generally, Michigan snowmobilers perceived low risks in snowmobiling on county/state
road shoulders, while 43% of the fatalities occurred on county/state or seasonal roads (Michigan
DNR, 2008; 2009; 2010; 2011; 2012; 2013). The fact that Lynch (2000) found very similar
results in regards to snowmobilers’ low perception of risk from riding on county/state road
shoulders suggests the risks of riding on county/state road shoulders should be further
emphasized.
Another physical approach that holds promise is the development of additional rail-trails
without car/truck traffic that directly connect community to community and are part of the
designated snowmobile trial network. Typically, businesses where snowmobilers can access
goods (e.g., gasoline, food, etc.) and services (lodging, vehicle repair, etc.) are located on or very
near these corridors. The development of this initial transportation link in the 19th and early 20th

115

century often defined where businesses were initially located. Today, they increasingly provide
for tourist traffic flow in both winter and non-snow months by a wide variety of non-motorized
trail users such as bicyclists who use these corridors in the non-snow months. With these trails
having well groomed surfaces for snowmobiling, the risk of riding on county/state road
shoulders and cars/trucks on the road can be avoided or at least reduced through these alternate
connections to goods, services and communities as well as providing a scenic riding venue.
Construct voluntary safety patrol programs though local snowmobile clubs and organizations to
reduce aberrant riding behaviors
Despite intensified patrol and visibility, Michigan DNR conservation officers and
deputies from County sheriffs and city police departments cannot fully cover the extensive
snowmobile trail and county/state highway right-of-ways across Michigan. A National Mountain
Bike Patrol Program initiated by International Mountain Biking Association (IMBA) focusing on
safe and responsible riding provides a potential model for monitoring snowmobiling activities in
the state of Michigan. The voluntary safety patrol program can recruit volunteers who are trained
with skills and exhibit safe snowmobiling behavior as well as being proficient in first aid, CPR,
minor snowmobile repair, and knowledgeable about the area to provide information, directions,
and accurate reports of trail conditions or hazards. The volunteer patrollers should follow
guidelines and rules developed by Michigan DNR or Michigan Snowmobile Association (MSA),
wear a patrol vest, and carry necessary items to ensure quality and visibility. The volunteer
patrollers can also partner with a Michigan DNR conservation officer or local unit deputy that
can enforce riding rules and regulations. A cooperative model currently used in Michigan
involving non-motorized trail users involves the Kent County Sheriff’s Department and patrol by
trained volunteer cyclists on the Michigan DNR’s Fred Meijer White Pine Trail State Park and
116

Kent County’s Kent Trails. It uses 80 volunteers who patrol on scheduled basis in cooperation
with the Sheriff’s department, who assist trail users and act as eyes and ears for the sheriff’s
department (Kent County, Michigan, 2014).
Create a snowmobile accident digital profile with geographic coordinates
A geospatial profile with accurate geographic coordinates of those snowmobile accidents
resulting in fatalities or injuries that need medical attention will provide comprehensive data for
accident investigation. This will facilitate better understanding of possible environmental and
human factors involved with serious accidents as well as key spatial data (e.g., relationship to
designated snowmobile trails, designated snowmobile trailheads, roadways, etc.). By utilizing
accurate locations of snowmobile accidents coupled with accident types and factors, analysts will
be able to detect problematic areas on the snowmobile trail system and county/state road rightof-way where managerial actions can be applied accordingly. A similar effort in Michigan’s
Silver Lake State Park was done by using off-road vehicle (ORV) accident profiles with GPSed,
mapped locations, which successfully detected accident hot spots and revealed areas in need of
managerial actions such as speed zones and unidirectional traffic areas. In turn this had a positive
association with reducing ORV accidents, providing a safer sand dune riding experience (Wu,
Borland, & Nelson, 2011). Geospatial technology such as Global Positioning System (GPS) and
GIS are becoming increasingly affordable and user-friendly. A hand-held, recreational grade
GPS unit offers reasonable accuracy in mapping accident locations even in heavily wooded
areas, especially in the leaf-off winter season. Developing such a database requires easy-to-use
technology, effort in marking accurate GPS locations and protocols of recording accident details.
Appropriate training for those personnel responding to accidents, unified accident reporting
forms, and state-wide central database management are crucial beginning steps.
117

Planning a complete package of development including extension of snowmobile trails and
related amenities in northern Lower Peninsula
The high spatial interaction received by regions three and four indicated that the northern
Lower Peninsula is the most popular snowmobile destination, with substantial trail mileage,
adequate snowfall, abundant natural resources, and close proximity to most of Michigan’s
snowmobile population which resides in southeastern Michigan. Other research has shown that
linking special events and tourist attractions nearby with a supporting transportation network as a
fully developed tourism package created a regional approach with strong tourism
competitiveness that can obtain an increasing influx of visitation (Lee, et al., 2013; Kim &
Fesenmaier, 1990). It is important not only to extend the snowmobile trails, but also better link
the goods and services supporting the activity (i.e., lodging, gas stations, and restaurants) and the
infrastructure to deliver the services (i.e., roads, internet) to the trail network. Furthermore, a
horizontal marketing effort, cooperating with nearby local skiing and snowboarding resorts,
special winter events, and other regional tourism attractions, will increase the level of
attractiveness to winter recreation enthusiasts. The tourism support system is valuable to nonsnow tourism as well, providing the services, goods and infrastructure to other natural resource
recreation based tourism year round.
Develop marketing strategies targeting out-of- state snowmobilers
A similar tourism package to that of northern Lower Peninsula should be considered for
snowmobile trails in the Upper Peninsula. This research found that expert snowmobilers were
willing to travel a long distance (i.e., 445 miles from the Detroit area) in order to access an area
with longer trail mileage, more snowfall and outstanding natural resource assets. The economic

118

benefit brought on by these expert snowmobilers was significant. The natural resources in the
Upper Peninsula, with properly developed services and amenities, could attract more out-of-state
snowmobilers. When considering non-residents, snowmobile trail investments in both the Upper
Peninsula and northern Lower Peninsula have significant potential as both regions are close to
key non-resident markets with the western Upper Peninsula close to Wisconsin, Illinois,
Minnesota and Iowa and the eastern Upper Peninsula and the northern Lower Peninsula close to
Illinois, Indiana, Ohio and Pennsylvania. Studies have shown that snowmobilers from
Wisconsin, Illinois, Minnesota, and Iowa visited both western and eastern Upper Peninsula for
snowmobiling, while many snowmobilers from Illinois, Indiana, Ohio, and Pennsylvania
snowmobiled in Lower Peninsula in winter 2007-08 (Wu, Nelson, & Chang, 2012). Considering
the Canadian market from Toronto and surrounding environs, the Eastern Upper Peninsula and
the Northeastern Lower Peninsula have the greatest potential. This also provides an opportunity
for regional tourism entities such as the Upper Peninsula Tourism Association, West Michigan
Tourism Association, Sunrise Side Tourism Association, etc. to market both snow and non-snow
attractions to visitors who may already be familiar with Michigan in one or two seasons, but not
all four and to complement existing travel patterns.

Research Limitation
In this study, the key limitation is that the state-wide survey was originally designed to
study 2007-08 Michigan trail permit purchasers’ use of public natural resources and attitudes to
multiple managerial actions, instead of measuring recreational specialization and its influences.
Though the measurement of recreational specialization closely followed the multi-dimensional
structure suggested by the literature, a more comprehensive construct including self-rated skill
level and centrality-to-life may yield additional richness in profiling snowmobilers’ within-group
119

social world. Program managers and leisure scholars could benefit from an instrument that was
originally and specifically tailored to meet the objectives of this research.
Secondly, this study utilized Michigan snowmobilers who purchased a trail permit in the
snowmobile season of 2007-8 in an attempt to understand the within-group social world of
snowmobilers and how their respective characteristics influenced their perceived snowmobiling
risks and interaction with snowmobile opportunities. The snowmobiler sample in this study
represents a small segment of the total range of the worldwide snowmobiling population
available. However, it was a valid sample to describe and explain the behavior of Michigan
resident snowmobilers entitled to lawfully ride on the designated snowmobile trail system. Data
regarding snowmobile trail length and snow precipitation were also limited to a single year.
From this point of view, the construct and research procedures described in this research would
benefit from a broader snowmobiling population and a longer temporal selection of supporting
resources.
In addition, using the most snowmobile friendly point of each destination region to
measure travel distance between each snowmobiler’s residential origin and destination may have
distorted the true travel distance. Unlike many other outdoor recreational activities where
participants stay in a relatively small geographic area for the activity, snowmobilers usually
trailer their sleds to a starting point and travel on the snowmobile trail system for a long distance.
The trail systems are interconnected and often cover several counties and there are even cross
state links. Therefore, it is difficult to pinpoint one location to represent the area or region
snowmobilers cruised within. However, utilizing the centroid of the county with the greatest
product of average snowfall and trail mileage to represent each snowmobiler’s destination within
the region did provide a scientifically defendable and pragmatic approach to estimate travel
120

distance. While the measured distance is an objective estimation instead of the precise travel
distance for each visit, it is a useful approximation. Future research may benefit from more
precise origin destination measurement using GPS devices with respondents or some similar
technology.
Lastly, while the calibrations and analysis procedures were based on peer-reviewed
journal papers and considered reasonable in addressing research questions, the literature also
suggested other measurements and processes. These included different measures of travel cost
and destination attributes or a more complex gravity model and statistic method. This is fertile
ground for future research inquiry regarding the relationships between snowmobilers’ withingroup social world and the perception of risk and spatial interaction.

Future Research
This study should be viewed as a starting point for examining the recreational
specialization of snowmobilers, their perception of risks, and the spatial interaction between the
availability and accessibility of snowmobiling opportunities and demand. While certain
relationships among three key concepts were revealed by this research, future research should
focus on the following.

First, future research should focus on a broader range of snowmobilers to further
investigate the within-group social world of the participants and the activity. A study with the
sample of snowmobilers from other jurisdictions in the Upper Midwest such as states of
Wisconsin and Minnesota, as well as Canada, and Scandinavia countries will provide a more
complete coverage of the snowmobile population. This diverse research sample will enhance the
scholarly understanding of snowmobile within-group social world and perhaps reveal
121

distinctions among jurisdictions/origins. Further, conducting a longitudinal study with
snowmobilers to investigate the role of snowmobiling in their range of leisure pursuits and at
various stages of their life will greatly improve not only the knowledge of snowmobiler social
world, but also the development and progress of recreational specialization in this activity. In
addition, little work has been done in relationship to other motorized adventure recreation (e.g.,
ORV riding), which would benefit from a study fully designed and tailored to understanding
recreation specialization. This would be complicated by the distinctly different mature of
different types of off-road vehicles (e.g., motorcycles, ATVs, larger specialty vehicles such as
dune buggies, etc.).

Such studies could include a more focused survey instrument with additional key
skill/knowledge and centrality-to-life measurements. Other statistical analysis methods such as
structural equation modeling and logistic regression may also be applied to investigate further
the influence of recreational specialization on perception of risks and spatial interaction. Which
behavioral, skill/knowledge, and psychological attachment dimension has more explanatory
power to those two concepts can be examined in detail. In the attempt to provide a more holistic
understanding of snowmobile social world, qualitative methods provide another alternative that
can be used to obtain insights regarding snowmobilers’ leisure pursuit. Research questions such
as “what does snowmobiling mean to snowmobilers?”, “how do snowmobilers progress through
recreational specialization continuum?”, or “how does snowmobiling experience in the early
childhood affect snowmobilers’ leisure participation in their later life?” have yet to be
investigated. An in-depth interview or a participatory action research may offer answers to the
questions stated above.

122

Second, to reduce the incidence and severity of snowmobiler accidents requires an
examination of cognitive, affective, and behavioral dimensions of snowmobiling besides
perceptions of driving risks. More detailed general demography, likelihood of risk-taking in
general, one’s propensity for sensation seeking, and circumstantial decision making are just a
few constructs utilized in accident and traffic research in studying risky driving behaviors.
Different taxonomic approaches in segmenting snowmobilers such as the subgroups produced by
cluster analysis may offer different perspectives of assessing the influence of recreational
specialization on perception of risk. Other action research such as using a riding stimulator, GPS
tracking mechanisms and helmet cameras may provide meaningful behavioral results outside of
the behavioral intentions or reported behaviors discussed in a survey.

Third, while spatial availability and accessibility as destination attributes and friction
functions in the traditional gravity model presented a general indication of spatial interaction
between snowmobile supply and demand, other calibrations such as destination attractiveness,
visitor preferences, and different means of friction parameters and exponentials will deepen
scholarly understanding of spatial interaction between snowmobilers and their destination trail
network. Friction parameters such as travel time, travel cost, cognitive distance, road network
distance, or population weighted distance, will provide other selections in a gravity model. Using
more detailed travel and destination information such as a specific destination down to the trail
head or specific city/town will also increase the accuracy of the model. Moreover, the association
between recreational specialization and spatial interaction can be improved by a more complex
gravity model such as the two-step floating catchment area (2SFCA) method that looks at
differential travel distances or time thresholds, as has been used when measuring health care
accessibility (Luo & Wang, 2003). Snowmobilers who resided within a distance threshold
123

receive one friction parameter while those lived outside of the threshold receive another, such
that the severe skewness of measured spatial interaction distribution can be partially accounted
for.

Lastly, using observed visitation, in this case snowmobile days, coupled with origin and
destination attributes, may provide the opportunity to model spatial interaction between origin
and destination. This is the ultimate target in investigating spatial interaction. Once the model is
proven accurate and valid, inertia effects can be applied to explain how increasing the length of
snowmobile trails in one area or decreasing friction in terms of distance would affect general
snowmobile travel flow between demand and supply. The constructed model will provide
valuable evidence and reference for future resource management and program planning.

Conclusion
The study improves the understanding of the within-group social world of snowmobilers.
Further, as no published literature has studied this social world and directly examined the effects
of recreational specialization on recreationalists’ perceived risks and spatial interaction with
recreation resources, this exploratory research clearly adds to the literature. It does this by
explaining how personal characteristics are associated with an individual’s cognitive
understanding of risky riding behaviors, as well as their preference for travel destinations. The
within-group social world of snowmobilers presents program managers with insights useful in
developing marketing plans, policies, and managerial strategies to protect health and safety. In a
dynamic tourism environment with the aging baby boomers and incoming tourists brought in by
the “Pure Michigan” marketing campaign, program managers, policy makers, and law
enforcement must embrace the on-going change in snowmobiles, trails and those who ride. They

124

need to utilize newly developed technology to more efficiently and effectively manage natural
resources and recreation programs. Tailored managerial actions for targeted populations and
informational messages in trusted modes and venues are important strategies for sustaining
recreation programs, protecting rider safety and responding to change.

Partnering with local enforcement units and recruiting enthusiastic volunteers to become
helping hands and eyes and ears on the snowmobile trails are additional alternatives in
safeguarding riding enjoyment. Geographic information systems with new spatial analytical tools
provide powerful, commonplace, and easier-to-use analysis to study recreation behaviors from a
spatial aspect. Social media offers an effective marketing reach at relatively minimal cost to
those seeking winter excitement living outside of Michigan.

Finally, thinking in the longer term with predicted climate change, snowmobiling will be
on the forefront of adaptation to a warmer climate. A Canadian climate change impact
assessment predicted that snowmobile season will reduce by 24% in the 2020s and 33% in the
2050s at Sudbury, Ontario, which is located about 160 miles east of Sault Ste Marie, Michigan at
a similar latitude with the Upper Peninsula. Orillia, Ontario, 260 miles east of Grayling,
Michigan (located at a similar latitude with the northern Lower Peninsula) was projected to lose
44% of the snowmobile season in the 2020s and 56% in the 2050s (McBoyle, Scott, & Jones,
2007). Therefore, trail venues in the Upper Peninsula, linking heavier snowfall with colder
temperatures than the Lower Peninsula and the rest of the Great Lakes region and still benefiting
from lake effect precipitation and cloud cover, could make Michigan’s Upper Peninsula a great
national and not just a regional snowmobiling destination in the years to come. Snowmobiling
and the trail system are important assets for the State of Michigan. Continuing research efforts

125

and the best practices of managing social and natural resources are keys to a sustainable, welladministered snowmobile program of the future.

126

APPENDIX

127

Questionnaire
1. Please complete the following table to describe the people in your household and their involvement in
snowmobiling. If there are no people in a certain category, please write 0 for that category.

# in household

# who snowmobiled
Winter 2007-08

# who snowmobiled
Winter 2008-09

# who have completed
a snowmobile safety class

Adults 18 or older
Children 12-17
Children 11younger
If no one from your household snowmobiled in Michigan during Winter 2007-08 or Winter 2008-09, please
skip to Question 7.
2. For each snowmobile your household owns, please complete the table below about last year’s and this year’s
snowmobiling. Please circle the correct y=yes or n=no regarding use, trail permits and registration and be sure to fill
in the number of miles each machine was driven in MI.

Model
year

Used in MI Used in MI
winter
winter
CC 2007-08? 2008-09?
y
n
y
n
y
n
y
n
y
n
y
n
y
n
y
n
y
n
y
n

Had 2008
MI Trail
Permit?
y
n
y
n
y
n
y
n
y
n

Has 2009 Current
Approx # MI Approx. # MI
MI Trail MI
miles driven miles driven so far
Permit? Registration winter 07-08 winter 08-09
y n
y n
y
n
y
n
y
n
y
n
y
n
y
n
y
n
y
n

3. Please estimate, using the table below, how many Michigan SNOWMOBILE DAYS you and household
members operated your snowmobile(s) in Michigan last winter and so far this winter. A SNOWMOBILE DAY IS
EACH DAY OR PART OF A DAY ONE SNOWMOBILE WAS OPERATED in Michigan unless it was primarily
used to support other activities like ice fishing. For a region or winter in which your household did not snowmobile,
leave that spot blank. Please add the total days for each year on the last line, even if it is 0.
Region (see map)
MI Snowmobile
MI Snowmobile
Days Last Winter
Days This Winter
(2007-08)
(2008-09)
1 Western UP
2 Eastern UP
3 NW NLP
4 NE NLP
5 Southern LP
Total Snowmobile Days
3a. Did you use any of your snowmobiles to support ice fishing, trapping or some other recreational activity in
winter 2007-08? ___yes ___no … or winter 2008-09? ___yes ___no
4. How many of these TOTAL SNOWMOBILE DAYS from Q3 were mostly spent on the Michigan designated
snowmobile trail system (trails marked with orange diamonds)?
Last winter (2007-08) ____# days
This winter (2008-09) _____# days
5. How many gallons of gasoline did your household use in all of its snowmobiles in Michigan during winter 200708 _______ # gallons and so far this winter (2008-09) _____________# gallons.

128

6. Of the total MI SNOWMOBILE DAYS last winter and this winter (last row in Q 3 above), how many were from
which type of trip?
Type of trip
Days involving overnight stays away from
permanent home
Days not involving overnight stays away from
permanent home but with travel of 100 or
more miles one way to snowmobiling site/area
Days not involving overnight stays and less
than 100 miles from home
ALL SNOWMOBILING DAYS (should
equal “Total Days” from last row of Q3)

Last Winter (2007-08)
# days

So far this winter (2008-09)
# days

# days

# days

# days

# days

# days

# days

Now, for your most recent Michigan snowmobiling outing when you stayed overnight away from home or
snowmobiled more than 100 miles from your permanent home, please answer questions 7-13. This outing
likely occurred during winter 2008-09.
7. When did the MOST RECENT outing begin? Month ____
8. In which region of the state was it
primarily focused? (Please check one)

Date ____ Year ____

__ Western UP __ Eastern UP
__Northwestern LP
__Northeastern LP
__Southern LP

9. How many people from your household ______# and how many of your snowmobiles _____#
were involved during that MOST RECENT outing?
10. How many total snowmobile days were your machines used for snowmobiling
(not as support for fishing) during that MOST RECENT outing?
__________# days
11. Did the outing involve an overnight stay away from your permanent home? __Yes __No (go to Q 12)
11a. If you stayed overnight away from home, where did you stay? Please check
all that apply.
____ Own second home
____ Motel/hotel/rental cottage or cabin
____ Camped
____ At friend’s/relative’s
____ Other (please explain_________________________________________)
11b. In total, how many nights did you spend away from home on this trip? _____# nights
12. How much money did your household spend on the entire trip during that MOST RECENT outing? Please
complete the table below for spending at home in preparation, traveling to and from the snowmobiling area, and in
the area where you rode. If you spent nothing on an item, please leave it blank.

Grocery & convenience store food/drink
Tow vehicle expenses (gasoline, repairs, etc.)
Snowmobile expenses (gasoline, repairs, etc)
Restaurant and bar meals and drinks
Sporting goods (bait, fishing tackle, etc.)
Lodging (motel, rental cabin, etc.)
All other items (clothing, souvenirs, etc.)

At Home
$
.00
$
.00
$
.00
NA
$
.00
NA
$
.00

129

En Route
$
.00
$
.00
$
.00
$
.00
$
.00
$
.00
$
.00

Local Area
$
.00
$
.00
$
.00
$
.00
$
.00
$
.00
$
.00

13. Whom did these expenditures cover? ____# from your household + ____ # others?
14. What were your household’s MI snowmobiling related expenses in the last 12 months in the following
categories? If there were no expenses for a category in MI, leave category blank.
Expense Item
Purchase of snowmobile
Purchase of snowmobile equipment other
than snowmobile (trailer, clothing, etc.)
Snowmobile repair/maintenance
Insurance on your snowmobile(s)
Off-season storage costs

Dollars spent in last 12 months
$
$
$
$
$

Now, I’d like to ask some questions about Michigan snowmobiling and its management.
15. Please rate the following services/situations regarding the Michigan DNR’s snowmobile program. Use a scale of
1 to 5, with 5 as very good, 4 as good, etc. Please CIRCLE the appropriate number for each item.
Very
Very
Services/Situations____
Good___Good__OK___Poor___Poor__Why your Rating?_______
Public trail grooming
5
4
3
2
1 __________________________
DNR snowmobile trail maps
5
4
3
2
1 ___________________________
Snowmobile trails
5
4
3
2
1 __________________________
Snowmobile trail signage
5
4
3
2
1 __________________________
Public trail design
5
4
3
2
1 __________________________
Trailhead parking/staging
5
4
3
2
1 __________________________
Snowmobile law enforcement
5
4
3
2
1 __________________________
Snowmobile safety education
5
4
3
2
1 __________________________
DNR snowmobile website
5
4
3
2
1 __________________________

16. Overall, how satisfied are you with Michigan snowmobiling on a scale of 1 (highly dissatisfied) to 5 (highly
satisfied)? ____# rating
What is the ONE most important reason for your rating? ________________________________
17. Is the public snowmobile trail system in Michigan overcrowded? ____ Yes

____ No

18. Please list the ONE improvement you would most like to see in the Michigan snowmobile trail system.
_____________________________________________________________________________________
19. Would you be willing to pay more to snowmobile in Michigan to pay for this suggested improvement?
____ Yes

____ No

____ Undecided

19a. If yes, please check the highest additional amount you would annually
be willing to pay per snowmobile for this suggested improvement?
__$5 __$10 __$15

130

__$20

20. It has been proposed that the State of Michigan acquire permanent easements or invest in longer term leases for
the portions of the designated snowmobile trail system that are on private lands if the current owners are willing. It
has also been proposed that such an approach should be used to expand the current snowmobile trail system. Please
circle your level of support for each of the following options on a scale of 1-5, with 5 as strongly support, 4 as
moderately support, down to 1 strongly oppose.
Action
More long term trail leases held by non-profit
snowmobile organizations/clubs
More long term trail leases held by State of
MI
Permanent trail easements held by non-profit
snowmobile organizations/clubs
Permanent trail easements held by State of MI
Expand the mileage of the designated MI
snowmobile trail system

Strongly
Support
5

Moderately
Support
4

Neutral
3

Moderately
Oppose
2

Strongly
Oppose
1

5

4

3

2

1

5

4

3

2

1

5
5

4
4

3
3

2
2

1
1

21. Would you be willing to pay more to snowmobile in Michigan to pay for longer term leases or permanent
easements for more of the snowmobile trail system?
____ Yes ____ No ____ Undecided
21a. If yes, please check the highest additional amount you would annually
be willing to pay per snowmobile for long term leases/permanent easements? __$5

__$10

__$15 __$20

22. Over winter 2007-08, 25 snowmobilers were killed during Michigan snowmobiling. The DNR is considering
potential ways to reduce fatalities. Please circle your level of support for each of these on a scale of 1 to 5, with 5
strongly support, 4 moderately support, etc.
Fatality Reduction Initiative
Strongly Moderately Neutral Moderately Strongly
support
support
oppose
oppose
Enforced snowmobile speed limit on all
5
4
3
2
1
public lands, designated trails and waters
Enforced snowmobile speed limit on all
5
4
3
2
1
designated trails only
Enforced snowmobile speed limit where
5
4
3
2
1
posted on designated trails only
Mandatory safety training for all first year
5
4
3
2
1
snowmobile operators regardless of age
Licensing required of all snowmobile
5
4
3
2
1
operators
Mandatory personal liability insurance for all
5
4
3
2
1
snowmobile operators
Snowmobile curfew/closed hours on
5
4
3
2
1
designated trail system
More intensive enforcement of existing
5
4
3
2
1
snowmobile regulations
23. Please list any other ideas you have about ways to reduce fatalities?_____________________________
______________________________________________________________________________________
24. Would you be willing to pay more to snowmobile in
Michigan to have fatality reduction initiatives?

____ Yes ____ No

131

____ Undecided

24a. If yes, please check the highest additional amount per snowmobile
you would annually be willing to pay to have fatality reduction initiatives? __$5

__$10

__$15

__$20

25. In total, considering all potential improvements to the Michigan snowmobile program discussed above including
the one improvement you would most like to see, efforts to make a more permanent and perhaps larger designated
snowmobile trail system and fatality reduction initiatives, how much more in total would you be annually willing to
pay per snowmobile?
__ $0 __$5
__$10 __$15 __$20 __$25 __$30
26. How many times during winter 2007-08 in Michigan were you or members of your household checked or
stopped by a law enforcement officer while snowmobiling? ____#
Winter of 2008-09?_____#
If you or another household member were stopped or checked by an enforcement officer one or more times,
please check which type of officer(s) was involved.
___ DNR Conservation Officer
___County/City/Other local unit officer/deputy
27. Did you or members of your household see, but were not stopped or checked by a law enforcement officer while
snowmobiling in Michigan during winter 2007-08?____ Yes ____ No
winter 2008-09?____Yes ____No
28. Please rate the level of danger to snowmobilers from the following, based on your snowmobiling experiences.
Use a rating scale of 1 to 5, with 5 being extremely dangerous, 4 highly dangerous, etc. and circle the appropriate
number for each.
Behavior/Situation
Operation of snowmobile by
person who has been drinking
but is not legally intoxicated
(0.01-0.07 blood alcohol)
Operation of snowmobile by a
legally intoxicated person
(0.08 or higher blood alcohol )
Speed of snowmobile
Driver lacking skill in
operating machine
Cars/trucks on seasonal roads
Public trail conditions
Public trail design
Other uses of designated
snowmobile trails (e.g. dog
sledding, cross country skiing)
Snowmobiling on county/state
road shoulders

Extremely
Dangerous

Highly
Dangerous

Moderately
Dangerous

Slightly
Dangerous

Not
Dangerous

5

4

3

2

1

5

4

3

2

1

5
5

4
4

3
3

2
2

1
1

5
5
5

4
4
4

3
3
3

2
2
2

1
1
1

5

4

3

2

1

5

4

3

2

1

29. List other behaviors/situations that you feel are extremely/ highly dangerous. _____________________
______________________________________________________________________________________

132

I would like to finish with some general questions about you.
30. Please check all of the recreation activities in which you participated during the past 12 months.
__Snowmobiling
__Camping
__Hiking
__Cross country skiing

__ORV riding
__Canoeing/kayaking
__Wildlife viewing
__Paved trail/road biking

31. What is your 5 digit home zip code? _________

__Open water fishing
__Power boating
__Ice fishing
__Pick wild mush./berries

__Hunting
__Horseback riding
__Mountain biking
__Trapping

32. What is your age? _____years

33. At what age did you begin snowmobiling? ________ years
34. Are you a member of a snowmobiling association or club? __No __Yes (please list each organization
____________________________________________________________________________________)

Please write any other comments you have about Michigan snowmobiling here.
Please mail the completed questionnaire back to me in the postage paid envelope provided. Thanks for your
assistance in helping to improve Michigan snowmobiling.
Dr. Chuck Nelson
131 Natural Resources Building
Michigan State University
East Lansing, MI 48824
(517) 432-0272

133

Initial Survey Cover Letter

January 15, 2009
Dear Michigan Snowmobile Trail Permit Purchaser:
The Michigan Department of Natural Resources (DNR) and Michigan State University (MSU)
are cooperating to better understand Michigan snowmobile use and users. This research study is
a follow-up to one done a decade ago by MSU for the DNR. Using that baseline information, it
provides the opportunity to track trends in snowmobiling since the late 1990s. You have been
randomly selected as one of the 3,000 trail permit purchasers to be sampled from the winter
2007-08 list of almost 225,000 snowmobile trail permit purchasers.
The enclosed questionnaire asks about your household’s snowmobiling during last winter (200708) and this winter. It also asks about your most recent snowmobiling outing (most likely this
winter). In addition, it provides the opportunity for you to evaluate Michigan’s snowmobile
program, suggest program improvements and enhance snowmobiling safety. Please take the 1520 minutes needed to complete the questionnaire and mail it back to me in the postage paid
envelope.
Your participation is voluntary. You can withdraw or refuse to answer any question without penalty.
There are no known risks associated with your participation in this study. Rather, your participation
in this study will contribute to a better understanding of Michigan snowmobiling and provide
guidance for future program direction.
Your responses will be kept confidential and your name will not be associated with any results.
Your privacy will be protected to the maximum extent allowable by law. If you have any questions
about this project at any time, please call Dr. Chuck Nelson, Associate Professor at MSU by phone
(517) 432-0272 or by email at nelsonc@msu.edu.
Thanks for helping to better understand and improve Michigan snowmobiling.
Sincerely,

Dr. Chuck Nelson,
Associate Professor
Enc.

134

Second Mailing Survey Cover Letter

February 15, 2009
Dear Michigan Snowmobile Trail Permit Purchaser:
About a month ago I sent you and some other randomly selected snowmobilers a survey about
your experiences and opinions on Michigan snowmobiling. While I have received completed
surveys from many others, I have not received yours. If our correspondence has crossed in the
mail, please accept my apologies for contacting your again. If you have not yet responded, please
do so using the enclosed survey. Your opinions are important. My goal is to be able to accurately
represent the collective views of those who snowmobile in Michigan in my report to the DNR.
Let your voice be heard. Please take the 15-20 minutes needed to complete the survey and then
mail it to me in the postage paid envelope. I will not be contacting you again about this matter.
Your participation is voluntary. You can withdraw or refuse to answer any question without penalty.
There are no known risks associated with your participation in this study. Rather, your participation
in this study will contribute to a better understanding of Michigan snowmobiling and provide
guidance for future program direction.
Your responses will be kept confidential and your name will not be associated with any results.
Your privacy will be protected to the maximum extent allowable by law. If you have any questions
about this project at any time, please contact Dr. Chuck Nelson, Associate Professor at MSU by
phone (517) 432-0272 or by email at nelsonc@msu.edu.
Thanks for helping to better understand and improve Michigan snowmobiling.
Sincerely,

Dr. Chuck Nelson,
Associate Professor
Enc.

135

BIBLIOGRAPHY

136

BIBLIOGRAPHY

Anderson, J. E. (1979). A theoretical foundation for the gravity equation. American economic
review, 69(1), 106-116.
Batty, M. (2007). "Spatial interaction." Encyclopedia of Geographic Information Science. Ed.
Thousand Oaks, CA: SAGE, 2007.417-19. SAGE Reference Online. Web. 20 Jan. 2012.
Baxter, M. J. (1979). The interpretation of the distance and attractiveness components in models
of recreational trips. Geographical Analysis, 11(3), 311-315.
Baxter, M. J., & Ewing, G. O. (1986). A framework for the exploratory development of spatial
interaction models: a recreation travel example. Journal of leisure research, 18(4), 320336.
Beilman, G. J., Brasel, K. J., Dittrich, K., Seatter, S., Jacobs, D. M., & Kevin Croston, J. (1999).
Risk factors and patterns of injury in snowmobile crashes. Wilderness & Environmental
Medicine, 10(4), 226-232.
Bertram, D. (2007). Likert scales. Calgary, Alberta, Canada: Retrieved May, 18, 2012.
Bricker, K. S., & Kerstetter, D. L. (2000). Level of specialization and place attachment: An
exploratory study of whitewater recreationists. Leisure Sciences, 22(4), 233-257.
Bryan, H. (1977). Leisure value systems and recreational specialization: The case of trout
fishermen. Journal of Leisure Research, 9(3), 174-187.
Bryan, H. (2000). Recreational specialization revisited. Journal of Leisure Research, 32(1), 1821.
Buchanan, T. (1985). Commitment and leisure behavior: A theoretical perspective. Leisure
Sciences, 7(4), 401-420.
Buckley, R. (2010). Adventure tourism management. Boston, MA: Butterworth-Heinemann.
Buckley, R. (2012). Rush as a key motivation in skilled adventure tourism: Resolving the risk
recreation paradox. Tourism Management, 33(4), 961-970.
Cheung, H. K. (1972). A say-use park visitation model. Journal of Leisure Research, 4(2), 13956.
Cole, J. S., & Scott, D. (1999). Segmenting participation in wildlife watching: A comparison of
casual wildlife watchers and serious birders. Human Dimensions of Wildlife, 4(4), 44-61.
137

Curry, L. A., & Youngblade, L. M. (2006). Negative affect, risk perception, and adolescent risk
behavior. Journal of applied developmental psychology, 27(5), 468-485.
Deery, H. A. (1999). Hazard and risk perception among young novice drivers. Journal of Safety
Research, 30(4), 225-236.
Delamater, P. L., Messina, J. P., Shortridge, A. M., & Grady, S. C. (2012). Measuring
geographic access to health care: raster and network-based methods. International
journal of health geographics, 11(1), 1-18.
Demirhan, G. (2005). Mountaineers’ risk perception in outdoor-adventure sports: a study of sex
and sports experience 1. Perceptual and motor skills, 100(3c), 1155-1160.
Ditton, R. B., Loomis, D. K., & Choi, S. (1992). Recreational specialization: Reconceptualization from a social worlds perspective. Journal of Leisure Research, 24(1),
33-51.
Donnelly, M. P., Vaske, J. J., & Graefe, A. R. (1986). Degree and range of recreational
specialization: Toward a typology of boating related activities. Journal of Leisure
Research, 18(2), 81-95.
Dunn, O. J. (1964). Multiple comparisons using rank sums. Technometrics, 6(3), 241-252.
Dyck, C., Schneider, I., Thompson, M., & Virden, R. (2003). Specialization among mountaineers
and its relationship to environmental attitudes. Journal of Park and Recreation
Administration, 21(2), 44-62.
Ewert, A. (1989). Outdoor adventure pursuits: Foundations, models and theories. Scottsdale,
AZ: Publishing Horizons.
Ewert, A., & Hollenhorst, S. (1994). Individual and setting attributes of the adventure recreation
experience. Leisure Sciences, 16(3), 177-191.
Fesenmaier, D. R., & Lieber, S. R. (1985). Spatial structure and behavior response in outdoor
recreation participation. Geografiska Annaler. Series B. Human Geography, 131-138.
Giles-Corti, B., & Donovan, R. J. (2002). The relative influence of individual, social and
physical environment determinants of physical activity. Social science & medicine,
54(12), 1793-1812.
Guagliardo, M. F. (2004). Spatial accessibility of primary care: concepts, methods and
challenges. International journal of health geographics, 3(1), 3.
Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis,
5th. NY: Prentice Hall International.
138

Hartigan, J. A., & Wong, M. A. (1979). Algorithm AS 136: A k-means clustering algorithm.
Journal of the Royal Statistical Society. Series C (Applied Statistics), 28(1), 100-108.
Hatfield, J., & Fernandes, R. (2009). The role of risk-propensity in the risky driving of younger
drivers. Accident analysis and prevention, 41(1), 25.
Hua, C. I., & Porell, F. (1979). A critical review of the development of the gravity model.
International Regional Science Review, 4(2), 97-126.
Hvenegaard, G. T. (2002). Birder specialization differences in conservation involvement,
demographics, and motivations. Human Dimensions of Wildlife, 7(1), 21-36.
International Snowmobile Manufacturers Association. (2013, June 2). Snowmobiling quick facts.
Retrieved from http://www.snowmobile.org/pr_snowfacts.asp
Jett, J. S., Thapa, B., & Ko, Y. J. (2009). Recreational specialization and boater speed
compliance in manatee zones. Human Dimensions of Wildlife, 14(4), 278-292.
Kent County, Michigan (2014, June 2). Volunteer Courtesy Trail Patrol. Retrieved from
https://www.accesskent.com/Sheriff/courtesytrail.htm
Keum, K. (2010). Tourism flows and trade theory: A panel data analysis with the gravity model.
The Annals of Regional Science, 44(3), 541-557.
Khadaroo, J., & Seetanah, B. (2008). The role of transport infrastructure in international tourism
development: A gravity model approach. Tourism management, 29(5), 831-840.
Khan, A. A. (1992). An integrated approach to measuring potential spatial access to health care
services. Socio-Economic Planning Sciences, 26(4), 275-287.
Kim, S. I., & Fesenmaier, D. R. (1990). Evaluating spatial structure effects in recreation travel.
Leisure Sciences, 12(4), 367-381.
Kuentzel, W. F., & Heberlein, T. A. (1992). Does specialization affect behavioral choices and
quality judgments among hunters? Leisure Sciences, 14(3), 211-226.
Kuentzel, W. F., & Heberlein, T. A. (2006). From novice to expert? A panel study of
specialization progression and change. Journal of Leisure Research, 38(4), 496-512.
Kuentzel, W. F., & McDonald, C. D. (1992). Differential effects of past experience,
commitment, and lifestyle dimensions on river use specialization. Journal of Leisure
Research, 24(3), 269-287.
Lee, S. H., Choi, J. Y., Yoo, S. H., & Oh, Y. G. (2013). Evaluating spatial centrality for
integrated tourism management in rural areas using GIS and network analysis. Tourism
Management, 34, 14-24.
139

Lee, J. H., & Scott, D. (2004). Measuring birding specialization: A confirmatory factor analysis.
Leisure Sciences, 26(3), 245-260.
Lee, J. H., & Scott, D. (2006). For better or worse? A structural model of the benefits and costs
associated with recreational specialization. Leisure Sciences, 28(1), 17-38.
Leung, S., & Starmer, G. (2005). Gap acceptance and risk-taking by young and mature drivers,
both sober and alcohol-intoxicated, in a simulated driving task. Accident Analysis &
Prevention, 37(6), 1056-1065.
Little, B. R. (1976). Specialization and the varieties of environmental experience. In
Experiencing the environment (pp. 81-116). Springer, US.
Lix, L. M., Keselman, J. C., & Keselman, H. J. (1996). Consequences of assumption violations
revisited: A quantitative review of alternatives to the one-way analysis of variance F test.
Review of educational research, 66(4), 579-619.
Luo, W. & Qi, Y. (2009). An enhanced two-step floating catchment area (E2SFCA) method for
measuring spatial accessibility to primary care physicians. Health & Place, 15(4), 11001107.
Luo, W. & Wang, F. (2003). Measures of spatial accessibility to health care in a GIS
environment: synthesis and a case study in the Chicago region. Environment and
Planning B: Planning and Design, 30, 865-884.
Lynch, J. A. (2000). Factors influencing snowmobilers' aberrant behaviors: a comparison of
1996-97 convicted snowmobile law violators and other snowmobilers (Doctoral
dissertation, Michigan State University. Department of Park, Recreation and Tourism
Resources).
Machin, M. A., & Sankey, K. S. (2008). Relationships between young drivers’ personality
characteristics, risk perceptions, and driving behaviour. Accident Analysis & Prevention,
40(2), 541-547.
Manning, R. E. (2011). Studies in outdoor recreation: Search and research for satisfaction, 3rd
Ed. Corvallis, OR: Oregon State University Press.
McBoyle, G., Scott, D., & Jones, B. (2007). Climate change and the future of snowmobiling in
non-mountainous regions of Canada. Managing Leisure, 12(4), 237-250.
McFarlane, B. L. (1996). Socialization influences of specialization among birdwatchers. Human
Dimensions of Wildlife, 1(1), 35-50.
McFarlane, B. L. (2004). Recreational specialization and site choice among vehicle-based
campers. Leisure Sciences, 26(3), 309-322.

140

McIntyre, N., & Pigram, J. J. (1992). Recreational specialization reexamined - The case of
vehicle-based campers. Leisure Sciences, 14(1), 3-15.
Michigan Department of Natural Resource. (2013, June 2). Snowmobiling in Michigan.
Retrieved from http://www.michigan.gov/dnr/0,4570,7-153-10365_14824---,00.html
Michigan Department of Natural Resource. (2014, June 2). Hunter Education. Retrieved from
http://www.michigan.gov/dnr/0,1607,7-153-10363_39267---,00.html
Michigan Department of Natural Resources. (2003). Snowmobile program 2002-2003 season
report. Law Enforcement Division, Michigan Department of Natural Resources, Lansing,
MI.
Michigan Department of Natural Resources. (2008). Snowmobile fatality report. Law
Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Michigan Department of Natural Resources. (2009). Snowmobile fatality summary 2008-2009.
Law Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Michigan Department of Natural Resources. (2010). Snowmobile fatality summary 2009-2010.
Law Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Michigan Department of Natural Resources. (2011). Snowmobile fatality summary 2010-2011.
Law Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Michigan Department of Natural Resources. (2012). Snowmobile fatality summary 2011-2012.
Law Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Michigan Department of Natural Resources. (2013). Snowmobile fatality summary 2012-2013.
Law Enforcement Division, Michigan Department of Natural Resources, Lansing, MI.
Morgan, C., & Stevens, C. A. (2008). Changes in perceptions of risk and competence among
beginning scuba divers. Journal of Risk Research, 11(8), 951-966.
National Safety Council (2014). Injury facts, 2014 Edition. Itasca, IL. See also:
http://www.nsc.org/
Needham, M. D. & Little, C. M. (2013). Voluntary environmental programs at an alpine ski area:
Visitor perceptions, attachment, value orientations, and specialization. Tourism
Management, 35, 70-81.
Needham, M. D., Vaske, J. J., Donnelly, M. P., & Manfredo, M. J. (2007). Hunting
specialization and its relationship to participation in response to chronic wasting disease.
Journal of Leisure Research, 39(3), 413.

141

Nelson, C., Lynch, J. and D. Stynes. (1998). An assessment of snowmobiling in Michigan by
snowmobilers with Michigan trail permits. Report to the Michigan Department of
Natural Resources Forest Management and Law Enforcement Divisions. East Lansing:
Michigan State University, Department of Park Recreation and Tourism Resources.
Nelson, C., Wu, I.-C., Stynes, D., & McCarty, E. (2009). Assessment of snowmobiling in
Michigan by snowmobilers with Michigan trail permits, 2009. Department of
Community, Agriculture, Recreation, and Resource Studies, Michigan State University,
East Lansing, MI.
Nourbakhsh, T. A. (2008). A qualitative exploration of female surfers: Recreational
specialization, motivations, and perspectives (Doctoral dissertation, California
Polytechnic State University).
Oh, C. O., & Ditton, R. B. (2006). Using recreation specialization to understand multi-attribute
management preferences. Leisure Sciences, 28(4), 369-384.
Oh, C. O., Ditton, R. B., Anderson, D. K., Scott, D., & Stoll, J. R. (2005). Understanding
differences in nonmarket valuation by angler specialization level. Leisure Sciences,
27(3), 263-277.
Oh, C. O., Lyu, S. O., & Hammitt, W. E. (2012). Predictive linkages between recreation
specialization and place attachment. Journal of Leisure Research, 44(1), 70-87.
Oh, C. O., Sorice, M. G., & Ditton, R. B. (2010). Exploring progression along the recreation
specialization continuum using a latent growth approach. Leisure Sciences, 33(1), 15-31.
Öström, M., & Eriksson, A. (2002). Snowmobile fatalities: Aspects on preventive measures from
a 25-year review. Accident Analysis & Prevention, 34(4), 563-568.
Pang-Ning, T., Steinbach, M., & Kumar, V. (2006). Introduction to data mining. In Library of
Congress.
Penchansky, R., & Thomas, J. W. (1981). The concept of access: definition and relationship to
consumer satisfaction. Medical care, 19(2), 127-140.
Pierz, J. J. (2003). Snowmobile injuries in North America. Clinical orthopaedics and related
research, 409, 29-36.
Radke, J., & Mu, L. (2000). Spatial decompositions, modeling and mapping service regions to
predict access to social programs. Geographic Information Sciences, 6(2), 105-112.
Rodrigue, J. P., Comtois, C., & Slack, B. (2013). The geography of transport systems. Routledge.
Rowe, B., Milner, R., Johnson, C., & Bota, G. (1992). Snowmobile-related deaths in Ontario: a
5-year review. CMAJ: Canadian Medical Association Journal, 146(2), 147.
142

Salz, R. J., & Loomis, D. K. (2005). Recreation specialization and anglers’ attitudes towards
restricted fishing areas. Human Dimensions of Wildlife, 10(3), 187-199.
Sawilowsky, S. & Blair, R. C. (1992). A more realistic look at the robustness and Type II error
properties of the t test to departures form population normality. Psychological Bulletin,
111, 352-360.
Schuurman, N., Berube, M., & Crooks, V. A. (2010). Measuring potential spatial access to
primary health care physicians using a modified gravity model. The Canadian
Geographer/Le Géographe canadien, 54(1), 29-45.
Scott, D., & Lee, J. H. (2010). Progression, stability, or decline? Sociological mechanisms
underlying change in specialization among birdwatchers. Leisure Sciences, 32(2), 180194.
Scott, D., & Shafer, C. S. (2001). Recreational specialization: a critical look at the construct.
Journal of Leisure Research, 33(3), 319-343.
Scott, D., & Thigpen, J. (2003). Understanding the birder as tourist: Segmenting visitors to the
Texas Hummer/Bird Celebration. Human Dimensions of Wildlife, 8(3), 199-218.
Scott, D., Baker, S. M., & Kim, C. (1999). Motivations and commitments among participants in
the Great Texas Birding Classic. Human Dimensions of Wildlife, 4(1), 50-67.
Scott, D., Ditton, R. B., Stoll, J. R., & Eubanks Jr, T. L. (2005). Measuring specialization among
birders: Utility of a self-classification measure. Human Dimensions of Wildlife, 10(1), 5374.
Sen, A., & Sööt, S. (1981). Selected procedures for calibrating the generalized gravity model.
Papers in Regional Science, 48(1), 165-176.
Sen, A. K., & Smith, T. E. (1995). Gravity models of spatial interaction behavior. Heidelberg:
Springer.
Shafer, C. S., & Hammitt, W. E. (1995). Purism revisited: Specifying recreational conditions of
concern according to resource intent. Leisure Sciences, 17(1), 15-30.
Smith, J. W., Burr, S. W., & Reiter, D. K. (2010). Specialization among off-highway vehicle
owners and its relationship to environmental worldviews and motivations. Journal of
Park and Recreation Administration, 28(2), 57-73.
Stevens, J. (2009). Applied multivariate statistics for the social sciences. Taylor & Francis US.
Stewart, R. L., & Black, G. B. (2004). Snowmobile trauma: 10 years' experience at Manitoba's
tertiary trauma centre. Canadian journal of surgery, 47(2), 90.

143

Stynes, D. (2009). State and regional economic impacts of snowmobiling in Michigan.
Department of Community, Agriculture, Recreation, and Resource Studies, Michigan
State University, East Lansing, MI.
Sugiyama, T., Francis, J., Middleton, N. J., Owen, N., & Giles-Corti, B. (2010). Associations
between recreational walking and attractiveness, size, and proximity of neighborhood
open spaces. American Journal of Public Health, 100(9).
Thapa, B., Graefe, A. R., & Meyer, L. A. (2006). Specialization and marine based environmental
behaviors among SCUBA divers. Journal of Leisure Research, 38(4), 601-615.
Vaske, J. J. (2008). Survey research and analysis: Applications in parks, recreation and human
dimensions. State College, PA: Venture Publishing.
Virden, R. J., & Schreyer, R. (1988). Recreational specialization as an indicator of environmental
preference. Environment and Behavior, 20(6), 721-739.
Weber, K. (2001). Outdoor adventure tourism: A review of research approaches. Annals of
Tourism Research, 28(2), 360-377.
Weiler, B., & Hall, C. M. (1992). Special interest tourism. London: Belhaven Press.
Widener, M. J., Farber, S., Neutens, T., & Horner, M. W. (2013). Using urban commuting data
to calculate a spatiotemporal accessibility measure for healthy food environment studies.
In Transportation Research Board 92nd Annual Meeting (No. 13-0825).
Wilson, A. G. (1967). A statistical theory of spatial distribution models. Transportation
research, 1(3), 253-269.
Witt, S. F., & Witt, C. A. (1995). Forecasting tourism demand: A review of empirical research.
International journal of Forecasting, 11(3), 447-475.
Won, D., Bang, H., & Shonk, D. J. (2008). Relative importance of factors involved in choosing a
regional ski destination: Influence of consumption situation and recreational
specialization. Journal of Sport & Tourism, 13(4), 249-271.
Wöran, B., & Arnberger, A. (2012). Exploring relationships between recreation specialization,
restorative environments and mountain hikers’ flow experience. Leisure Sciences, 34(2),
95-114.
Wu, I.-C., Borland, K., & Nelson, C. (2011). Proceedings from 2011 Northeastern Recreation
Research Symposium (NERR): Identifying the spatial pattern of off-road vehicle
accidents in Silver Lake State Park. Bolton Landing, NY, U.S.A. (in review)
Wu, I.-C., & Nelson, C. (2012). Presentation from 2012 Northeastern Recreation Research
Symposium (NERR): Mapping your customers to improve marketing decisions: Using
144

Michigan snowmobilers to expend tourist markets. The 2012 Northeastern Recreation
Research Symposium, Cooperstown, NY, U.S.A.
Zhang, X., Lu, H., & Holt, J. B. (2011). Modeling spatial accessibility to parks: a national study.
International journal of health geographics, 10(1), 31.

145