DISCRETE CHOICE MODELS OF HUNTING AND FISHING IN MICHIGAN
By
Scott Knoche

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

ABSTRACT
DISCRETE CHOICE MODELS OF HUNTING AND FISHING IN MICHIGAN
By
Scott Knoche
In this dissertation I employ multiple discrete choice modeling methods and data
collection approaches to make both methodological and empirical contributions which focus on
a variety of management-relevant programs and policies. In dissertation chapter 1, I examine
the potential for researchers to estimate recreation demand models using widely available
secondary data on recreation site choice that lacks some income and trip details. I find that
substituting zip-code median income for individual-reported income, and constructing
estimates of trips based on days spent hunting and the distance of the hunting site from the
individual’s residence, generate welfare estimates that are similar to results using individualreported information on income and trips. This research indicates potential for utilizing a wide
range of previously ignored recreation site choice data for recreation demand modeling
purposes. Dissertation chapters 2 and 3 are supported by the procedure developed in
dissertation chapter 1. Dissertation chapter 2 estimates a discrete choice model of ruffed
grouse hunting in Michigan with the objective of estimating both the economic benefits of all
publicly accessible hunting land in Michigan for ruffed grouse hunters and the potential
changes to ruffed grouse hunter economic benefits as a result of a proposal to ban the use of
firearms in approximately 67,000 acres of national forest wilderness area. Economic benefits of
publicly accessible hunting land for ruffed grouse hunters in Michigan was estimated to be
about $30 million in 2008, with economic benefits of the 67,000 acres of national forest

wilderness area to ruffed grouse hunters estimated at about $45,000. Dissertation chapter 3
estimates the economic benefits of the USDA Conservation Reserve Program for pheasant
hunters in Michigan by linking a previously developed spatially explicit, landscape based model
of pheasant sightings to a discrete choice model of pheasant hunter site choice. Chapter 3 also
estimates the potential economic benefits generated through the initiation of a multi-agency
and stakeholder pheasant habitat restoration plan. Results show that economic benefits for
pheasant hunters depend critically on restoration site selection. Dissertation chapter 4 uses a
choice experiment survey of Michigan trout anglers to examine the willingness of anglers to
make tradeoffs between changes in driving distance to a fishing site and changes in attributes
available at fishing sites. On average, trout anglers prefer higher catch rates, shorter travel
distances to a fishing site, and are highly averse to strict fishing regulations such as catch-andrelease only and artificial flies only regulations. However, there is evidence of preference
heterogeneity within the trout angler population for these regulations, with some anglers
preferring (all else equal) to fish in areas with these restrictions. This distribution of angler
preferences is used to examine the proportion of anglers who can be made better off when
strict regulations induce catch-related quality improvements at a fishing site.

To Rebecca

iv

ACKNOWLEDGEMENTS

First, I would like to thank my major professor and Ph.D. committee chair Frank Lupi.
Frank has had a tremendous influence over my development as a researcher and has provided
superior guidance and encouragement over the past years, dating back to my Master’s program
at MSU. Thanks also to my other Ph.D. committee members John Hoehn, Dan Hayes and Dan
Kramer for their support and feedback on research, especially previous drafts of this
dissertation. This research would not have been possible without support from the Michigan
Department of Natural Resources (MDNR). Brian Frawley of the MDNR Wildlife Division
provided assistance with obtaining wildlife harvest survey data used to support much of this
research, and Al Stewart of MDNR Wildlife provided support and encouragement for my
economic analyses of upland gamebird hunting. Many thanks to personnel from MDNR
Fisheries Division, in particular Troy Zorn and the MDNR Coldwater Resource Steering
Committee, for providing input and guidance with respect to the development of a survey of
Michigan trout anglers. Other friends and colleagues contributed to specific aspects of this
research. Jody Simoes and Scott Weiksel assisted with various aspects of the development and
implementation of the Michigan trout angler survey. Ashley Suiter provided critical GIS support
which facilitated the analysis of the economic benefits of restored lands for pheasant hunters.
Jonathon Siegle provided much needed help with trout survey data entry and management.
Research within this dissertation was made possible by funding through MSU Ag-Bio Research,
the Michigan Department of Natural Resources and Trout Unlimited. Finally, I would like to
thank my wife, Rebecca Held Knoche, for her support and love throughout the entire process.
v

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................................. viii
LIST OF FIGURES .............................................................................................................................. x
KEY TO ABBREVIATIONS................................................................................................................ xii
INTRODUCTION .............................................................................................................................. 1
ACCESS ......................................................................................................................................... 3
HABITAT MANAGEMENT AND GAME/FISH POPULATIONS ........................................................ 4
REGULATIONS .............................................................................................................................. 6
REFERENCES ................................................................................................................................ 8
CHAPTER 1: WELFARE IMPLICATIONS OF USING IMPERFECT SECONDARY DATA TO ESTIMATE
RECREATION DEMAND MODELS.................................................................................................. 10
1.1 Introduction ......................................................................................................................... 10
1.2 Literature Review ................................................................................................................ 14
1.3 Methods .............................................................................................................................. 17
1.3.1 Economic Theory .......................................................................................................... 17
1.3.2 Empirical Model ............................................................................................................ 19
1.3.3 Baseline Travel Cost Models - Archery Hunting and Firearm Hunting ......................... 20
1.3.4 Model Re-Estimation Using Proxy for Income ............................................................. 22
1.3.5 Model Re-Estimation Using Proxy for Trips.................................................................. 22
1.4 Results ................................................................................................................................. 30
1.5 Discussion ............................................................................................................................ 35
APPENDIX .................................................................................................................................. 39
REFERENCES .............................................................................................................................. 41
CHAPTER 2: ESTIMATING THE ECONOMIC BENEFITS OF PUBLICLY ACCESSIBLE LAND FOR
RUFFED GROUSE HUNTERS .......................................................................................................... 46
2.1 Introduction ......................................................................................................................... 46
2.2 Study Area ........................................................................................................................... 49
2.2.1 Policy Scenario .............................................................................................................. 51
2.3 Methods .............................................................................................................................. 53
2.3.1 Data............................................................................................................................... 53
2.3.2 Economic Model ........................................................................................................... 55
2.4 Results ................................................................................................................................. 56
2.5 Discussion ............................................................................................................................ 62
2.6 Management Implications .................................................................................................. 65
REFERENCES .............................................................................................................................. 67

vi

CHAPTER 3: HARVESTING BENEFITS FROM HABITAT RESTORATION – THE INFLUENCE OF
LANDSCAPE POSITION ON ECONOMIC BENEFITS TO PHEASANT HUNTERS .............................. 72
3.1 Introduction ......................................................................................................................... 72
3.2 Methodology ....................................................................................................................... 76
3.2.1 Ecological Production Function .................................................................................... 76
3.2.2 Discrete Choice Recreation Demand Model ................................................................ 77
3.2.3 Estimating Economic Benefits for Site Quality Changes .............................................. 79
3.3 Survey and Attribute Data ................................................................................................... 80
3.3.1 Computing Pheasant Sightings using the Nielson Model ............................................. 85
3.4 Results ................................................................................................................................. 90
3.4.1 Pheasant Hunter Benefits of CRP Land ........................................................................ 91
3.4.2 Michigan Pheasant Restoration Initiative (MPRI) Policy Scenario ............................... 92
3.5 Discussion ............................................................................................................................ 99
REFERENCES ............................................................................................................................ 103
CHAPTER 4: A DISCRETE CHOICE EXPERIMENT APPROACH EXAMINING TROUT ANGLER
PREFERENCES FOR FISHING SITE ATTRIBUTES .......................................................................... 110
4.1 Introduction ....................................................................................................................... 110
4.2 Study Area ......................................................................................................................... 112
4.3 Material and Methods....................................................................................................... 115
4.3.1 Discrete Choice Method ............................................................................................. 115
4.3.2 Stated Preference Discrete Choice Survey ................................................................. 117
4.3.3 Random Utility Theory and Model Estimation ........................................................... 126
4.3.4 Mixed Logit Model ...................................................................................................... 127
4.3.5 Estimating Willingness to Pay and Preference Heterogeneity ................................... 129
4.4 Results ............................................................................................................................... 130
4.4.1 Management Regulation Policy Scenarios ................................................................. 134
4.5 Discussion .......................................................................................................................... 136
APPENDICES............................................................................................................................. 141
APPENDIX A ......................................................................................................................... 142
APPENDIX B.......................................................................................................................... 146
APPENDIX C.......................................................................................................................... 159
APPENDIX D ......................................................................................................................... 161
APPENDIX E .......................................................................................................................... 171
APPENDIX F .......................................................................................................................... 176
APPENDIX G ......................................................................................................................... 178
REFERENCES ............................................................................................................................ 181

vii

LIST OF TABLES

Table 1.1. Peer-reviewed literature examining impacts of assumptions regarding the
opportunity cost of time, driving costs, and the nature of trips, on welfare estimates .............. 15
Table 1.2. OLS regression of the number of hunting trips taken to the site during the season on
the number of days taken to the site during the season and the distance to the hunting site
from the individual’s residence .................................................................................................... 26
Table 1.3. Ad-hoc trips proxy and empirically-based trips proxy variables .................................. 28
Table 1.4. Baseline Firearm and Archery model results using hunter-reported information on
trips and income ........................................................................................................................... 31
Table 1.5. Differences in per-trip and seasonal welfare estimates for publicly accessible hunting
land between baseline models estimated using hunter reported income and models reestimated using aggregate census data for income ..................................................................... 32
Table 1.6. Per-trip and economic benefits for firearm hunters and archery hunters for
empirically-based and ad-hoc proxies for the number of hunting trips ...................................... 34
Table 1. Deer hunting travel cost model results using seven empirically-based estimates of trips
as a function of days and distance ................................................................................................ 40
Table 2. Deer hunting travel cost model results using seven assumption-based estimates of trips
as a function of days and distance ................................................................................................ 40
Table 2.1. Michigan ruffed grouse hunter demographics and hunting site choices by region of
residence, 2008–2009 ................................................................................................................... 58
Table 2.2. Conditional logit results for a travel cost model of ruffed grouse hunting in Michigan
....................................................................................................................................................... 60
Table 3.1. Nested logit model of pheasant hunting in Michigan ................................................. 91
Table 3.2. Seasonal welfare and per-acre welfare of MPRI restoration siting scenarios ............. 97

viii

Table 4.1. Fishing site attribute levels for choice experiment.................................................... 121
Table 4.2. Fishing site attributes variables in indirect utility function ....................................... 127
Table 4.3. Mixed logit model results for Michigan trout fishing choice experiment ................. 132
Table 4.4. Per-trip willingness to pay for statistically significant (P<0.05) fishing site attribute
variables ...................................................................................................................................... 134
Table 4.5. Current proportion of anglers receiving positive utility from regulations (scenario 1)
and the proportion of anglers receiving positive utility when the regulation is accompanied by a
catch-related site quality improvement (scenarios 2-4). ........................................................... 136
Table A.1. Summary of trout angler survey response and disposition ...................................... 144
Table A.2. Survey weighting........................................................................................................ 145
Table E.1. Paired choice alternatives and attribute levels ......................................................... 172
Table F.1. Pearson product moment correlations for attributes within and across choice
alternatives in choice pairs ......................................................................................................... 177
Table G.1. Mail survey choice scenario groupings by survey number ....................................... 179
Table G.2. Internet survey choice scenario groupings by survey number ................................. 180

ix

LIST OF FIGURES

Figure 1.1. Michigan hunting site choices .................................................................................... 20
Figure 1.2. Average number of days spent hunting per firearm and archery (bow) hunting trip by
distance from residence ............................................................................................................... 24
Figure 3.1. Michigan lower peninsula counties and landscape parcels used to estimate pheasant
sightings ........................................................................................................................................ 87
Figure 3.2. County-level estimates of pheasant abundance from the Nielson model and a
MDNR mail carrier pheasant brood survey and the delineation between regional nests in nested
logit model .................................................................................................................................... 89
Figure 3.3 MPRI priority pheasant restoration areas within Michigan’s lower peninsula ........... 93
Figure 3.4. Parcel-level economic benefit estimates from converting 1,600 acres of agricultural
land to CRP herbaceous land ........................................................................................................ 98
Figure 4.1. Major categories of Michigan trout fishing regulations ........................................... 113
Figure 4.2. Example Michigan stream trout fishing site choice scenario .................................. 119
Figure 4.3. Coldwater trout streams in Michigan ....................................................................... 123
Figure B.1. Mail survey page 1 ................................................................................................... 147
Figure B.2. Mail survey page 2 .................................................................................................... 148
Figure B.3. Mail survey page 3 .................................................................................................... 149
Figure B.4. Mail survey page 4 .................................................................................................... 150
Figure B.5. Mail survey page 5 .................................................................................................... 151
Figure B.6. Mail survey page 6 .................................................................................................... 152

x

Figure B.7. Mail survey page 7 .................................................................................................... 153
Figure B.8. Mail survey page 8 .................................................................................................... 154
Figure B.9. Mail survey page 9 .................................................................................................... 155
Figure B.10. Mail survey page 10 ................................................................................................ 156
Figure B.11. Mail survey page 11 ................................................................................................ 157
Figure B.12. Mail survey page 12 ................................................................................................ 158
Figure C.1. Choice experiment format #2 ................................................................................... 160
Figure D.1. First mail contact (front)........................................................................................... 162
Figure D.2. First mail contact (back) ........................................................................................... 163
Figure D.3. Envelope for first mail contact ................................................................................. 164
Figure D.4. Second mail contact (postcard front) ....................................................................... 165
Figure D.5. Second mail contact (postcard back) ....................................................................... 166
Figure D.6. Third mail contact (postcard front example) ........................................................... 167
Figure D.7. Third mail contact (postcard back) ........................................................................... 168
Figure D.8. Fourth mail contact (letter front) ............................................................................. 169
Figure D.9. Fourth mail contact (letter back).............................................................................. 170

xi

KEY TO ABBREVIATIONS

AAA

American Automobile Association

Ag

Agriculture

BBS

Breeding Bird Survey

CFA

Commercial Forest Act

CRP

Conservation Reserve Program

CS

Consumer Surplus

Diff

Difference

ha

hectare

IJI

Interspersion and Juxtaposition Index

LWCF

Land and Water Conservation Fund

Max

Maximum

MDNR

Michigan Department of Natural Resources

Min

Minimum

MPRI

Michigan Pheasant Restoration Initiative

MPS

Mean Patch Size

NGO

Non Governmental Organization

NLCD

National Land Cover Dataset

Obs

observations

OLS

Ordinary Least Squares

Stats

Statistics

xii

USDA

United States Department of Agriculture

USDOI

United States Department of Interior

WTP

Willingness to Pay

xiii

INTRODUCTION
Hunting and fishing are popular outdoor recreational activities in the United States, with
33.1 million people fishing and 13.7 million people hunting at least once in 2011 (USDOI 2013).
The quality of these recreational experiences is affected by resource manager decisions
regarding the allocation of scarce budgetary dollars across various management activities to
sustain viable fish and wildlife populations. Understanding individual preferences for aspects of
hunting and fishing experiences is critical for enabling state and federal fisheries and wildlife
management agencies to identify management actions which would provide the greatest
benefit to hunters and anglers. Characterizing these preferences in terms of the monetary
tradeoffs individuals are willing to make for quality changes in hunting and fishing site
attributes can improve resource allocation and facilitate cost-benefit analysis.
The primary objective of this dissertation is to utilize discrete choice statistical modeling
tools to examine the preferences of hunters and anglers for different aspects of hunting and
fishing experiences. Through the discrete choice approach, a researcher can examine hunter or
angler preferences and tradeoffs for attributes of hunting and fishing locations by analyzing the
choice of a hunting or fishing site from a choice set of two or more sites, with each site
described by relevant site characteristics (e.g., travel cost, public land, catch/harvest rate).
There are two primary types of discrete choice approaches – revealed preference and stated
preference – with both approaches used in this dissertation. The revealed preference discrete
choice approach involves analyzing preferences by using individuals’ actual choices (observed
or reported), whereas the stated preference discrete choice approach is a survey-based
approach which elicits preferences through individuals’ selection of a preferred alternative
1

from a set of alternatives, with each alternative’s attributes described explicitly within the
choice format.
The discrete choice analysis within this dissertation begins in chapter 1 with a
methodological contribution in which a procedure is developed to utilize previously collected
data to estimate revealed preference discrete choice models of hunter site choice. Federal and
state fisheries and wildlife management agencies routinely administer surveys to collect
information on wildlife harvest, and such data offers substantial potential for cost-effectively
using revealed preference approaches to examine hunter preferences and estimate economic
benefits associated with hunting site attributes. The analysis in dissertation chapter 1 provides
evidence that travel cost model welfare estimates obtained through the use of assumptions
necessary to overcome wildlife harvest data limitations are comparable to welfare estimates
obtained when using data which does not require such assumptions. In dissertation chapters 2,
3, and 4, I focus on using discrete choice approaches to address pressing and critical aspects of
management concern: Access, catch/harvest rate, habitat management, and regulations.
Abundant game or fish species in a particular location is of little benefit to hunters or anglers if
they are unable to access the land or a body of water. Correspondingly, abundant access for
hunting and fishing matters little if fish and game species are not present in sufficient numbers
to make targeting them a worthwhile endeavor. Finally, hunting and fishing regulations act as
potential constraints to interactions hunters and anglers have with fish and wildlife resources.
Below, I provide background and detail the rationale for understanding hunter and angler
preferences for a) access, b) habitat management & catch/harvest rate, and c) regulations, and
I briefly describe the contribution of this dissertation with respect to these three categories.
2

ACCESS
Access to both publicly-owned and privately-owned lands has been identified as two of
the most important issues facing hunting today. A 2002 survey of licensed hunters nationwide
reported that 22% and 18% of hunters identified access to public land and private land,
respectively, as being one of the two most important issues facing hunting today (Duda et al.
2004). Actions at both the national and state level have been undertaken to facilitate access to
public and private lands. At the national level, the United States Department of Agriculture’s
Voluntary Public Access and Habitat Incentive Program provided grants to 25 states totaling
$17.8 million to encourage private landowners to make their land available to the public for
hunting. Further, President Obama launched the America’s Great Outdoors initiative in 2010,
which recommends full funding of $900 million per year (by 2014) for the Land and Water
Conservation Fund, a federal funding program which over the past 40 years has enabled the
purchase of about 3 million acres of recreation lands and the funding of over 29,000 projects to
develop basic recreational facilities on public lands. In Michigan, publicly-owned hunting access
options include State Forest (1.5 million ha), National Forest (1.2 million ha), State Game and
Wildlife areas (200,000 ha), State Parks and Recreation areas (107,000 ha), two National
Lakeshores (52,000 ha), and United States Fish and Wildlife Service Land (47,000 ha).
Additionally, about 890,000 ha of privately-owned, publicly accessible land is available to
hunters through Michigan’s Commercial Forest Act (CFA), which provides tax incentives for
private landowners (predominantly timber harvesting companies) to allow hunting and angling
on their land.

3

Given the importance to hunters of being able to access lands and the high importance
natural resource managers are currently placing on facilitating access, there has been a
surprising lack of attention in the published literature regarding the relationship of hunting
access to hunter site choice and the economic benefits hunters derive from this access.
Research within examines the economic benefits of publicly accessible hunting lands for ruffed
grouse hunters in Michigan, helping to fill this gap in the literature while building on my
previous research on the economic benefits of publicly accessible lands for deer hunters
(Knoche and Lupi 2012). Within this dissertation, I use the revealed preference travel cost
method to estimate economic benefits with respect to ruffed grouse hunting in Michigan for
the three main types of publicly accessible hunting land: Federally-owned land, State-owned
land, and privately-owned, publicly accessible CFA land. Additionally, I perform a managementrelevant policy analysis by examining the implications to ruffed grouse hunters from the closing
of a portion of federally-owned land to firearm hunters.

HABITAT MANAGEMENT AND GAME/FISH POPULATIONS
Natural resource managers employ habitat and fish/game management strategies to
produce satisfactory catch/harvest outcomes from participating in hunting and fishing
activities. Duda et al. (2004) found that 18% of hunters identified loss of habitat as being one of
two most important issues facing hunting today, whereas 13% of hunters identified game
management as being one of two most important issues. Extensive and expensive habitat
restoration and conservation approaches which have the potential to improve catch/harvest

4

outcomes desired by hunters and anglers are undertaken by federal, state, local and non-profit
organizations. According to Bernhardt et al. (2005), between 14 and 15 billion dollars were
spent on over 37,000 river/stream restoration projects in the continental U.S. from 1990 to
2003, many of which have the potential to improve fisheries. The United States Department of
Agriculture (USDA) Conservation Reserve Program (CRP) has provided farmers and landowners
with rental payments and cost-share assistance to engage in beneficial land management
practices totaling over $41.3 billion since the CRP was signed into law in 1985 (USDA 2013).
CRP land has been shown by a number of researchers (see, e.g., Gould and Jenkins 1993,
Kantrud 1993, Riley 1995, Nielson et al. 2008 and Schroeder and Vander Haegen 2011) to be
beneficial for wildlife species pursued by hunters. Fish propagation and stocking is a commonly
used tool used by state and federal agencies to improve catch-related attributes for
recreational fishing, such as catch rate and availability of particular species. In 2004, state and
federal agencies reported stocking an estimated 1.75 billion fish in the waters of the U.S.
(Halverson 2008).

Allocating budgetary resources across management alternatives with the objective of
maximizing benefits to hunters and anglers requires an understanding of hunter and angler
preferences for catch-related aspects of hunting and fishing activities. Research within this
dissertation links a spatially explicit landscape-based ecological production function of pheasant
abundance to a revealed-preference discrete-choice model of pheasant hunters’ site choices
providing valuable information on the economic benefits of USDA Conservation Reserve
Program Land to pheasant hunters in Michigan. The economic benefits of all CRP land to
pheasant hunters in Michigan’s lower peninsula is estimated (relative to all CRP land being
5

eliminated), and a management-relevant analysis is also provided through the estimation of the
economic benefits accruing to pheasant hunters through plausible restoration scenarios
occurring through the implementation of the Michigan Pheasant Restoration Initiative (MPRI).

Additionally, this dissertation explores the aspects of the Michigan stream trout fishing
experience which are most important to anglers. Through the use of a stated preference choice
experiment, I analyze angler fishing site choice as a function of the attributes available at
different possible fishing locations. Characterizing angler preferences for fishing site attributes
by using distance as a common trade-off metric facilitates the comparison of angler preferences
for these attributes. Understanding relative preferences for catch attributes can help fisheries
managers more effectively allocate limited budgetary resources to stocking programs and
habitat management actions likely to improve the catch-related aspects of the stream trout
fishing experience most preferred by anglers.

REGULATIONS
Natural resource managers implement and enforce a variety of regulations with the
objective of limiting excessive impacts to fish and wildlife populations and resulting reductions
in the quality of recreational opportunities for a wide variety of stakeholders. Fisheries and
wildlife managers regulate hunters and anglers through restrictions on harvest, species
targeted, and gear/method which vary along spatial and temporal dimensions. Hunters and
anglers may have different levels of tolerance for such regulations. For anglers with slightly or
moderately negative preferences for regulations, regulation-induced improvements in catch
6

and harvest may make some of these anglers better off than a scenario in which regulations are
relaxed but catch/harvest rates are low. Additionally, managers of publicly-owned lands may
be confronted with recreational use conflicts between different user groups and thus consider
restricting outdoor recreation by user groups (e.g., firearm hunters) at specific locations. An
informed decision regarding whether or not to implement such restrictions would involve
examining the impacts to all user groups affected (both positively and negatively) from such
restrictions.

It is important for resource managers to understand the potential impacts of regulations
on hunter and angler site choice and their enjoyment of hunting and fishing activities. Research
within this dissertation explores stream trout angler preferences for fishing site attributes and
estimates the proportion of anglers who have positive and negative preferences for restrictions
on trout harvest and tackle restrictions. This then allows for an examination of whether and to
what extent trout anglers negatively impacted by regulations can be made better-off through
regulation-induced improvements in catch-related attributes. Finally, this dissertation
examines the potential welfare impacts to hunters from a recreational use conflict which
results in the loss of hunting rights. Drawing upon a recent contentious debate over firearm
hunting in Michigan National Forest Wilderness areas, I examine the impact to ruffed grouse
hunters from eliminating 67,000 acres of National Forest land as a hunting option.

7

REFERENCES

8

REFERENCES

Bernhardt, E.S., M.A. Palmer, J.D. Allan, G. Alexander, K. Barnas, S. Brooks et al. 2005.
Synthesizing US River Restoration Efforts. Science 308, 636–37.
Duda, M., P.E. De Michele, C. Zurawski, M. Jones, J.E. Yoder, W. Testerman, A. Lanier, S.J.
Bissell, P. Wang, and J.B. Herrick. 2004. Issues Related to Hunting and Fishing
Access in the United States: A Literature Review, Responsive Management,
Harrisonburg, VA.
Gould, J.H., and K.J. Jenkins. 1993. Seasonal Use of Conservation Reserve Program Lands by
White-Tailed Deer in East-Central South Dakota. Wildlife Society Bulletin 21: 250-255.
Halverson, M.A. 2009. Stocking Trends: A Quantitative Review of Governmental Fish
Stocking in the United States, 1931 to 2004. Fisheries 33: 69–75.
Kantrud, H.A. 1993. Duck Nest Success on Conservation Reserve Program Land in the Prairie
Pothole Region. Journal of Soil and Water Conservation 48: 238-242.
Knoche, S., and F. Lupi. 2012. The Economic Value of Publicly Accessible Deer Hunting Land.
Journal of Wildlife Management 76: 462–470.
Nielson, R.M., L.L. McDonald, J.P. Sullivan, C. Burgess, D.S. Johnson, D.H. Johnson, S. Bucholtz,
S. Hyberg, and S. Howlin. 2008. Estimating the Response of Ring-Necked Pheasants
(Phasianus colchicus) to the Conservation Reserve Program. The Auk 125: 434-444.
Riley, T.Z. 1995. Association of the Conservation Reserve Program with Ring-Necked Pheasant
Survey Counts in Iowa. Wildlife Society Bulletin 23: 386-390.
Schroeder, M.A., and W.M. Vander Haegen. 2011. Response of Greater Sage-Grouse to the
Conservation Reserve Program in Washington State. Studies in Avian Biology 38: 517530.
United States Department of Agriculture (USDA) Farm Service Agency. 2013. Conservation
Reserve Program – Rental payments by Fiscal Year. Retrieved from:
http://www.fsa.usda.gov/FSA/webapp?area=home&subject=copr&topic=rns-css
U.S. Department of the Interior [USDOI], Fish and Wildlife Service, and U.S. Department of
Commerce, U.S. Census Bureau. 2013. 2011 National Survey of Fishing, Hunting, and
Wildlife-associated Recreation. Washington, D.C., USA.

9

CHAPTER 1: WELFARE IMPLICATIONS OF USING IMPERFECT SECONDARY DATA TO ESTIMATE
RECREATION DEMAND MODELS

1.1 Introduction
Economists, through the application of non-market valuation methods, contribute to
the efficient management of environmental goods and services by providing resource managers
and policy makers with economic benefits estimates associated with changes in environmental
quality. For environmental catastrophes such as a major oil spill, the potential for substantial
reductions in economic welfare resulting from area closures and/or environmental quality
changes can easily justify costly non-market valuation studies to obtain accurate and precise
willingness-to-pay estimates used to inform compensation decisions and support litigation
efforts. Outside of the public spotlight, however, resource managers are routinely making
lower-visibility budgetary allocation decisions to improve environmental quality characteristics,
while often lacking the funds necessary to inform these decisions through a full-scale study of
stakeholder preferences and economic benefits. It is incumbent upon applied economists
working within the natural resource management field to find cost-effective ways to contribute
meaningfully to this decision-making process. In some situations, the benefits transfer
approach can be a simple and cost-effective way to contribute to the discussion – however,
spatial and/or temporal differences between the initial study and the intended area of
application may be too great to apply this method with a reasonable degree of confidence. As
a third alternative, utilizing “secondary data” collected by resource managers may be an

10

attractive option, providing up-to-date, location-specific information on individual preferences
for recreation sites, while avoiding the high costs associated with primary data collection.
Resource managers concerned with the sustainable management of natural resources
often collect spatially explicit information regarding the nature and intensity of recreational use
to quantify user impact on a resource. Such information is regularly collected by state and
federal agencies through hunter harvest and angler creel surveys, and may also be available for
mountain biking, hiking, white-water rafting and other outdoor recreational activities where
user-impacts and congestion are issues of management concern. In particular, hunter harvest
surveys appear to be a particularly robust source of spatially explicit recreation site choice data.
Through the Pittman-Robertson Wildlife Restoration Act (1937), each state in the U.S. has the
primary responsibility and authority over the hunting of wildlife within state boundaries, and
many of these states administer regular wildlife harvest surveys that help monitor game
populations and assist in the development of hunting regulations. While the specifics of these
hunter harvest surveys vary across states and game species of interest, the surveys generally
collect data on the hunting site choice (e.g., county or game management unit) of hunters, the
number of days spent hunting, and quantity of targeted species harvested. For example, in
Oregon, each hunter is required to report harvest and effort information (i.e., number of days
hunted and location hunted) for each hunting tag purchased. The Michigan Department of
Natural Resources (MDNR) surveys big game hunters (e.g., deer, elk, bear) and small game
hunters (e.g., ruffed grouse, pheasant, rabbit, and squirrel) to estimate hunter participation,
harvest, and hunting effort. Colorado Division of Wildlife conducts surveys of big game and
small game hunters regarding hunting area, species harvested, and number of days spent
11

hunting. While states have the primary authority over managing resident wildlife, the U.S.
Federal government has the ultimate responsibility for developing regulations regarding
migratory bird hunting. Migratory bird hunters are required by federal law to register with the
Migratory Bird Harvest Information Program; hunters are then randomly selected and mailed
surveys (survey return is voluntary). The survey asks hunters to provide information regarding
their migratory bird hunting experience, such as the locations hunted and the number of days
spent hunting at each location.
At first glance, secondary data that captures individuals’ recreation site selection
decisions during some time period (e.g., calendar year, hunting/fishing season) would seem
suitable for analysis using a revealed preference non-market valuation approach such as the
multiple-site random utility travel cost model. However, as this data collection is generally
designed to understand the nature and extent of user-impacts on natural resources, rather
than to understand the motivations of recreationists, this can result in challenging data
measurement and data omission issues from the perspective of a researcher seeking to apply
non-market valuation techniques. A common limitation of such recreational site use data is the
measurement of site choice in terms of the number of days spent at a site during a particular
time period, rather than the number of trips taken to the site. If it cannot be reasonably
assumed that all trips are single day trips, but instead that a day at a recreation site may be a
component of a multiple day trip of undetermined length, constructing the travel cost variable
(generally considered to be a function of trip-level expenses such as driving costs and
opportunity costs of time) becomes problematic. Another common limitation faced by
researchers who might consider using secondary data to estimate a travel cost model is a lack
12

of information regarding the income of the individual. The opportunity cost of time associated
with traveling to and from a site is considered to be key component of the price of a trip, and
this is generally estimated as a fraction of an individual’s wage rate (Parsons 2003).
The research within examines the possibility of making available a large body of
imperfect site choice data for use by recreation demand modelers on limited budgets to
estimate economic benefits associated with popular outdoor recreational activities such as
hunting (13.7 million U.S. hunters; 282 million days hunting [USDOI 2011]). I examine the
potential to overcome two common limitations of secondary data sets with respect to
estimating travel cost models of recreational site choice – 1) missing information on individual
income, and 2) site choice information measured in days rather than trips. Using data from a
2003 survey of Michigan deer hunters (response rate = 68%), I estimate a travel cost model of
deer hunter site choice using complete data on hunter income and trips and compare the
economic benefits from this baseline model to economic benefits when the model is reestimated using a variety of variable proxies as substitutes for the missing data on income and
trips. The empirical strategy is as follows. First, baseline travel cost models of firearm and
archery hunting are estimated using all available data (i.e., hunter-reported income and hunterreported trips). Next, variable proxies for individual income and the number of trips
(measurements which would be missing in much of secondary data) are constructed and these
proxies are used in model re-estimation. The proxy for income is constructed by assigning each
individual the U.S. Census-based median household income for their zip code of residence.
Proxies for missing trips are constructed using a variety of possible functional relationships
between the number of trips to a hunting site and both the number of days spent at the
13

hunting site and the distance from the individual’s residence to the hunting site. The impacts
on economic benefits resulting from assumptions on income and trips are compared to the
impacts from researcher best-judgments and assumptions commonly employed by recreation
demand modelers.

1.2 Literature Review
To provide context to the research findings regarding the differences in economic
benefits resulting from the use of proxy variables, it is helpful to examine the existing literature
to understand the welfare implications of researcher-imposed best-judgments and
assumptions. First, though it is fairly common within the recreation demand literature to assign
an individual the mean or median U.S. Census income of a geopolitical region when lacking
individual-specific data on income (see, e.g., Heberling and Templeton 2009, Knoche and Lupi
2007, Knoche and Lupi 2012, Knoche and Lupi 2013, Whitehead et al. 2009, Henderson et al.
1999, Jakus et al. 1997, Cameron 1992), I am aware of no studies which examine the welfare
implications of this assumption. Additionally, there does not appear to be any literature which
examines the potential for making use of a large body of outdoor recreation participation data
which measures recreation site use in terms of days rather than trips.

14

Table 1.1. Peer-reviewed literature examining impacts of assumptions regarding the
opportunity cost of time, driving costs, and the nature of trips, on welfare estimates
Authors

Baseline
Comparison
Estimates
Opportunity Cost of Time (fraction of wage rate)
Cesario (1976)
0
1/3
11
1/4
1
Bishop and Heberlein
0
(1979)
1/2
1
Wilman and Pauls (1987)
1/3
1
6
Rockel and Kealy (1991)
3/10
3/5
7
1/4
4
Bowker et al. (1996)
0
1/2
4
3/10
48
Layman et al. (1996)
0
3/5
48
1
48
1/4
4
Zawaki et al. (2000)
0
1/2
4
1/4
14
Bin et al. (2005)
1/3
1
14
Hynes et al. (2009)
0
1
4
1/3
2
Edwards et al. (2011)
0
1
2
Driving Costs
Bowker et al. (1996)
Imputed
Reported
4
Imputed
Reported
12
English and Bowker (1996)
$0.25/mile
2
$0.15/mile
$0.35/mile
2
Layman et al. (1996)
Imputed
Reported
64
Prayaga et al. (2010)
Imputed
Reported
2
Nature of Trips
Multi-Destination
Multi-Destination
Mendelsohn et al. (1992)
2
Excluded
Included
Parsons and Wilson (1997) Incidental Excluded Incidental Included
2
Loomis et al. (2000)

Incidental Excluded

CS Diff.
15% to 35%
143%
300%
15% to 101%
10% to 20%
2% to 52%
43% to 127%
-14% to 67%
34% to 163%
81% to 193%
13% to 104%
41% to 259%
-38% to 17%
17% to 391%
29% to 85%
101% to 120%
349% to 450%
5% to 50%
-21% to 570%
61% to 64%
123% to 130%
7% to 113%
-66% to -56%
65% to 77%
4% to 8%

Incidental Included
1
72%
Single Destination
1
1516%
Multi-Destination
Excluded
Loomis (2006)
Excluded
All Trips Included
1
137%
Notes: CS diff. is difference in consumer surplus between the baseline & comparison model
assumptions, imputed costs are per-mile driving cost estimates, reported costs are travel costs reported
by individuals, & Estimates is the number of comparisons between baseline & comparison model.

15

The effect of many other researcher-imposed best-judgments and assumptions on
coefficient estimates and welfare measures have been examined within the recreation demand
literature, and meta-analyses within this literature (see Zandersen and Tol 2009, Shrestha and
Loomis 2003, Rosenberger and Loomis 2001, Walsh et al. 1992, and Smith and Kaoru 1990 a,b)
report that researcher decisions and modeling assumptions can substantively impact welfare
estimates. Table 1.1 provides a survey of the literature on the welfare impacts from various
assumptions (though this is not intended to be an exhaustive list). Numerous studies were
identified which provide information on the welfare impacts resulting from assumptions on
components of the travel cost variable, with many researchers reporting ranges of welfare
estimates resulting from using different percentages of an individual’s wage rate to calculate
their opportunity cost of time. For example, researchers commonly compare welfare estimates
obtained through valuing the opportunity cost of travel time at 0% of wage rate (i.e., there is no
time cost to the traveler), 100% of wage rate, and some fraction of the wage rate. In general,
researchers find that the percentage selected can have a substantial impact on welfare
estimates. When using 0% of the wage rate as the baseline estimate, valuing opportunity cost
of time at 100% of an individual’s wage rate results in welfare differences that are generally
greater than 100%, and as high as 490%. Also well-examined in the literature are the welfare
impacts associated with different approaches for estimating an individual’s transportation
costs. Decisions on whether to use individual-reported travel cost vs. imputed travel costs
(imputed costs generally incorporate per-mile estimates of driving costs obtained from
American Automobile Association), and which per-mile rate to apply to miles traveled when
using imputed travel costs also produce a wide range of welfare estimates.
16

Another area of research examining the impact of researcher-imposed best-judgments
and assumptions involves the treatment of recreational trips also consisting of incidental or
intended trips to another location. The specific issue is that for such trips, it is generally not
possible to apportion travel costs amongst different activities. As such, many studies (see
Englin et al. 2008, Bin et al. 2005, Chakraborty and Keith 2000, Fix and Loomis 1998, Bowker et
al. 1996, and Beal 1995) drop multiple-destination and/or multiple-purpose trips from their
analysis. However, Freeman (1993) notes that dropping such trips can result in an
underestimate of consumer surplus. As such, researchers use their best judgment in deciding
whether to drop trip observations that are suspected to involve incidental or intended trips to
another location. The decision whether to drop such trips can have a large impact on welfare
estimates.

1.3 Methods
1.3.1 Economic Theory
Random Utility Maximization (RUM) is the theoretical framework underlying the
recreation demand model used in this research. The utility an individual realizes from
recreating at a particular site (site i) is expressed mathematically in equation 1.
1)

𝑣𝑛𝑖 = 𝛽𝑡𝑐 𝑡𝑐𝑛𝑖 + 𝛽𝑞 𝑞𝑖 + 𝑒𝑛𝑖
The utility individual n receives from visiting site i is 𝑣𝑛𝑖 , 𝑡𝑐𝑛𝑖 is the trip cost for reaching

site i, 𝑞𝑖 is a vector of site characteristics, 𝑒𝑛𝑖 is a random error term capturing unmeasured

17

characteristics associated with site i, and 𝛽𝑠 are parameters (Haab and McConnell 2002;
Parsons 2003).
An individual chooses the option which offers the highest utility. Thus, site k would be
chosen from the set of all possible sites i when
2)

𝛽𝑡𝑐 𝑡𝑐𝑛𝑘 + 𝛽𝑞 𝑞𝑘 + 𝑒𝑛𝑘 ≥ 𝛽𝑡𝑐 𝑡𝑐𝑛𝑖 + 𝛽𝑞 𝑞𝑖 + 𝑒𝑛𝑖 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 𝑖𝑛 𝐶

where C is the set of possible recreation sites in an individual’s choice set.
A conditional logit model (which is derived from a random utility model with extreme
value type 1 - distributed random components) is used to estimate the probability individual n
chooses site k .
𝐶

3)

Prob(𝑛𝑘) = exp(𝛽𝑡𝑐 𝑡𝑐𝑛𝑘 + 𝛽𝑞 𝑞𝑘 / ∑ exp(𝛽𝑡𝑐 𝑡𝑐𝑛𝑖 + 𝛽𝑞 𝑞𝑖 )
𝑖=1

Parameter estimation proceeds via maximum likelihood estimation by estimating the log of the
likelihood function in equation 4, where Prob(𝑛𝑖) is the logit form from equation 3 and 𝑟𝑖𝑛 is
equal to one if site i is chosen by individual n, and equal to zero otherwise (Parsons 2003).
𝑁

4)

𝐶

𝐿 = ∏ ∏ Prob(𝑛𝑖)𝑟𝑖𝑛
𝑛=1 𝑖=1

The analysis focuses on examining how economic benefits estimates differ when proxies for
trips and income are used in place of trips and income obtained from survey responses. To this

18

end, equation 5 is used to estimate the change in per-trip economic benefits ( Sˆn ) resulting
from a quality change at one or multiple sites (𝑞́𝑖 ) , and with estimated parameters (𝛽̂ ).
𝐶

5)

𝐶

𝑆̂𝑛 = [ln{∑ exp(𝛽̂𝑡𝑐 𝑡𝑐𝑛𝑖 + 𝛽̂𝑞 𝑞𝑖́)} − ln{∑ exp(𝛽̂𝑡𝑐 𝑡𝑐𝑛𝑖 + 𝛽̂𝑞 𝑞𝑖 )}]/−𝛽̂𝑡𝑐
𝑖=1

𝑖=1

1.3.2 Empirical Model
Our empirical modeling strategy is designed to explore the potential for using
researcher-based best judgments to overcome data limitations when estimating recreation
demand models. To accomplish this, economic benefits estimates derived from travel cost
models which utilize hunter-reported data on income and trips (from a 2002 survey of Michigan
deer hunters) are contrasted with economic benefits derived from models re-estimated using a
variety of proxies for trips and income. The per-trip and seasonal economic benefits from the
existence of publicly accessible hunting land (relative to all publicly accessible land being
eliminated) are used as a basis for comparing the baseline models to the re-estimated proxybased models. Sections 1.3.3 – 1.3.5 proceed as follows. First, I describe the formulation of
two “Baseline” travel cost models which incorporate hunter-reported trips and hunter-reported
income in models which predict the selection of a hunting site as a function of a travel cost
variable and other hunting site attributes. Then, I describe the approach for constructing
proxies for income and trips, which allows for model re-estimation and subsequent economic
benefits comparison.

19

1.3.3 Baseline Travel Cost Models - Archery Hunting and Firearm Hunting
This analysis uses data from a 2003 survey of Michigan deer hunters. This survey data
contains information on the number of hunting trips taken and the number of days spent
hunting at each hunting location during the 2002 hunting season for two methods of deer
hunting: shotgun/rifle and archery. Separate baseline travel cost models for the shotgun/rifle
(henceforth firearm) and archery seasons are estimated using hunter-reported trips to a
Michigan county to construct the site choice variable and hunter-reported income to construct
the relevant component of the travel cost variable. Estimating separate travel cost models for
different types of hunting activities helps identify the extent to which differences in economic
benefits resulting from the use of proxy variables for income and trips may be contingent upon
the nature of the recreational experience.
Figure 1.1. Michigan hunting site choices

20

In the baseline archery and firearm travel cost models, the probability of site choice is
estimated as a function of the following hunting site attributes: travel cost, acreage of the three
types of publicly accessible hunting land, deer population, and two regional categorical
variables indicating whether the county is located in the Upper Peninsula or Northern Lower
Peninsula of Michigan (see Figure 1.1 for the delineation between these regions). The three
types of publicly accessible hunting land are federally-owned land, state-owned land, and
privately-owned, publicly accessible Commercial Forest Act (CFA) land. Federally-owned land in
Michigan consists of National Forest land, U.S. Fish and Wildlife Service land, and National
Lakeshore land, and State-owned land consists of State Forest land, State Parks and Recreation
Areas, and State Game and Wildlife Areas. The CFA provides tax incentives to private
landowners who retain and manage forest for long-term timber production; enrollees are
required to make this land available for hunters and anglers.
The variable Price is calculated as the sum of round-trip driving costs and the
opportunity cost of time associated with traveling to and from a hunting site. The round-trip
driving costs were obtained by multiplying the per-mile cost of operating a vehicle (31 cents;
AAA 2002) by the total estimated round-trip mileage. Round-trip mileage is estimated by
calculating the driving distance between the hunter’s residence and the centroid of the county
to which the trip was taken. The opportunity cost of time was estimated to be 1/3 of the
individual’s hourly wage rate (Parsons 2003), multiplied by the estimated number of hours the
individual spent traveling to and from the hunting site (estimated via PC Miler® software). For
the baseline firearm and archery models, the individual’s reported income was used to

21

estimate the opportunity cost of time. Deer population in the county was obtained from
estimates via the Michigan Department of Natural Resources.

1.3.4 Model Re-Estimation Using Proxy for Income
As noted in the review of the literature, some researchers confronted with missing data
for individual income assign individuals aggregate U.S. Census level income estimates based on
geopolitical regions such as zip code of residence. To examine the welfare implications of such
income proxies, a proxy for individual income is constructed by assigning each individual the
2000 U.S. Census estimate of median household income for their zip code of residence.
Economic benefits estimates from the baseline archery and firearm models are then compared
to benefits estimates from re-estimated archery and firearm models which substitute the
census-based income proxy into the relevant component of the travel cost variable.

1.3.5 Model Re-Estimation Using Proxy for Trips
Travel cost model estimation is problematic when site choice information is measured
by the number of days spent at a recreation site as opposed to the number of trips taken to the
recreation site. However, the number of trips taken to a recreation site might reasonably be
assumed to be a function of non-missing data such as the number of days spent at the
recreation site and the distance from the individual’s residence to the recreation site. I
hypothesize that individuals hunting in areas further from their residence would spend more
22

days hunting at a recreation site per trip, due to the higher costs associated with traveling to
these further sites (see McConnell 1992 for a similar argument). Prior to constructing proxy
variables for trips, I empirically examine the relationship between the number of days spent
hunting per hunting trip and the distance from their residence to the hunting location.

23

Figure 1.2. Average number of days spent hunting per firearm and archery (bow) hunting trip by distance from residence
4

Mean days per Trip

3

2
Firearm
Bow

1

0

n=994 n=488

n=204 n=110

n=111 n=52

n=82 n=41

< 50

≥ 50 & < 100

≥ 100 & < 150

≥ 150 & < 200

n=67 n=38

≥ 200 & < 250

Distance to Hunting Site From Residence (in miles)

24

n=93 n=40

≥ 250

Figure 1.2 provides evidence of an increasing relationship between the number of days
the individual spends hunting per trip and the distance of the hunting site from the individual’s
residence for both archery and firearm hunters. Archery and firearm hunters spend
approximately one day hunting per hunting trip for hunting sites under 50 miles from their
residence, whereas for long distance trips (>250 miles), archery hunters spend close to three
days hunting per hunting trip and firearm hunters report spending close to four days hunting
per hunting trip.
This assumption of increasing days per hunting trips as a function of distance is tested
through the use of Ordinary Least Squares (OLS) regression for both firearm and archery
hunters (specified below).

6)

𝑇𝑟𝑖𝑝𝑠𝑖𝑗 = 𝛼0 + 𝛽1 𝐷𝑎𝑦𝑠𝑖𝑗 + 𝛽2 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗 + 𝜀𝑖𝑗 , for all 𝑇𝑟𝑖𝑝𝑠𝑖𝑗 ≥ 1

In equation 6, 𝑇𝑟𝑖𝑝𝑠𝑖𝑗 is the number of trips taken by individual i to hunting site j during the
2002 hunting season, 𝐷𝑎𝑦𝑠𝑖𝑗 is the number of days spent hunting site j during the 2002 season,
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗 is the (one-way) distance in miles from the individual i’s residence to hunting site j,

and 𝜀𝑖𝑗 is an error term.

25

Table 1.2. OLS regression of the number of hunting trips taken to the site during the season on
the number of days taken to the site during the season and the distance to the hunting site
from the individual’s residence
Firearm Model
Independent
Variables

Archery Model

Coefficient

P-Value

Coefficient

P-Value

Days

0.323

P<0.01

0.941

P<0.01

Distance

-0.021

P<0.01

-0.026

P<0.01

Constant

5.252

P<0.01

2.183

P<0.01

Model Statistics
# of Obs.

1551 hunters

769 hunters

R-squared

0.269

0.718

OLS regression results in Table 1.2 are consistent with Figure 1.2 – individuals
economize on costs when hunting far-away sites by spending more days hunting per hunting
trip. This trend is evident in both the Firearm and Archery OLS models, with the coefficient on
Distance being negative and statistically significant (P<0.01) for both of these models. In
summary, the greater the distance traveled from an individual’s residence to a hunting site,
holding days spent at a recreation site constant, the fewer trips taken to that hunting site.
In the analysis supported by Figure 1.2 and Table 1.2 above, I avoided a more in-depth
analysis of the relationship between trips, days, and distance – desiring simply to confirm the
hypothesized direction of the relationship. Armed with the general information presented

26

above, seven ad-hoc relationships (see Table 1.3) were constructed which relate the number of
hunting trips taken to a site to the number of hunting days reported by the individual and the
distance required to travel to that site. Then, using the same interval scheme as with the adhoc calculations, seven empirically-based proxies for the number of trips were constructed by
calculating the mean trips-per-day ratio for a given distance interval level (obtained through the
survey data) and multiplying this ratio by the number of days reported hunting in the survey.
These empirically-based trip proxies will be used to serve as a basis of comparison for the adhoc assumptions. As these proxies incorporate actual information on hunter-reported
trips/days ratio (rather than relying on ad-hoc formulations of this relationship), economic
benefits estimates from these empirically-based proxies are expected to more closely
approximate benefits estimates from the baseline models.

27

Table 1.3. Ad-hoc trips proxy and empirically-based trips proxy variables
Site Choice
Proxy Variable
7)

8)

9)

10)

Ad-hoc Trips Proxy

Empirically-based Trips Proxy

Trips_A

𝐷𝑎𝑦𝑠, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 50
= {𝐷𝑎𝑦𝑠
, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 50
2

∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 50] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 50
∑ 𝐷𝑎𝑦𝑠
=
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 50] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 50
∑ 𝐷𝑎𝑦𝑠
{

Trips_B

𝐷𝑎𝑦𝑠, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
= {𝐷𝑎𝑦𝑠
, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 100
2

∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
∑ 𝐷𝑎𝑦𝑠
=
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 100] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 100
∑ 𝐷𝑎𝑦𝑠
{

Trips_C

𝐷𝑎𝑦𝑠, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
= {𝐷𝑎𝑦𝑠
, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150
2

∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
∑ 𝐷𝑎𝑦𝑠
=
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150
∑ 𝐷𝑎𝑦𝑠
{

𝐷𝑎𝑦𝑠, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75
𝐷𝑎𝑦𝑠
, 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
=
2
𝐷𝑎𝑦𝑠
, 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150
{
3

∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75
∑ 𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
= 𝐷𝑎𝑦𝑠 ∗ [
|75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150] , 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
∑ 𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 150
∑ 𝐷𝑎𝑦𝑠
{

Trips_D

28

Table 1.3. (cont’d)
Site Choice
Proxy Variable

11) Trips_E

Ad-hoc Trips Proxy

𝐷𝑎𝑦𝑠,
𝐷𝑎𝑦𝑠
, 100 ≤
=
2
𝐷𝑎𝑦𝑠
{ 3 ,

Empirically-based Trips Proxy
𝐷𝑎𝑦𝑠 ∗ [

𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200

= 𝐷𝑎𝑦𝑠 ∗ [

𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 200
{

∑ 𝑇𝑟𝑖𝑝𝑠
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
∑ 𝐷𝑎𝑦𝑠

∑ 𝑇𝑟𝑖𝑝𝑠
|100 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200] , 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200
∑ 𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 200] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 200
∑ 𝐷𝑎𝑦𝑠

∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75
∑ 𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150] , 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
∑ 𝐷𝑎𝑦𝑠
=
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|150 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 225] , 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 225
∑ 𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 225] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 225
∑ 𝐷𝑎𝑦𝑠
{

12) Trips_F

𝐷𝑎𝑦𝑠,
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 75
𝐷𝑎𝑦𝑠
, 75 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 150
2
= 𝐷𝑎𝑦𝑠
, 150 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 225
3
𝐷𝑎𝑦𝑠
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 225
{ 4 ,

13) Trips_G

∑ 𝑇𝑟𝑖𝑝𝑠
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
𝐷𝑎𝑦𝑠,
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 100
∑ 𝐷𝑎𝑦𝑠
𝐷𝑎𝑦𝑠
∑ 𝑇𝑟𝑖𝑝𝑠
, 100 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200
𝐷𝑎𝑦𝑠 ∗ [
|100 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200] , 100 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200
2
∑ 𝐷𝑎𝑦𝑠
𝐷𝑎𝑦𝑠
=
=
∑ 𝑇𝑟𝑖𝑝𝑠
, 200 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 300
3
𝐷𝑎𝑦𝑠 ∗ [
|200 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 200] , 200 ≤ 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 < 300
∑ 𝐷𝑎𝑦𝑠
𝐷𝑎𝑦𝑠
𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 300
∑ 𝑇𝑟𝑖𝑝𝑠
{ 4 ,
𝐷𝑎𝑦𝑠 ∗ [
|𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 300] , 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ≥ 300
∑ 𝐷𝑎𝑦𝑠
{
𝐷𝑎𝑦𝑠 ∗ [

29

Equations 7 through 13 (Trips_A through Trips_G) generate ad-hoc and empiricallybased proxy estimates for individual trips to a hunting site using days spent hunting and
distance to the hunting site. For example, the Ad-Hoc Trips Proxy for equation 9 (Trips_C) is
calculated using the assumption that each day spent hunting less than 50 miles from the
individuals residence resulted in one trip, whereas for days spent hunting at a location greater
than or equal to 50 miles from the hunters residence, two days spent hunting resulted in one
trip being taken. The Empirically-based Trips Proxy is calculated by first calculating the mean
trips – days ratio for each distance interval, then multiplying this ratio by the number of days
spent hunting by the individual within the distance interval. For example, the Empirically-based
Trips Proxy for equation 9 (Trips_C) is calculated by first estimating the mean trips/days ratio
for all hunters taking a trip to a county within 50 miles of their residence, with this ratio then
multiplied by the number of days reported hunting by each individual taking a trip within this
distance interval. Standard rounding procedures (i.e., ≥ 0.5 rounds up to nearest integer, < 0.5
rounds down to nearest integer) were used. When an individual reported hunting one day
(regardless of distance), that was considered to be a single trip.

1.4 Results
Table 1.4 provides the baseline model estimates for both the firearm and archery travel
cost models. Coefficient signs are generally as expected – Price is negative and Deer (deer
population) is positive for both models, and these coefficients are statistically significant

30

(P<0.001). The three types of publicly accessible hunting land (Federal, State, and CFA) are
positive and statistically significant (P<0.001) in each model.
Table 1.4. Baseline Firearm and Archery model results using hunter-reported information on
trips and income
Firearm
Archery
Baseline
Baseline
Model
Model
Independent Variables
Price

-0.0324*

-0.0376*

Deer(10,000 deer)

0.2217*

0.1502*

Federal (10,000 acres)

0.0352*

0.0598*

State(10,000 acres)

0.0779*

0.0845*

CFA(10,000 acres)

0.1046*

0.0898*

Size(100 sq. miles)

-0.1396*

-0.1019*

UP_Dum

2.4177*

1.6627*

NLP_Dum

0.5778*

0.5200*

Log-Likelihood

-30,850

-27,245

Observations

654

1355

Model Statistics

Notes: A * indicates statistical significance at the P<0.001 level. Null hypothesis
that all parameters equaled 0 was rejected for both models (P<0.0001)

Using economic benefits from the existence of publicly accessible hunting land as the basis of
comparison, I compare the economic benefits from the firearm and archery baseline models to
re-estimated firearm and archery models in which hunter-reported income is replaced by the
zip code level U.S. Census estimate of median household income. Welfare estimates from
31

these re-estimated models closely approximate the welfare estimates derived from baseline
firearm and archery models in which hunter-reported income are used (see Table 1.5). Per-trip
and seasonal estimates of the economic benefits of publicly accessible hunting land for both
firearm and archery hunters in Michigan varies by about 3 percent depending on whether
individual-reported income is used (as in the baseline models) or median zip code income is
substituted for individual-reported income (as in the re-estimated models).
Table 1.5. Differences in per-trip and seasonal welfare estimates for publicly accessible hunting
land between baseline models estimated using hunter reported income and models reestimated using aggregate census data for income
Re-estimated Models
Baseline
Using Census-based
% difference
Models
Median Income Proxy
Per-Trip

Seasonal

Firearm

$15.18

$14.74

-2.9%

Archery

$12.88

$12.43

-3.5%

$81,226,000

$78,717,000

-3.1%

Total

Next, I compare the welfare estimates using the empirically-based and ad-hoc trip
proxies with welfare estimates obtained via the baseline models. Per-trip and seasonal welfare
estimates for publicly accessible hunting land are obtained by re-estimating the firearm and
archery models using 14 different proxies for trips for each model (seven ad-hoc proxies and
seven empirically-based proxies) and then estimating the economic benefits of publicly
accessible hunting land. The 14 model re-estimations using ad-hoc trip proxies and the 14
model re-estimations using the empirically-based proxies all generated coefficient estimates

32

with the same sign as the baseline models using individual-reported information for income
and trip, and all results were statistically significant (P<0.001).1
For the firearm and archery models re-estimated using the ad-hoc and empirically-based
proxies for hunter trips, Table 1.6 shows that as the number of distance intervals increases,
welfare estimates for these trip proxies more closely approximate results from the baseline
models. Welfare estimates using the empirically-based proxies for trips differ from baseline
estimates by 1% - 31%, generally more closely approximating baseline estimates than the adhoc estimates (which has differences ranging from 3% - 44%). This is expected, as the
empirically-based trip proxies utilize estimates of the trips-days ratio calculated from survey
responses, whereas the ad-hoc proxies do not use such information. At the three and four
distance interval level, welfare measures obtained when using ad-hoc and empirically-based
proxies more closely align with welfare estimates from the baseline model. The four ad-hoc
assumptions for both the firearm and archery models produce welfare estimates that are
within 10% of the baseline models welfare estimates.

1

See Appendix A, Table A.1. and Table A.2. for coefficient estimates and model statistics for the
28 model estimations.
33

Table 1.6. Per-trip and economic benefits for firearm hunters and archery hunters for empirically-based and ad-hoc proxies for the
number of hunting trips
Baseline
Model

Per-Trip Firearm
Hunter Economic
Benefits from
Public Land

Per-Trip Archery
Hunter Economic
Benefits from
Public Land

Total Seasonal
Economic
Benefits for
Firearm and
Archery Hunters
From Public Land
(millions)

Two Intervals

Trip Proxies

Three Intervals

Four Intervals

Trips_A

Trips_B

Trips_C

Trips_D

Trips_E

Trips_F

Trips_G

Empiricallybased

$17.53
(15.48%)

$15.01
(-1.12%)

$19.91
(31.16%)

$15.57
(2.57%)

$15.13
(-0.33%)

$15.47
(1.91%)

$14.91
(-1.78%)

Ad-hoc

$20.07
(32.21%)

$19.55
(28.79%)

$21.86
(44.01%)

$15.38
(1.32%)

$16.57
(9.16%)

$14.55
(-4.15%)

$16.04
(5.66%)

Empiricallybased

$15.55
(20.73%)

$14.29
(10.95%)

$15.35
(19.18%)

$14.00
(8.70%)

$14.13
(9.70%)

$14.15
(9.86%)

$14.07
(9.24%)

Ad-hoc

$15.65
(21.51%)

$15.56
(20.81%)

$17.14
(33.07%)

$13.04
(1.24%)

$13.92
(8.07%)

$12.65
(-1.79%)

$13.86
(7.61%)

Empiricallybased

$95.2
(17.16%)

$83.5
(2.74%)

$103.4
(27.33%)

$84.9
(4.53%)

$83.6
(2.88%)

$84.8
(4.45%)

$82.6
(1.75%)

Ad-hoc

$104.6
(28.79%)

$102.5
(26.23%)

$114.1
(40.51%)

$82.3
(1.29%)

$88.4
(8.81%)

$78.5
(-3.39%)

$86.3
(6.29%)

$15.18

$12.88

$81.2

34

1.5 Discussion
This research examines the welfare impacts associated with operationalizing imperfect
data sets for the purposes of estimating travel cost models of recreation demand. I show that
employing assumptions on hunter trips and income when faced with missing data produces
welfare estimates that closely approximate welfare measures produced by travel cost models
that use individual-reported information on income and trips. Additionally, substituting censusbased income measures for individual-reported income impacts welfare measures by about 34%. Relative to the welfare implications of researcher assumptions and best judgments
detailed in Table 1.1, using a census-based income proxy in place of individual-reported income
has little effect. When assessing whether it is appropriate to assign an individual a censusbased median income proxy based on their geographic region of residence, it may be helpful to
search for and review information that compares socio-demographics of the group of interest
to the general population. Research by the National Shooting Sports Foundation (2010) found
that median hunter income was within 5% of state median income for 12 of 17 states
examined, further supporting the use of aggregate income measures when valuing hunting
experiences in different states. Given that survey item non-response to income questions is
typically between 20% and 40% (Moore et al. 1997, Juster and Smith 1997, and Yan et al. 2005),
the availability of aggregate census data on individual income may allow researchers to utilize a
large proportion of survey responses in which individuals choose not to answer income-related
questions.
The ad-hoc assumptions regarding the relationship between the number of days spent
hunting per trip and the distance of the site from an individual’s residence produced welfare
35

estimates that were increasingly close to baseline model estimates as the number of distance
intervals increased. The large positive differences of hunter welfare for the two interval
scenarios, relative to estimates from the baseline model, is likely due to these ad-hoc
assumptions failing to account for the propensity of individuals hunting far from their residence
to spend longer than two days hunting during a trip. As with the proxy for income, using ad-hoc
proxies for trips, particularly at the three or four interval level, resulted in welfare differences
from the baseline model which are much smaller than the differences identified in Table 1.1.
Ultimately, when assessing the feasibility of using secondary data measuring site choice in
terms of days rather than trips, researchers should keep in mind that other commonly used
researcher assumptions generally have a much larger impact on welfare measures.
The applicability of these findings to other hunting activities (and potentially to other types
of outdoor recreation) depends on whether individuals participating in other types of hunting
activities similarly adjust the number of hunting days per trip based on the distance to a
hunting site. Comparing welfare estimates between the firearm and archery models may
provide insight into whether this is the case, as there are potentially important differences in
the nature of the archery and firearm deer hunting experiences. An obvious difference is that
firearm and archery hunters use different weapons, which may require different levels of
sophistication and practice to attain a skill level sufficient to harvest an animal. Miller and
Graefe (2000) compared the degree of hunter specialization across seven different hunting
experiences (rifle deer, muzzleloader deer, archery deer, waterfowl, pheasant, grouse and
turkey) finding that the activity with the highest degree of specialization was archery deer
hunting, twice the mean degree of specialization of rifle deer hunting, which had the lowest
36

mean specialization of the seven hunting activities examined. Furthermore, Miller and Graefe
(2001) found dissimilarities in satisfaction models across rifle, archery, and muzzleloader deer
hunters, which the researchers interpreted as underlying difference in orientation towards
hunting between hunters in these three groups. Finally, the nature of recruitment into archery
hunting suggests that archery deer hunters are a distinctive subset of firearm deer hunters who
are seeking additional and/or different hunting experiences. According to Duda and Bisell
(2001), 82 percent of archery hunters actively hunted with firearms prior to taking up bow
hunting, with bow hunters having on average 12 years of experience in firearm hunting before
taking up bow hunting.
These differences in firearm and archery hunters and their respective hunting experiences,
along with the similarity of welfare measures between the archery and firearm models,
suggests that tradeoffs between trips and days as a function of distance may be similar across
different types of outdoor recreation experiences. It is important to note that different
assumptions on the cost and nature of recreational trips (as outlined in Table 1.1) have
considerably more impact on welfare estimates then do the assumptions on trips and individual
income. The wealth of wildlife harvest data available from the federal government and state
wildlife management agencies provides the opportunity to use similar assumptions to those
used within this dissertation chapter to estimate travel cost models for recreational hunting
experiences in different regions and for different species. Furthermore, the finding that
individuals facing high costs of reaching a site will take fewer trips while spending more days on
site per-trip might reasonably applied to other types of outdoor recreational experiences as
well. While care must be taken in each situation, this research supports the use of assumptions
37

on income and trips similar to those used in this paper to estimate travel cost models using
widely available but imperfect secondary data on recreation site choice.

38

APPENDIX

39

Table 1. Deer hunting travel cost model results using seven empirically-based estimates of trips as a function of days and distance
TRIPS_A
TRIPS_B
TRIPS_C
TRIPS_D
TRIPS_E
TRIPS_F
TRIPS_G
Vars Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery
Price -0.028 -0.034* -0.030 -0.035 -0.026 -0.035 -0.030 -0.037 -0.30 -0.036 -0.030 -0.037 -0.030 -0.036
Pop 0.223 0.173* 0.234 0.187 0.255 0.208 0.237 0.182 0.236 0.187 0.236 0.182
0.233 0.185
Size -0.154 -0.131* -0.136 -0.125 -0.169 -0.151 -0.152 -0.134 -0.142 -0.129 --0.152 -0.133 -0.139 -0.127
CFA 0.106 0.110* 0.102 0.104 0.109 0.103 0.107 0.104 0.102 0.101 0.106 0.105
0.096 0.096
NAT 0.041 0.067* 0.041 0.065 0.044 0.068 0.044 0.067 0.041 0.066 0.043 0.067
0.040 0.066
State 0.065 0.079* 0.059 0.078 0.071 0.088 0.064 0.086 0.062 0.081 0.064 0.085
0.066 0.083
UP
2.746 1.701* 2.792 1.763 2.506 1.665 2.794 1.747 2.787 1.691 2.790 1.773
2.619 1.623
NLP 0.653 0.624* 0.736 0.692 0.802 0.706 0.695 0.674 0.768 0.709 0.699 0.664
0.795 0.714
Stats
Log L -27,987 -32,483 -26,546 -31,585 -32,778 -32,260 -26,480 -30,425 -26,507 -31,174 -26,432 -30,609 -26,395 -31,105
Obs
654
654
654
654
654
654
654
654
654
654
654
654
654
654
Table 2. Deer hunting travel cost model results using seven assumption-based estimates of trips as a function of days and distance
TRIPS_A
TRIPS_B
TRIPS_C
TRIPS_D
TRIPS_E
TRIPS_F
TRIPS_G
Vars Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery Firearm Archery
Price -0.025 -0.033 -0.025 -0.033 -0.024 -0.031 -0.031 -0.040 -0.031 -0.037 -0.033 -0.041 -0.036 -0.045
Pop 0.227 0.175 0.236 0.193 0.252 0.208 0.237 0.178 0.237 0.189 0.237 0.177
0.252 0.212
Size -0.155 -0.131 -0.145 -0.128 -0.108 -0.147 -0.152 -0.128 -0.152 -0.129 -0.154 -0.129 -0.198 -0.175
CFA 0.105 0.111 0.104 0.106 0.108 0.107 0.106 0.097 0.106 0.094 0.104 0.098
0.117 0.106
NAT 0.041 0.068 0.041 0.065 0.043 0.064 0.043 0.068 0.043 0.0662 0.045 0.068
0.048 0.070
State 0.062 0.079 0.059 0.077 0.068 0.083 0.065 0.089 0.065 0.086 0.070 0.089
0.085 0.108
UP
2.609 1.744 2.587 1.751 2.455 1.622 2.773 1.670 2.773 1.530 2.697 1.640
3.230 2.018
NLP 0.667 0.615 0.738 0.677 0.781 0.683 0.710 0.675 0.710 0.722 0.743 0.680
0.995 0.928
Stats
Log L -31,978 -29,414 -33,710 -31,474 -35,938 -33,594 -27,973 -26,690 -29,973 -28,937 -26,759 -26,117 -28,301 -28,406
Obs
654
654
654
654
654
654
654
654
654
654
654
654
654
654

40

REFERENCES

41

REFERENCES

Beal, D.J. 1995. A Travel Cost Analysis of the Value of Carnarvon Gorge National Park for
Recreational Use. Review of Marketing and Agricultural Economics. 63.2: 292-303.
Bin, O., C.E. Landry, C.L. Ellis, and H. Vogelsong. 2005. Some Consumer Surplus Estimates for
North Carolina Beaches. Marine Resource Economics 20: 145-161.
Bishop, R.C., and T.A. Heberlein. 1979. Measuring Values of Extramarket Goods: Are Indirect
Measures Biased? American. Journal of Agricultural. Economics 61.5: 926-930.
Bowker, J.M., D.B.K. English, and J.A. Donovan. 1996. Toward a Value for Guided Rafting on
Southern Rivers. Journal of Agricultural and Applied Economics 28.2: 423-432.
Cameron, T. A. 1992. Combining Contingent Valuation and Travel Cost Data for the Valuation of
Nonmarket Goods. Land Economics 68.3: 302-317.
Cesario, F. J. 1976. Value of Time in Recreation Benefit Studies. Land Economics: 52.1: 32-41.
Chakraborty, L. and J.E. Keith. 2000. Estimating the Recreation Demand and Economic Value of
Mountain Biking in Moab, Utah: An Application of Count Data Models. Journal of
Environmental Planning and Management. 43.4: 461-469.
Coyne, A.G., and W.L. Adamowicz. 1992. Modeling Choice of Site for Hunting Bighorn Sheep.
Wildlife Society Bulletin 20.1: 26-33.
Duda, M.D., and S.J. Bissell. 2001. Bowhunting participation, trends, satisfactions and marketing
options. Harrisonburg, VA: Responsive Management.
Edwards, P.E.T., G.R. Parsons, and K.H. Myers. 2011. The Economic Value of Viewing Migratory
Shorebirds on the Delaware Bay: An Application of the Single Site Travel Cost Model
Using On-Site Data. Human Dimensions of Wildlife 16.6: 435-444.
Englin, J., T.P. Holmes, and J. Lutz. 2008. Wildfire and the Economic Value of Wilderness
Recreation. In The Economics of Forest Disturbances: Wildfires, Storms and Invasive
Species., eds. T. P. Holmes, J. P. Prestemon, and K. L. Abt, pp. 191-208. Forestry
Sciences.
English, D.B.K., and J.M. Bowker. 1996. Sensitivity of Whitewater Rafting Consumers Surplus to
Pecuniary Travel Cost Specifications. Journal of Environmental Management. 47.1: 7991.

42

Fix, P., and J.B. Loomis. 1998. Comparing the Economic Value of Mountain Biking Estimated
using Stated and Revealed Preference. Journal of Environmental Planning and
Management. 41.2: 227-236.
Freeman, A.M. 1993. The Measurement of Environment and Resource Values: Theory and
Methods. Washington, D.C.: Resources for the Future.
Haab T.C., and K.E. McConnell. 2002. Valuing Environmental and Natural Resources:
Econometrics of Non-Market Valuation. Northhampton, MA: Edward Elgar Publishing.
Heberling, M.T., Templeton, J.J. 2009. Estimating the economic value of national parks with
count data models using on-site, secondary data: The case of the great Sand Dunes
National Park and Preserve. Environmental Management. 43: 619–627.
Henderson, M.M., K.R. Criddle, and S.T. Lee. 1999. The Economic Value of Alaska’s Copper River
Personal Use and Subsistence Fisheries. Alaska Fishery Research Bulletin 6.2: 63-69.
Hynes, S., N. Hanley and C.O’Donoghue. 2009. Alternative Treatments of the Cost of Time in
Recreational Demand Models: An Application to Whitewater Kayaking in Ireland.
Journal of Environmental Management. 90.2: 1014-1021.
Jakus, P.M., M. Downing, M.S. Bevelimer, J.M. Fly. 1997. Do Sportfish Consumption Advisories
Affect Reservoir Anglers' Site Choice? Agricultural and Resource Economics Review. 26:
196-204.
Juster, F.T., and J.P. Smith. 1997. Improving the Quality of Economic Data: Lessons from the HRS
and AHEAD. Journal of the American Statistical Association 92.440: 1268-1278.
Knoche, S., Lupi, F., 2013. Economic Benefits of Publicly Accessible Land for Ruffed Grouse
Hunters. Journal of Wildlife Management. 77: 1294-1300.
Knoche, S., and F. Lupi. 2012. The Economic Value of Publicly Accessible Deer Hunting Land.
Wildlife Society Bulletin 76.3: 462-470.
Knoche, S., and F. Lupi. 2007. Valuing Deer Hunting Ecosystem Services from Farm Landscapes.
Ecological Economics 64.2: 313-320.
Layman, R.C., J.R. Boyce, K.R. Criddle. 1996. Economic Valuation of the Chinook Salmon Sport
Fishery of the Gulkana River, Alaska, under Current and Alternative Management Plans.
Land Economics 72.1: 113-128.
Loomis, J. B. “A Comparison of the Effect of Multiple Destination Trips on Recreation Benefits
as Estimated by Travel Cost and Contingent Valuation Methods.” Journal of Leisure
Research 38,1(2006): 46-60.

43

Loomis, J.B., S. Yorizane and D.M. Larson. 2000. Testing Significance of Multi-Destination and
Multi-Purpose Trip Effects in a Travel Cost Method Demand Model for Whale Watching
Trips. Agricultural and Resource Economics Review 29.2: 183–191.
McConnell, K.E. 1992. Onsite Time in the Demand for Recreation. American Journal of
Agricultural Economics 74: 918-25.
Mendelsohn, R., Hof, J., Peterson, G., and Johnson, R. 1992. Measuring Recreation Values with
Multiple Destination Trips. American Journal of Agricultural Economics 74.4: 926-933.
Miller, C.A. and A.R. Graefe. 2000. Degree and Range of Specialization Across Related Hunting
Activities. Leisure Sciences 22.3:195-204.
Miller, C.A. and A.R. Graefe. 2001. Effect of Harvest Success on Hunter Attitudes Towards
White-tailed Deer Management in Pennsylvania. Human Dimensions of Wildlife 6.3: 189203.
Moore, J.C., L.L. Stinson, and E.J. Welniak Jr. 1997. Income Measurement Error in Surveys: A
Review. Statistical Research Report, U.S. Bureau of the Census, Washington, D.C.
National Shooting Sports Foundation [NSSF]. 2010. A Portrait of Hunters and Hunting License
Trends: National Report. Newton, CT.
Parsons, G.R., and A.J. Wilson.1997. Incidental and Joint Consumption in Recreation Demand.
Agricultural and Resource Economics Review 26.1: 1-6.
Parsons, G.R. 2003. The Travel Cost Model. In A Primer on Nonmarket Valuation, eds.,
P.A.Champ, K.J. Boyle, and T.C. Brown, pp. 269-329. Dordrecht, The Netherlands:
Kluwer Academic Publishers.
Pittman-Robertson Wildlife Restoration Act of 1937, 16 U.S.C. 669-669k.
Prayaga, P., J. Rolfe, and N. Stoeckl. 2010. The Value of Recreational fishing in the Great
BarrierReef, Australia: A Pooled Revealed Preference and Contingent Behaviour Model.
Marine Policy 34.2: 244-251.
Rockel, M.L. and M.J. Kealy. 1991. The Value of Nonconsumptive Wildlife Recreation in the
United States. Land Economics 67.4: 422-434.
Rosenberger, R.S., and J.B. Loomis. 2001. Benefits Transfer of Outdoor Recreation Use Values.
General Technical Report RMRS-GTR-72, Rocky Mountain Research Station, U.S.D.A.
Forest Service, Fort Collins, CO.
Shrestha, R.K., and J.B. Loomis. 2003. Meta-analytic Benefit Transfer of Outdoor Recreation
Economic Values: Testing Out-of-Sample Convergent Validity. Environmental and
Natural Resource Economics 25.1: 79-100.
44

Smith, V.K., and Y. Kaoru. 1990a. What Have We Learned Since Hotelling’s Letter? A MetaAnalysis. Economics Letters 32: 267-272.
Smith, V.K., and Y. Kaoru. 1990b. Signals or Noise? Explaining the Variation in Recreation
Benefit Estimates. American Journal of Agricultural Economics 72: 419-433.
U.S. Department of the Interior [USDOI], Fish and Wildlife Service, and U.S. Department of
Commerce, U.S. Census Bureau. 2013. 2011 National Survey of Fishing, Hunting, and
Wildlife-associated Recreation. Washington, D.C., USA.
Walsh, R.G., D.M. Johnson, and J.R. McKean. 1992. Benefit Transfer of Outdoor Recreation
Demand Studies, 1968-1988. Water Resources Research 28:707-713.
Whitehead, J.C., Poulter, B., Dumas, C.F., Bin, O. 2009. Measuring the Economic Effects of Sea
Level Rise on Shore Fishing. Mitigation and Adaptation Strategies For Global Change 14:
777-792.
Wilman, E.A., and R.J. Pauls. 1987. Sensitivity of Consumers’ Surplus Estimates to Variation in
the Parameters of the Travel Cost Model. Canadian Journal of Agricultural Economics
35: 197-212.
Yan, T., M. Jans, and R. Curtin. 2006. Changes in Nonresponse to Income Questions.
Proceedings Joint Statistical Meetings/American Association for Public Opinion Research.
Montreal, Canada.
Zandersen, M., and R.S.J. Tol. 2009. A Meta-analysis of Forest Recreation Values in Europe.
Journal of Forest Economics 15: 109-130.
Zawacki, W. T., A. Marsinko, and J.M. Bowker. 2000. A Travel Cost Analysis of Nonconsumptive
Wildlife-Associated Recreation in the United States. Forest Science 46.4: 496-506.

45

CHAPTER 2: ESTIMATING THE ECONOMIC BENEFITS OF PUBLICLY ACCESSIBLE LAND FOR
RUFFED GROUSE HUNTERS2

2.1 Introduction
Hunting small game is a popular outdoor recreational activity in the United States, with
4.5 million people spending about 51 million days and $2.5 billion hunting rabbit, squirrel,
pheasant, grouse, and other small mammals and birds in 2011 (U.S. Department of the Interior
[USDOI] 2011). Monetary expenditures incurred by small game hunters on trips and equipment
have the potential to generate substantial economic impacts to local communities, with many
of these benefits accruing to rural economies of particular concern and interest to politicians
and policy makers. As a result, much attention has been paid to these retail expenditures and
resulting economic impacts, which include multiplier effects (initial spending and recirculation
of those expenditures throughout an economy) and the jobs generated by the initial spending
and re-spending of these dollars. From 2000–2003, retail sales related to small game hunting in
the United States averaged $443 million annually, generating annual multiplier effects of about
$1.2 billion and over 10,100 full and part-time jobs (Southwick Associates 2007).
Economic impacts such as those detailed above are widely available and can be useful
for characterizing the importance of outdoor recreational activities such as small game hunting

2

A version of this chapter was published as “Knoche, S., and F. Lupi. 2013. Economic Benefits of
Publicly Accessible Land for Ruffed Grouse Hunters. Journal of Wildlife Management 77: 1294-1300.”.
In that publication, welfare measures from the sample were extrapolated to the population by
computing per-trip willingness to pay measures, which were then averaged across sample members
and multiplied by the number of resident ruffed grouse hunters. Here, consistent with the other
dissertation chapters, welfare measures were extrapolated using weighted average per-trip willingness
to pay where the weights were each person’s trips, which was multiplied by the number of resident
ruffed grouse hunters.
46

to regional or local economies. However, economic impacts figures are not appropriate for
analyzing alternative management regimes within a cost-benefit context. For example, were
small game hunting not an option, individuals would spend their money on other activities and
these expenditures would still generate economic impact, just to different industries likely
affecting different regions and different people. Thus, such expenditures simply reflect
monetary transfers, and are not appropriate for use in informing the deployment of scarce
management resources to improve the welfare of outdoor recreationists. Estimating the
economic benefits (i.e., maximum willingness to pay minus total costs) accruing to outdoor
recreation participants from aspects of the recreational experience is the appropriate approach
for evaluating the net benefits of costly alternative management regimes. Estimating the
economic benefits associated with aspects of outdoor recreation requires the use of nonmarket valuation techniques, which have often been used to estimate economic benefits
associated with big game hunting (e.g., white-tailed deer (Odocoileus virginianus)[Schwabe et
al. 2001; Knoche and Lupi 2007, 2012], elk (Cervus Canadensis) [Fried et al. 1995], moose (Alces
alces) [Sarker and Surry 1998], and pronghorn antelope (Antilocapra americana) [Boxall 1995]).
However, other than ring-necked pheasant (Phasianus colchicus) hunting (see Shulstad and
Stoevener 1978, Adams et al. 1989, Remington et al. 1996, Hansen et al. 1999), little research
has been conducted on the economic benefits associated with the small game hunting
experience. In particular, the popularity of hunting ruffed grouse (Bonasa umbellus) on public
land relative to many other small game species (Frawley 2012) offers a unique opportunity for
economic valuation and policy analysis.

47

Wildlife managers and policy makers at both the federal and state levels have enacted
programs to increase and enhance the provision of publicly accessible lands for hunting and
other types of outdoor recreation. Helland (2006) reports that 11 states (primarily west of the
Mississippi) have Walk-In Hunting Access (WIHA) programs that facilitate access to private land.
According to Benson (2001), 40% of state wildlife management agencies have a program that
assists hunters and landowners with access-related issues for hunted and/or non-hunted
species, with 96% of agency administrators reporting that access to private land is important
for their organizations’ objectives. At the federal level, President Obama launched the
America’s Great Outdoors initiative in 2010, which recommends full funding of $900 million per
year (by 2014) for the Land and Water Conservation Fund (LWCF), a federal government
funding program which over the past 40 years has enabled the purchase of 1.2 million hectares
of recreation lands and the funding of over 29,000 projects to develop basic recreational
facilities on public lands. In 2012, the LWCF distributed more than $42 million in grants for land
acquisition and recreation development purposes, which subsequently leveraged an additional
$48 million from state, local, and private sources. Hunting-specific federal initiatives such as
the United States Department of Agriculture’s Public Access and Habitat Incentive Program
(VPA-HIP) provided grants to 25 states totaling $17.8 million to encourage private landowners
to make their land available to the public for hunting.
Despite efforts by state and federal governments to 1) improve or increase access to
publicly-owned recreation land, and 2) to implement programs that incentivize landowners to
allow public hunting access on their privately owned lands, few research efforts (other than
Knoche and Lupi [2012]) have examined the economic benefits accruing to hunters as a result
48

of changing amounts of publicly accessible land for hunters. To address these gaps in the
literature, the travel cost valuation method was used to determine which site attributes (e.g.,
harvest rate, public land acreage) influence a ruffed grouse hunter’s site selection decision.
Economic benefits to ruffed grouse hunters were then estimated for each type of publicly
accessible hunting land that is a statistically significant predictor of hunter site selection. This
method allowed the estimation of economic benefits of publicly accessible land to grouse
hunters by examining the tradeoffs individuals make between the amount and nature of
publicly accessible land at a hunting site and the cost incurred in traveling to that hunting site.

2.2 Study Area
The ruffed grouse is the most popular upland game bird in the upper Great Lakes region,
with about 193,000 hunters spending over 1.5 million days hunting ruffed grouse in Michigan
and Wisconsin in 2007, compared to 116,000 hunters spending about 800,000 days hunting
pheasant in these two states (Dhuey 2008, Frawley 2008). Ruffed grouse hunting participation
and harvest can vary substantially, as ruffed grouse in this region, as well as in Alaska and much
of Canada, have approximately 10-year cycles of abundance (Rusch et al. 1999), with hunter
numbers and ruffed grouse harvest fluctuating with respect to the population cycle (Dessecker
et al. 2006). During cyclic highs, about three million ruffed grouse are harvested evenly across
Minnesota, Wisconsin, and Michigan, with about 900,000 harvested during cyclic lows
(Dessecker and McCauley, 2001). Our survey uses data from the 2008 ruffed grouse hunting

49

season, during which ruffed grouse populations were rebounding from population lows
documented during 2004–2005 (Frawley and Stewart 2009).
In 2008 in the state of Michigan, an estimated 88,300 resident hunters spent 624,000
days harvesting 301,000 ruffed grouse (Frawley 2012). For the purposes of delineating ruffed
grouse population areas of interest in Michigan (see Figure 1.1 in dissertation chapter 1), the
Upper Peninsula and much of the Northern Lower Peninsula of Michigan (along with northern
portions of Wisconsin and Minnesota, and portions of southern Canada) are considered to be
part of the Boreal Transition Forest area, defined as a transitional forest landscape between
deciduous-dominated forests to the south and coniferous-dominated forests to the north
(Dessecker 2006). The remaining southern portion of the Lower Peninsula of Michigan, along
with the southern two-thirds of Wisconsin and Central Minnesota, are defined to be part of the
Prairie Hardwood Transition, a transitional landscape between prairies to the west and south
and deciduous-dominated forests to the east and north (Stewart et al. 2006). All forested
landscapes in these regions can support populations of ruffed grouse. However, mature forests
generally support only low-density populations of ruffed grouse, whereas ruffed grouse are
most prolific in young (early-successional) forests. The Michigan portion of the Boreal
Transition Forest area has a substantial amount of ruffed grouse habitat, with about 4.1 million
hectares of large-diameter forest (>12.7 cm) and 1.3 million hectares of small-diameter forest
(≤12.7 cm; Dessecker 2006). The southern portion of the Lower Peninsula of Michigan, which
has been heavily affected by humans via agricultural use and urban or suburban development,
has about one million hectares of large-diameter and 400,000 hectares of small-diameter forest
(Dessecker 2006).
50

Forests in both regions described above contain productive ruffed grouse habitat.
Additionally, much of this forest acreage is located on publicly accessible land, available to all
Michigan hunters for free access or for a small entrance fee. Publicly-owned hunting options in
Michigan include State Forest (1.5 million ha), National Forest (1.2 million ha), State Game and
Wildlife areas (200,000 ha), State Parks and Recreation areas (107,000 ha), 2 National
Lakeshores (52,000 ha), and United States Fish and Wildlife Service Land (47,000 ha).
Additionally, about 890,000 million ha of privately-owned, publicly accessible land is available
to hunters through Michigan’s Commercial Forest Act (CFA), which provides tax incentives for
private landowners (predominantly timber companies) to allow hunting and angling on their
land. Federally-owned land, state-owned land, and CFA land are the three major types of
publicly accessible hunting lands in Michigan, and serve as the focal point of this analysis (see
Knoche and Lupi 2012 for greater detail relating to publicly accessible hunting lands in
Michigan).

2.2.1 Policy Scenario
Currently, the vast majority of federally-owned and state-owned land in Michigan is
open to recreational hunting. However, national forests are legally responsible for providing
for a variety of uses and benefits by the Multiple Use and Sustained Yield Act (1960), which
states that “…it is the policy of Congress that the national forests are established and shall be
administered for outdoor recreation, range, timber, watershed, and wildlife and fish purposes.”
Managing land for multiple uses can result in stakeholder conflict when one stakeholder’s use
of the resource is affected by, or incompatible with, the use of the resource by another
51

stakeholder. Conflict between recreational users of national forest land (e.g., hikers,
birdwatchers, and snowshoers vs. gun hunters and snowmobilers) is at the heart of a recent
court ruling requiring the United States Forest Service to consider closing 14 semi-primitive
areas containing about 27,000 ha of the approximately 400,000 ha Huron-Manistee National
Forest to recreational gun hunting. The Huron-Manistee National Forest is popular with
outdoor recreationists (with about 4 million estimated visits during 2007), and recreational
hunting is one of the most popular recreational activities, with about 25% of Huron-Manistee
National Forest visits being for the primary purpose of hunting (U.S. Department of Agriculture
[USDA] 2009). Closing portions of this National Forest to gun hunting has the potential to
negatively affect hunters pursuing a multitude of game species found on these lands, including
white-tailed deer, black bear (Ursus americanus), wild turkey (Meleagris gallopavo), and ruffed
grouse. In a 27 January 2012 decision, the USDA Forest Service officially selected a
management alternative detailed within the Final Supplemental Environmental Impact
Statement (USDA 2012) that will continue to allow gun hunting at these 14 areas; however, this
decision has since been appealed and may not be settled for some time. This particular user
conflict scenario provides a useful setting for examining how economic benefits accruing to a
particular user group would be affected when access privileges are reduced or eliminated.

52

2.3 Methods
2.3.1 Data
To examine how the location of hunter trips are influenced by the different levels of
attributes at potential hunting sites, I first describe the various hunting site options within the
hunter’s choice set. For the purposes of this study, each of the 83 counties in Michigan is
defined as being a hunting site alternative (see figure 1.1 in dissertation chapter 1). Developing
a predictive model of hunter site selection and ultimately estimating the economic benefits to
hunters from hunting site attributes requires information on hunter site choice and trip
frequency to each site, and the levels of attributes available at each hunting site (e.g., harvest
rate of game species, amount of publicly accessible hunting land).
Information on hunter site choice and site visitation frequency is obtained from the
Michigan 2008–2009 Upland Game Harvest Report survey (Michigan Department of Natural
Resources 2009) administered by the Michigan Department of Natural Resources (MDNR). This
was a mail survey sent to 9,987 randomly selected people who were eligible to hunt small game
in Michigan, with up to two follow-up questionnaires sent to non-respondents. Questionnaires
were undeliverable to 242 people and were returned by 5,532 people, yielding a 57% response
rate. For this analysis, only Michigan residents who hunted ruffed grouse in Michigan (n =
1,280) are included. Information on the number of trips to a site is generally required to
estimate a travel cost model, however, the Upland Game Harvest Survey provides data on the
number of days spent hunting, as opposed to the number of trips. To convert the reported
number of days hunted into an estimate of hunting trips, I relied on relationships between

53

hunting trips, hunting days, and travel distance that were identified using data from a 2003
survey of Michigan deer hunters. Using this data to estimate multiple travel cost models in
which both reported trips and estimated trips are used as the dependent variable, I identified
several plausible formulations of this relationship that generated economic benefits estimates
that closely approximated benefits estimates obtained when using hunter-reported trips (see
dissertation chapter 1). The best formulation was used to convert grouse hunting days into
grouse hunting trips. Using this formulation converted 12,006 days hunting into 9,666 trips,
which is about 1.24 days per hunting trip.
The variable Price was computed to reflect the round-trip time costs and driving costs
from each hunter’s residence to each of the 83 different counties. For each survey respondent,
PC*Miler software was used to calculate the driving distance from the individual’s zip code of
residence to the zip code that is closest to the geographic center of each hunting site. Driving
cost was computed by multiplying the per-mile vehicle operating cost by the distance traveled.
The per-mile operating costs, excluding fixed costs such as insurance, were 40.2 cents per mile
based on data published annually by the American Automobile Association (2008). The other
component of travel costs was the opportunity cost of the time the individual spent to travel to
the site. For this, the individual’s address was used in conjunction with ArcGIS® software to
assign each individual the United States Census Bureau median household income for their
census tract of residence (Cameron 1992, Kahn 1998). Some individual’s addresses were not
identifiable; for these cases, individuals were assigned the median household income for their
zip code of residency. These Census-based median household income figures were used as
proxies for individuals’ annual wage rates. Using the same data and travel cost modeling
54

approach as described in the previous paragraph, I found that for two different types of
hunting, substituting median household income estimates (based on zip code of residence) for
individual-reported income had a negligible effect on economic benefits estimates (see
dissertation chapter 1). Though some debate continues regarding the appropriate portion of an
individuals’ wage rate to attribute to the time component of travel cost, the recreation
literature has generally accepted that the opportunity cost of time spent driving ranges from a
third of the hourly wage rate to the full wage rate (Parsons 2003). For this analysis, one-third of
the wage rate was used. Wage rate was computed by dividing the Census-based median
income figure by the number of work hours per year. Travel time was calculated using an
average trip speed of approximately 64 km (40 miles) per hour.
Data on seasonal ruffed grouse harvest data was obtained from the Michigan Upland
Game Harvest Report surveys implemented by the MDNR. In Michigan, the ruffed grouse
hunting season is approximately three months long, beginning 15 September and lasting
through 1 January, with the exception that ruffed grouse hunting is not permitted during the
firearm deer season (15–30 Nov). Using hunter survey responses from the 2006–2007, 2007–
2008, and 2008–2009 Michigan Upland Game Harvest Report surveys, a per-day ruffed grouse
harvest rate, averaged over three years, was produced for each county in Michigan.

2.3.2 Economic Model
The travel cost method is commonly used to estimate economic benefits associated
with outdoor recreation (Parsons 2003). Understanding the relationship between the cost

55

(price) of a recreational trip and the number of trips taken allows the construction of a demand
relationship between these two variables, which then can be used to estimate economic
benefits. Economic benefits, as defined in this paper, represent the difference between the
greatest amount an individual would be willing to pay to participate in the activity and the triprelated expenses actually incurred to participate. If an individual participates in the
recreational activity, then the activity must be worth at least as much as the cost incurred to
participate. The individual realizes an economic benefit when the amount the individual would
be willing to pay for the recreational experience exceeds the cost of the experience.
As my primary research interest was to estimate economic benefits associated with
publicly accessible hunting land, I used a multiple-site approach known as the random utility
travel cost model (Freeman 1993, Grijalva et al. 2002, Knoche and Lupi 2007). The random
utility travel cost model allows the estimation of the economic benefits of site attributes by
comparing the attribute levels to the costs of traveling to those hunting sites (Haab and
McConnell 2002, Lupi et al. 2003, Parsons 2003, Kotchen et al. 2006). The modeling of hunter
site choice, estimation of site attribute parameters, and calculation of economic benefits for
publicly accessible land within this paper closely follows the economic theory and econometric
methods described in detail in dissertation chapter 1, section 1.3.1.

2.4 Results
Despite the Upper Peninsula having only about 3% of Michigan’s residents, 27.4% of
Michigan ruffed grouse hunters live in the Upper Peninsula (Table 2.1). Likewise, about 13.5%
56

of Michigan residents reside in the Northern Lower Peninsula but 31.8% of ruffed grouse
hunters live in this region. The Southern Lower Peninsula has the majority of Michigan’s ruffed
grouse hunters but comprises a smaller share of the total population of their regions than do
the northern regions. Regional trends are also evident regarding hunting site selection. Upper
Peninsula residents are more likely to hunt in the county they live in and are more likely to hunt
in two or more counties, and they also drive the shortest distance to hunting sites. Southern
Lower Peninsula residents are least likely to hunt in their county of residence or hunt in two or
more counties, and they drive the furthest distance to hunting sites, whereas hunting site
selection figures of Northern Lower Peninsula residents lie between the other two regions.
Comparing Michigan ruffed grouse hunter demographics to Michigan hunters in general, we
see that the mean age of Michigan hunters (for all species) is 42, and percent male composition
of hunters is 92% (Frawley 2006), both of which are below the ruffed grouse hunter age and
gender statistics for all 3 regions in Michigan. However, care should be taken with these
demographic comparisons as Frawley’s figures are from the 2005 hunting seasons, and the
ruffed grouse figures for use in this analysis are from the 2008 hunting season.

57

Table 2.1. Michigan ruffed grouse hunter demographics and hunting site choices by region of
residence, 2008–2009
PUBLICLY
ACCESSIBLE
HUNTING LAND BY
REGION OF REGION (1000S HA)
RESIDENCE

HUNTER DEMOGRAPHICS AND HUNTING SITE CHOICES

%
%
Avg. distance
% living
hunting
Avg. %
hunting from residence
Federal State CFA in each
in county
Age male
in ≥2
to site hunted
region
of
counties
(km)
residence

Upper
Peninsula

724

699

887

27.4

47.3

94.9

87.2

29.3

52.5 km

Northern
Lower
Peninsula

888

989

17

31.8

49.0

96.6

64.6

27.8

108.6 km

Southern
Lower
Peninsula

6

155

2

40.8

48.3

98.3

12.5

21.6

307.4 km

Statewide

1,617

48.2 96.8

49.5

25.7

174.3 km

1,842 905

For the economic model of hunter site choices, the Price variable represents the implicit
price an individual hunter pays to hunt at a particular site. With most goods purchased in the
marketplace, as price of the good increases, holding all else constant, the quantity of the good
purchased decreases. The same holds true here; price had a negative effect on number of trips
taken by a hunter to a specific county (P<0.001), meaning that increases in travel cost to a
hunting site, all else equal, result in fewer hunting trips to that hunting site (Table 2.1). As
expected, the 3-year average of ruffed grouse harvested per day (Harvest) and amount of the 3
types of publicly accessible hunting land per county (CFA, Federal, and State) had a positive
effect on number of trips taken by a hunter to a specific county (P<0.001). The region-specific
58

dummy variables for the Upper Peninsula and the Northern Lower Peninsula both had a
positive effect on number of trips to a county (P<0.001). Because of the non-linearity of the
conditional logit model, the precise value of a single coefficient does not have a straightforward
interpretation. However, when comparing across variables measured by the same metric (e.g.,
hectares of the different types of publicly accessible hunting land), the relative magnitude of
these coefficients indicates the importance of a marginal increase in the level of a hunting site
attribute in generating economic benefits to hunters.

59

Table 2.2. Conditional logit results for a travel cost model of ruffed grouse hunting in Michigan
Variablea

Coefficient

SE

P-Values

Price

−0.029

0.0003

P<0.001

Harvest (an additional ruffed grouse)

1.373

0.086

P<0.001

Federal (10,000 ha)

0.072

0.0044

P<0.001

State (10,000 ha)

0.169

0.0066

P<0.001

CFA (10,000 ha)

0.148

0.0071

P<0.001

Size (100 sq. km)

−0.049

0.003

P<0.001

Upper Peninsula

4.502

0.098

P<0.001

Northern Lower Peninsula

2.475

0.057

P<0.001

Statistics
Log likelihood

−23,237

Observations

1,280 hunters

a

Price = travel costs; Harvest = 3-year average of grouse hunted per day;
Federal, State, and CFA = area of land in the county that is federally-owned
land, state-owned land, and publicly accessible, privately-owned Commercial
Forest Act, respectively; size = area of county; Upper Peninsula and Northern
Lower Peninsula represent region-specific dummy variables (reference region is
Southern Lower Peninsula).

Seasonal economic benefits for the three types of publicly accessible land individually in
this model are estimated to be about $14.0 million for state-owned land, $11.1 million for CFA
land, and $4.4 million for federally-owned land, with a total of $29.9 million in economic
benefits annually to ruffed grouse hunters from the existence of publicly accessible hunting
lands in Michigan. Average per hectare economic benefits for publicly accessible land are
greatest for CFA land ($12.29/ha), followed by state-owned land ($7.61/ha) and federallyowned land ($3.94/ha). I also examined the effects of different assumptions regarding the
60

percentage of wage rate and driving costs, as these assumptions will affect the economic
benefits estimates detailed above (e.g., English and Bowker 1996, Layman et al. 1996, Bin et al.
2005, Edwards et al. 2011). Relative to using 1/3 wage rate, economic benefits are reduced by
7% when using 1/4 of wage rate, whereas using 2/5 of wage rate increases economic benefits
by 6%. Similarly, decreasing the travel rate by approximately 16 km (10 miles) per hour reduces
economic benefits by 10%, whereas increasing the travel rate by approximately 16 km (10
miles) per hour increases economic benefits by 6%.
The vast majority of publicly accessible hunting land is in the Upper Peninsula and
Northern Lower Peninsula of Michigan, with public lands in these areas generating substantial
economic benefits for ruffed grouse hunters. The approximately 2.3 million ha of publicly
owned and privately-owned, publicly accessible hunting land in the Upper Peninsula of
Michigan generates about $24.1 million in annual economic benefits for ruffed grouse hunters,
and the approximately 1.4 million ha of publicly accessible hunting land in the Northern Lower
Peninsula generates about $4.8 million in annual economic benefits. Publicly accessible land in
the Southern Lower Peninsula generates about $160,000 in annual economic benefits. The
Upper Peninsula generates the greatest amount of average annual economic benefits per
hectare ($10.45/ha) followed by the Northern Lower Peninsula ($3.42/ha). Despite the relative
scarcity of publicly accessible hunting land in the Southern Lower Peninsula, combined with the
proximity of this public land to high population centers (e.g., Metropolitan Detroit, Lansing, and
Grand Rapids) where many hunters live, an average hectare of publicly accessible hunting land
in this region is valued at $0.82/ha, less than an average hectare in the Upper Peninsula or
Northern Lower Peninsula.
61

The above results illustrate the annual economic benefits to ruffed grouse hunters from
the existence of the current levels of publicly accessible hunting land, relative to each type of
publicly accessible hunting land (or all publicly accessible hunting land in each region) being
completely eliminated. Although obtaining a clearer picture of the total economic benefits
from publicly accessible hunting land allows managers to compare and contrast hunter
economic benefits to state and federal land management costs, it does not provide insight into
the extent marginal changes to the amount of publicly accessible hunting land would impact
economic benefits from ruffed grouse hunting. Estimating hunter economic benefits from
marginal changes in publicly accessible hunting land allows managers to better account for the
benefits and costs of more realistic policy scenarios in which hunters compete with other
stakeholders within the political and legal systems for preferential access to federal and stateowned lands. Eliminating approximately 27,000 hectares of National Forest land contained
within 14 semi-primitive areas in Michigan is estimated to reduce economic benefits to ruffed
grouse hunters by about $45,000 annually.

2.5 Discussion
Although the economic impacts of small game hunting generated by hunter retail
expenditures have been well documented, almost no insight exists into the economic benefits
small game hunters derive from their hunting experiences. This is surprising, given the
popularity of small game hunting for species such as squirrel (1.7 million hunters, 21 million
days), rabbit (1.5 million hunters, 17 million days spent hunting), quail (841,000 hunters, 9

62

million days), and grouse (812,000 hunters, 8 million days) (USDOI 2011). This research helps to
address this gap in the literature by estimating economic benefits hunters realize from the use
of publicly accessible hunting lands for ruffed grouse hunting.
Furthermore, this research is easily compared and contrasted with research by Knoche
and Lupi (2012), which provides estimates of the annual economic benefits of public land
accruing to firearm and archery deer hunters in Michigan. Annual economic benefits from
publicly accessible hunting land to the average Michigan resident ruffed grouse hunter is
estimated to be $338.98, which is about 310% and 280% greater than the annual economic
benefits for the average Michigan firearm deer hunter ($82.94) and archery deer hunter
($88.71), respectively (Knoche and Lupi 2012). These results are intuitive within the context of
the use of public land by deer and ruffed grouse hunters, given that 79% of ruffed grouse
hunters hunted on public land at least once during the 2009 hunting season (Frawley 2012),
compared to 41% for deer hunters during the 2006 hunting season (Frawley and Rudolph
2008). Additionally, increasing levels of public access variables, Federal, State, and CFA, were
statistically significant and positive predictors of hunter site selection for both the ruffed grouse
travel cost model presented in this paper as well as the two deer hunting travel cost models
(Knoche and Lupi 2012). Both studies provide evidence that hunters for deer and grouse have
stronger preferences for state-owned land and CFA land, relative to preferences for federallyowned land. Specifically, coefficients on Federal variables are markedly lower than State and
CFA coefficients in the three models described above, with State and CFA coefficients being
between 107% and 187% greater than the coefficients for Federal in the ruffed grouse model
and firearm deer hunting model results, and between 43% and 46% greater than Federal in the
63

archery deer hunting model results. State and CFA coefficients are relatively comparable, with
the greatest coefficient difference in the firearm model where the CFA coefficient was 26%
greater than the State coefficient. The multiple types of state land (e.g., state forests,
recreation areas, game/wildlife areas) may offer enhanced access opportunities through more
developed facilities and increased access points with CFA land offering similar access benefits
via access points and roads built and maintained to facilitate timber management. In contrast,
much of Michigan’s federally-owned land consists of large, contiguous blocks of national forest
land, which may present access challenges to recreational hunters.
Finally, despite the estimate that annual economic benefits to ruffed grouse hunters
would be reduced by about $45,000 as a result of the proposed policy to ban hunting on
approximately 27,000 hectares of national forest land, this amount is only a portion of the lost
annual economic benefits from such a policy. This land, which is located in the Northern Lower
Peninsula of Michigan, offers additional, diverse hunting opportunities including (but not
limited to) white-tailed deer, black bear, wild turkey, and small game species such as rabbit and
squirrel. The approximately 88,300 resident ruffed grouse hunters in Michigan are only a small
portion of the 753,000 people that hunted in Michigan in 2006 (USDOI 2006), suggesting that
reductions in economic benefits to all hunters might be many times the lost economic benefits
to ruffed grouse hunters. A rigorous cost-benefit analysis of this policy would need to include
these reduced benefits to these other hunters as result of the closure, as well as any changes in
economic benefits to individuals whose outdoor recreational experiences are positively
affected from the closure to hunting.

64

2.6 Management Implications
The results within are relevant to wildlife managers and policy makers who are
considering whether to increase or enhance access for ruffed grouse hunters to both publiclyowned and privately-owned, publicly accessible hunting land. When faced with difficult
decisions on how to best allocate limited public funds to provide the greatest benefit to
hunters, anglers, and other outdoor recreationists, policy makers can look at this research and
note that a relatively small group of outdoor recreationists (about 88,300 resident ruffed
grouse hunters) derive about $30 million in annual economic benefits from the availability of
publicly accessible hunting land. With the annual economic benefits associated with publicly
accessible land for deer hunting in Michigan estimated to be over $80 million (Knoche and Lupi
2012), along with the economic benefits of this land for hunting other game species in Michigan
(e.g., wild turkey, pheasant, waterfowl, black bear) and for participating in other types of
outdoor recreation (e.g., angling, hiking, camping, mountain biking, cross country skiing) still
unaccounted for, total economic benefits accruing to users of public lands in Michigan is
hundreds of millions of dollars per year. The large economic benefits of state-owned land
relative to federally-owned land and CFA land takes on more importance with the Governor of
Michigan recently signing into law Senate Bill No. 248 (2012) which caps the amount of land the
State of Michigan can own at 4.65 million acres (~1.88 million ha) until the legislature approves
a strategic plan for buying and selling land. Although when (or if) the Michigan legislature will
develop and approve a strategic plan is not clear, in the meantime the state is faced with sharp
constraints on acquiring land for outdoor recreation purposes, land that this analysis shows to
be quite valuable to a single segment of outdoor recreationists. Finally, this analysis also
65

indicates that a policy proposal to close about 27,000 ha of national forest land from hunting
would result in an annual loss of over $45,000 in economic benefits to ruffed grouse hunters in
Michigan. Given the relatively stronger preferences for state-owned and CFA land over
federally-owned land, it would be expected that similar management actions undertaken on
these lands would result in greater losses of economic benefits to ruffed grouse hunters.
Regardless, with the roughly 648,000 days spent hunting ruffed grouse in Michigan (Frawley
2012) being a small percentage of the estimated 11.9 million days spent hunting in Michigan in
2006 (USDOI 2006), reduced economic benefits to ruffed grouse hunters would likely comprise
only a small portion of the losses realized by all hunters.

66

REFERENCES

67

REFERENCES

Adams, R. M., O. Bergland, W.N. Musser, S.L. Johnson, and L.M. Musser. 1989. User Fees and
Equity Issues in Public Hunting Expenditures: The Case of Ring-necked Pheasant in
Oregon. Land Economics 65: 376–393.
American Automobile Association (AAA). 2008. Your driving costs. AAA Association
Communication, Heathrow, Florida, USA.
Benson, D.E. 2001. Survey of State Programs for Habitat, Hunting, and Nongame Management
on Private Lands in the United States. Wildlife Society Bulletin 29: 354–358.
Berrens, R.P., and R.M. Adams. 1998. Applying Contingent Valuation in the Design of Fee
Hunting Programs: Pheasant Hunting in Oregon Revisited. Human Dimensions of Wildlife
3: 11–25.
Bin, O., C.E. Landry, C.L. Ellis, and H. Vogelsong. 2005. Some Consumer Surplus Estimates for
North Carolina Beaches. Marine Resource Economics 20: 145–161.
Boxall, P.C. 1995. The Economic Value of Lottery-rationed Recreational Hunting. Canadian
Journal of Agricultural Economics 43: 119–131.
Cameron, T.A. 1992. Combining contingent valuation and travel cost data for the valuation of
nonmarket goods. Land Economics 68: 302–317.
Dessecker, D. 2006. BCR12: Boreal Hardwood Transition. In Ruffed Grouse Conservation Plan,
eds D. R. Dessecker, G. W. Norman, and S. J. Williamson. Association of Fish and Wildlife
Agencies, Resident Bird Working Group, Washington, D.C., USA.
Dhuey B. 2008. Small Game Harvest 2007-08. Wisconsin Department of Natural Resources,
Bureau of Science Services, Madison, Wisconsin, USA.
Edwards, P.E.T., G.R. Parsons, and K.H. Myers. 2011. The economic value of viewing migratory
shorebirds on the Delaware Bay: An application of the single site travel cost model using
on-site data. Human Dimensions of Wildlife 16: 435–444.
English, D.B.K., and J.M. Bowker. 1996. Sensitivity of Whitewater Rafting Consumers Surplus to
Pecuniary Travel Cost Specifications. Journal of Environmental Management 47: 79–91.
Frawley, B.J. 2006. Demographics, Recruitment, and Retention of Michigan Hunters: 2005
update. Wildlife Division Report 3462. Michigan Department of Natural Resources,
Lansing, Michigan, USA.
Frawley, B.J. 2008. 2007 Small Game Harvest Survey. Wildlife Division Report 3493. Michigan
Department of Natural Resources, Lansing, Michigan, USA.

68

Frawley, B.J., and B.A. Rudolph. 2008. 2006 Deer Hunter Opinion Survey. Wildlife Division
Report No. 3482. Michigan Department of Natural Resources, Lansing, Michigan, USA.
Frawley, B.J. 2012. 2009 Small Game Harvest Survey. Wildlife Division Report 3541. Michigan
Department of Natural Resources, Lansing, Michigan, USA.
Frawley, V.R., and C.A. Stewart. 2009. Ruffed Grouse and American Woodcock Status in
Michigan, 2009. Wildlife Division Report 3504. Michigan Department of Natural
Resources, Lansing, Michigan, USA.
Freeman, A.R. 1993. The Measure of Environmental and Resource Values. Resources for the
Future, Washington, D.C., USA.
Fried, B.M., R.M. Adams, R.P. Berrens, and O. Bergland. 1995. Willingness to Pay for a Change in
Elk Hunting Quality. Wildlife Society Bulletin 23: 680–686.
Grijalva, T.C., R.P. Berrens, A.K. Bohara, P. Jakus, and W.D. Shaw. 2002. Valuing the Loss of Rock
Climbing Access in Wilderness Areas: a National-Level, Random-Utility Model. Land
Economics 78: 103–120.
Haab, T., and K. McConnell. 2002. Valuing Environmental and Natural Resources: Econometrics
of Non-Market Valuation. Edward Elgar, Cheltenham, United Kingdom.
Hansen, L., P. Feather, and D. Shank. 1999. Valuation of Agriculture’s Multi-Site Environmental
Impacts: An Application to Pheasant Hunting. Agricultural and Resource Economics
Review 2: 199–207.
Helland, J. 2006. Walk-In Hunting Programs in Other States. Minnesota House of
Representatives Research Department, St. Paul, Minnesota, USA.
Kahn, M.E. 1998. A Household Level Environmental Kuznets Curve. Economics Letters 59: 269–
273.
Kotchen, M., M. Moore, F. Lupi, and E. Rutherford. 2006. Environmental Constraints on HydroPower: An Ex-Post Benefit-Cost Analysis of Dam Relicensing. Land Economics 82: 384–
403.
Knoche, S., and F. Lupi. 2007. Valuing Deer Hunting Ecosystem Services from Farm Landscapes.
Ecological Economics 64: 313–320.
Knoche, S., and F. Lupi. 2012. The Economic Value of Publicly Accessible Deer Hunting Land.
Journal of Wildlife Management 76: 462–470.
Layman, R.C., J.R. Boyce, and K.R. Criddle. 1996. Economic Valuation of the Chinook Salmon
Sport Fishery of the Gulkana River, Alaska, Under Current and Alternative Management
Plans. Land Economics 72: 113–128.

69

Lupi, F., J. Hoehn, and G. Christie. 2003. Using an Economic Model of Recreational Fishing to
Evaluate Benefits of Sea Lamprey (Petromyzon marinus) Control. Journal of Great Lakes
Research 29: 742–754.
Michigan Department of Natural Resources. 2009. Michigan 2008–2009 Upland Game Harvest
Report. Michigan Department of Natural Resources, Wildlife Division, Lansing, USA.
Parsons, G.R. 2003. The Travel Cost Model. In P. A. Champ, K. J. Boyle, and T. C. Brown, eds. A
primer on nonmarket valuation, pp 427–440. Kluwer Academic Publishers, Dordrecht,
The Netherlands.
Remington, T.E., M.J. Manfredo, J.J. Vaske, and S.M. Demasso. 1996. Fee Hunting Pheasants in
Colorado: Experimental Evidence. Human Dimensions of Wildlife 1: 51–59.
Rusch, D.H., J.R. Cary, and L.B. Keith. 1999. Pattern and Process in Ruffed Grouse Cycles.
Proceedings of Midwest Fish and Wildlife Conference 61: 238.
Sarker, R., and Y. Surry. 1998. Economic Value of Big Game Hunting: The Case of Moose Hunting
in Ontario. Journal of Forest Economics 4: 29–60.
Schwabe, K.A., P.W. Schuhmann, R. Boyd, and K. Dooridian. 2001.The Value of Changes in Deer
Season Length: An Application of the Nested Multinomial Logit Model. Journal of
Environmental and Resource Economics 19: 131–147.
Shulstad, R.N., and H.H. Stoevener. 1978. The Effects of Mercury Contamination in Pheasants
on the Value of Pheasant Hunting in Oregon. Land Economics 54: 39–48.
Southwick Associates. 2007. State and National Economic Effects of Fishing, Hunting and
Wildlife-Related Recreation on U.S. Forest Service-Managed lands. U.S. Forest Service,
Washington, D.C., USA.
Stewart, C.A., M. Donovan, and V.R. Frawley. 2007. BCR 23: Prairie hardwood transition. In D.R.
Dessecker, G.W. Norman, S.J. Williamson, eds. Ruffed grouse conservation plan, pp 56–
59. Association of Fish and Wildlife Agencies, Resident Bird Working Group, Washington,
D.C., USA.
U.S. Department of Agriculture [USDA]. 2009. National Visitor Use Monitoring Results: Huron
Manistee National Forest. U.S. Department of Agriculture Forest Service, Washington,
D.C., USA.
U.S. Department of Agriculture [USDA]. 2012. Record of Decision: Final Supplemental
Environmental Impact Statement. U.S. Department of Agriculture Forest Service,
Milwaukee, Wisconsin, USA.
U.S. Department of the Interior [USDOI]. 2011. National survey of fishing, hunting, and wildlifeassociated recreation. U.S. Fish and Wildlife Service, Washington, D.C., USA.

70

U.S. Department of the Interior [USDOI]. 2006. National survey of fishing, hunting, and wildlifeassociated recreation - Michigan. U.S. Fish and Wildlife Service, Washington, D.C., USA.

71

CHAPTER 3: HARVESTING BENEFITS FROM HABITAT RESTORATION – THE INFLUENCE OF
LANDSCAPE POSITION ON ECONOMIC BENEFITS TO PHEASANT HUNTERS

3.1 Introduction
The precarious state of the world’s ecosystems has been well documented, with the
Millennium Ecosystem Assessment (2005) characterizing 15 out of 24 ecosystem services as
either being in decline or being used unsustainably. Substantial efforts focused on halting and
reversing these declines have emerged, as evidenced by the tremendous growth of the
ecological restoration and conservation biology research disciplines (Young 2000, Suding 2011)
and the substantial financial resources dedicated to restoration and conservation efforts.
According to Bernhardt et al. (2005), between 14 and 15 billion dollars were spent on over
37,000 river/stream restoration projects in the continental U.S. from 1990 to 2003. The United
States Department of Agriculture (USDA) Conservation Reserve Program (CRP) has provided
farmers and landowners with rental payments totaling approximately $43 billion between the
years 1987 to 2014 (USDA 2014). Given the high cost of restoration and conservation efforts
and the limited budgets available to achieve objectives, it is critical to develop a better
understanding of whether and to what extent restoration and conservation efforts impact
human well-being. Establishing and accounting for key linkages between basic biophysical
structures and processes of the natural environment, the production of ecosystem functions
and services, and finally, benefits to humans can allow researchers to characterize and quantify
human benefits associated with costly habitat management actions (Haines-Young and Potchin
2010, de Groot et al. 2002). Characterizing the change in human well-being in monetary terms
facilitates a cost-benefit analysis of potential restoration projects, and estimating the economic
72

values of ecosystem services has been identified as a critical research area (Millennium
Ecosystem Assessment 2005, Carpenter et al. 2006, Carpenter et al. 2009).
Ecological restoration actions have the potential to generate economic benefits to
outdoor recreationists such as recreational hunters and anglers through improved catch or
harvest rates at hunting or fishing locations. Several research efforts have involved estimating
the economic benefits of changes in water quality for recreational anglers by establishing and
accounting for the linkages between water quality, angler catch rate, and economic benefits.
Lipton and Hicks (1999, 2003) and Massey et al. (2006) estimate the economic benefits
resulting from changes in water quality to recreational anglers in the Mid-Atlantic region by
linking expected catch models incorporating dissolved oxygen to recreation demand models of
angling site choice. Another area of research involves estimating the economic benefits of
water temperature mitigation strategies by linking water temperature, fish
population/expected catch, and angler economic benefits. Huang et al. (2012) estimate the net
economic benefits to recreational anglers from altered water releases and diversions designed
to mitigate stream temperatures, and Johnson and Adams (1988) use a contingent valuation
survey in conjunction with a fish production model to estimate the economic benefits to
recreational anglers associated with increased streamflow.
The effect of restoration siting decisions and land use change on ecosystem service
provision is receiving increasing attention from researchers with respect to a variety of
ecosystems, including wetlands (see, e.g., White and Fennessy 2005, Lesta 2009, MorenoMateos et al. 2010), marine reserves (Leslie et al. 2003), alpine environments (Fontana et al.

73

2009) and floodplain/riparian environments (Oetter et al. 2004, Johnson et al. 2007).
Researchers are also beginning to explore how spatial variability across landscapes affects the
economic values of ecological outputs (Barbier 2012, Barbier et al. 2008, Sanchirico and Mumby
2009), and recreation demand modelers have begun incorporating landscape metrics such as
landscape patch shape complexity, path variation, and total edge directly in recreation demand
models (see Bestard and Font 2009). However, there is a lack of research linking such spatially
explicit information regarding the provision of ecosystem services to outdoor recreationist (e.g.,
hunter, angler) behavior in order to evaluate economic benefits across a range of ecological
restoration. This is surprising, given that the increasingly sophisticated spatially explicit
ecological production functions which use landscape metrics to predict species abundance and
richness measures for small mammals (Gibson et al. 2004), large mammals (Maier et al. 2005),
avian species (Radford et al. 2005, Nielson et al. 2008), and fish and aquatic invertebrates (Pess
et al. 2002, Arbuckle and Downing 2002) could be adapted to generate region-specific species
abundance estimates for use in discrete choice recreation demand models of popular outdoor
recreation activities such as away-from-home wildlife viewing (22.5 million participants),
recreational fishing (33.1 million participants), and recreational hunting (13.7 million
participants) (USDOI 2011). The discrete choice recreation demand approach is well-suited for
incorporating spatially explicit outputs generated by an ecological production function as site
quality attributes entering into the individual’s recreation demand decision, thereby allowing a
tracing of economic benefits back to the original restoration actions. Accounting for these
linkages could help improve targeting of restoration efforts of existing large-scale conservation
programs such as the USDA CRP (Claassen et al. 2008) by providing resource managers with a
74

greater understanding of the spatial distribution of landscape characteristics that would
generate the largest conservation “bang for the buck”.
In this chapter, I estimate the economic benefits of CRP land for pheasant hunters in
Michigan by linking a spatially explicit model of pheasant sightings as a function of landscape
characteristics developed by Nielson et al. (2008) (henceforth referred to as the Nielson Model)
to a recreation demand model of pheasant hunting in Michigan. Further, I focus special
attention on the extent to which economic benefits accruing to hunters are contingent upon
the site-selection for CRP conservation actions. Agricultural landscapes have the potential to
provide a wide variety of valuable ecosystem services (Swinton et al. 2007, Knoche and Lupi
2007), and provision of ecosystem services on such landscapes can be aided by targeted CRP
conservation/restoration efforts on these lands. Researchers have found that CRP conservation
generate benefits for sought-after games species, including white-tailed deer (Gould and
Jenkins 1993) but especially for avian species such as waterfowl (Kantrud 1993, Reynolds 1994,
Reynolds 2001), sage grouse (Schroeder and Vander Haegen 2011), northern bobwhite (Riffell
et al. 2008) and ring-necked pheasant (Riley 1995, King and Savidge 1995, Haroldson et al.
2006, Nielson et al. 2008). However, despite the cost of the CRP and the documented
relationship of CRP land to species populations sought by hunters and wildlife-watchers, efforts
to estimate the economic benefits of CRP land to outdoor recreationists have been sparse. One
exception is Hansen et al. (1999), who use a recreation demand model to estimate that the
economic value of CRP land to pheasant hunters is about $80 million annually. Hansen et al.
(1999) use a reduced form approach in which landscape characteristics presumed to influence
pheasant population enter directly into the pheasant hunter site decision problem. While
75

reduced form approaches are often employed by economists when information is lacking
regarding the relationship between the biophysical structure/processes of an ecosystem and
ecosystem service outputs (see, e.g., Hicks, 2004), such approaches do not fully characterize the
complex nature of the ecological production process in which ecosystem services are generated
through the underlying biophysical structure/processes. Furthermore, Newbold and Massey
(2010) show that reduced form models can produce biased estimates of willingness to pay and
recommend caution when using such models for ecosystem valuation. The availability of the
ecological production function via the Nielson model overcomes the limitations associated with
the reduced form approach and permits the construction of region-specific pheasant sightings
which are a function of the amount and spatial distribution of relevant landscape characteristics
such as agricultural land and CRP. This enables the examination of how hunter economic
benefits vary contingent upon the selection of different geographic areas for CRP-related
restoration.

3.2 Methodology
3.2.1 Ecological Production Function
Equation 1 illustrates the ecological production function referred to as the Nielson
Model, which predicts pheasant sightings on a North American Breeding Bird Survey(BBS)3
route j as a function of a time trend (𝑦𝑟), vector of landscape characteristics (𝑐ℎ𝑗 ), a vector of
landscape size and distribution metrics (𝑚𝑗 ) within a 1000m buffer of BBS route j:
3

See Sauer et al. (2011) for summary of the North American Breeding Bird Survey and survey
results and analysis from 1966 – 2010.
76

1)

𝑃ℎ𝑒𝑎𝑠𝑎𝑛𝑡 𝑆𝑖𝑔ℎ𝑡𝑖𝑛𝑔𝑠𝑗 = exp[𝛼0 + 𝛼𝑦𝑟 𝑦𝑟 + 𝛼𝑐ℎ 𝑐ℎ𝑗 + 𝛼𝑚 𝑚𝑗 ].

The Nielson Model uses a Bayesian hierarchical modeling approach, fitting the model
using Markov-chain Monte Carlo methods to model BBS counts of pheasants as overdispersed
Poisson counts, with deviance information criterion (DIC) used as a guide to identify the most
parsimonious model. For further details on developing and estimating the Nielson Model see
the Nielson et al. (2008) paper; details relating to the adaptation and replication of the Nielson
model for use as a site quality variable for this analysis are presented in Section 3.1.

3.2.2 Discrete Choice Recreation Demand Model
The discrete choice recreation demand model approach is well-suited for incorporating
spatially explicit outputs generated by an ecological production function as site quality inputs
into the individual’s recreation demand decision, thereby allowing a tracing of economic
benefits back to the original restoration actions. Estimating changes in economic benefits
resulting from changes in recreation site attributes proceeds by examining the tradeoffs
individuals make between the levels of quality attributes available at recreation sites and the
costs associated with traveling to those sites (Parsons 2003, Haab and McConnell 2002). The
behavioral foundation for the discrete choice recreation demand model is random utility
theory, which posits that the indirect utility individual n receives from visiting a recreational site
j (𝑣𝑛𝑗 ) is a function of travel cost (𝑡𝑐𝑛𝑗 ) which varies by respondents as well as by recreational
site, a vector of site quality characteristics (𝑞𝑗 ), which vary only by recreational site, and 𝑒𝑛𝑗 ,
77

which is a random error term capturing unmeasured characteristics associated with a
recreational site:
2)

𝑣𝑛𝑗 = 𝛽𝑡𝑐 𝑡𝑐𝑛𝑗 + 𝛽𝑞 𝑞𝑗 + 𝑒𝑛𝑗 .

In equation 2, the larger a site quality parameter βq is in absolute value, the greater the degree
the site quality characteristic influences the indirect utility of individual n, whereas the sign on
the parameter reveals whether the site characteristic has a positive or negative impact on
individual n’s indirect utility. Economic theory suggests that indirect site utility decreases with
increases in travel costs, and increases with site characteristics thought to be desirable, such as
game species population.
Within random utility theory, an individual chooses to visit the site which offers the
highest indirect utility. Thus, site k would be chosen over all other possible sites j when
equation 3 holds, with C identifying the set of possible recreation sites in an individual’s choice
set:
3)

𝛽𝑡𝑐 𝑡𝑐𝑛𝑘 + 𝛽𝑞 𝑞𝑘 + 𝑒𝑛𝑘 ≥ 𝛽𝑡𝑐 𝑡𝑐𝑛𝑗 + 𝛽𝑞 𝑞𝑗 + 𝑒𝑛𝑗 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑗 𝑖𝑛 𝐶.

Since the indirect utility function described in equation 2 contains an error term which captures
un-measurable site characteristics (characteristics known by the individual but unobserved by
the researcher), equation 4 is used to estimate the probability individual n chooses site k:
4)

Prob (𝛽𝑡𝑐 𝑡𝑐𝑛𝑘 + 𝛽𝑞 𝑞𝑘 + 𝑒𝑛𝑘 ≥ 𝛽𝑡𝑐 𝑡𝑐𝑛𝑗 + 𝛽𝑞 𝑞𝑗 + 𝑒𝑛𝑗 )for all 𝑗 in C.

These probability functions can be interpreted as the expected demand functions for the sites.
A 2-level nested logit model is used, with equation 5 defining the probability an individual
78

chooses site 𝑘 in nest 𝑙 from the J possible sites, where 𝜃𝑚 are distributional parameters to be
estimated (Haab and McConnell 2002).

5)

Prob(𝑘, 𝑙) =

𝑣𝑗𝑙 𝜃𝑙−1
𝑣
𝐽𝑙
exp( 𝜃𝑘𝑙 )[∑𝑗=1
exp( 𝜃 )]
𝑙
𝑙
𝑣𝑗𝑚 𝜃
𝐽𝑚
𝐿
𝑚
∑𝑚=1[ ∑𝑗=1 exp(
𝜃𝑚 )]

.

The probability individual n takes a trip to site k in nest 𝑙 is a function of indirect utility 𝑣
associated with site k in nest 𝑙, as well as the indirect utility associated with substitute sites.
This form allows sites within a nest to have error terms that are more correlated with other
sites within the nest than with sites in a different nest. Maximum likelihood estimation was
used to identify the values for the coefficients which maximize the likelihood function, Li,
defined in equation 6. In equation 6, 𝑟𝑘𝑙 = 1 if individual n visited site k in nest 𝑙 and equals
zero otherwise, and Prob(𝑘, 𝑙) is the nested logit form from equation 5:
𝐿

6)

𝐽𝑙

𝐿𝑖 = ∏ ∏ Prob(𝑘, 𝑙)𝑟𝑘𝑙 .
𝑙=1 𝑘=1

3.2.3 Estimating Economic Benefits for Site Quality Changes
Once parameters in the indirect utility function have been estimated, it is then possible
to estimate the changes in economic benefits resulting from changes in the levels of
recreational site attributes or changes in available sites. The economic benefits associated with
a change at a particular site or sites is given by the difference in the expected maximum utility
with and without changes in the levels of site attributes and/or site access, and this difference
79

is divided by the marginal utility of income. When the site probability functions take the nested
logit form, this measure of economic benefits is defined in equation 7:
𝜃𝑚

7)

𝑊𝑇𝑃 =

1
ln
𝛽𝑡𝑐

′
𝑣𝑗𝑙′
𝑚
∑𝐿𝑚=1[∑𝐽𝑗=1
exp( 𝜃 )]

𝑙

𝜃𝑚

𝑣𝑗𝑙
𝑚
∑𝐿𝑚=1[∑𝐽𝑗=1
exp( 𝜃 )]
𝑚

.

In equation 7, WTP denotes the change in per-trip economic benefits resulting from a
quality and/or access change at one or multiple sites; 𝑣 are the indirect utilities for sites in the
original state and 𝑣 ′ are the indirect utilities for sites after the quality and/or access changes
(Haab and McConnell 2002). This is a “per-trip” measure in the sense that it’s units refer to
trips to all sites in the choice set. An estimate of the seasonal economic benefits to
recreationists from a quality change is obtained by multiplying WTP by the total number of trips
individuals made to the recreational sites within the choice set within the season of interest.

3.3 Survey and Attribute Data
Information on pheasant hunter site choice and site visitation frequency was obtained
from the 2008–2009 Michigan Upland Game Harvest Report mail survey (MDNR, 2009)
administered by the Michigan Department of Natural Resources (MDNR) (n=9,987; 57%
response rate)4. The analysis included only individuals (n = 682) who reported: a) hunting
pheasant during the 2008 Michigan pheasant hunting season, b) the county (or counties) in

4

Even though the survey achieved a reasonably high response rate, we do not have access to
the data on nonrespondents and a nonresponse study was not conducted.
80

which they hunted pheasant, and c) the number of days spent hunting in each county.
Consistent with the spatial resolution of the survey data, hunting sites are defined at the county
level. This survey provides hunting site choice information for 682 hunters who reported
spending a total of 3265 days hunting pheasant during the pheasant hunting season.
For each hunter, the survey data contains the number of days hunted at each county
visited by the hunter. To run the site choice model it was necessary to convert these days into
trips per county. If days were converted into trips using a simple mean number of days per trip,
it would cause difficulties if trips to farther sites are more likely to be multiple day trips. Instead
a distance-based conversion was used which was developed using data for other hunting
activities where both days and trips were known (see dissertation chapter 1). Dissertation
chapter 1 estimated travel cost models for multiple hunting activities and compared welfare
estimates using hunter-reported trips with estimates using a variety of plausible relationships
between trips, days and distance to hunting sites. That analysis found many distance-based
conversions that closely approximated those using hunter-reported trips and found one that
only had an approximately 1% difference. The latter conversion (also described and applied in
dissertation chapter 2) was applied to this data, generating a total estimate of 3070 pheasant
hunting trips for the 682 pheasant hunters in the model.
To account for multiple trips from the same individual when estimating the discrete
choice model, the frequency weighting procedure was used in Stata® to weight each observed
site choice by the estimated number of trips an individual took to the site. Individuals who took
trips to multiple sites have a choice set for each unique site visited, weighted by its respective

81

number of visits. The clustered-robust standard errors procedure is used to account for the
correlated error terms across different site choices by the same individual.
The vast majority of pheasant hunting in Michigan takes place in Michigan’s Lower
Peninsula. Of the 2008-2009 Upland Game Harvest Survey respondents reporting hunting
pheasant during the 2008-2009 Michigan pheasant hunting season (n=871), only 45 (5.2%) lived
in the Upper Peninsula. Further, only four of the 45 individuals living in the Upper Peninsula
reported hunting a county in the Lower Peninsula (8.9%) and only five of the 826 individuals
living in the Lower Peninsula reported hunting a county in the Upper Peninsula (0.6%). Thus, it
appears likely that most hunters do not consider counties outside their peninsula of residence
when evaluating potential pheasant hunting site options. Finally, the majority of the Upper
Peninsula is closed to pheasant hunting – 10 out of 15 counties in the Upper Peninsula have no
legal pheasant hunting season, with only one county in the Upper Peninsula having all of its
land open to pheasant hunting. Given both the limited pheasant hunting in Michigan’s Upper
Peninsula and limited travel between the Upper and Lower Peninsulas for the purpose of
pheasant hunting, only hunters and counties in Michigan’s Lower Peninsula are included in the
choice model. Because there are also likely to be regional differences within Michigan’s Lower
Peninsula, with similarities between hunting sites within regions and differences between
hunting sites across regions, two nests are defined in this nested logit model of pheasant
hunting. The first nest, called the No_Extended_Season nest, consists of counties in the
Northern Lower Peninsula and counties located in the western portion of the Southern Lower
Peninsula which do not have sizeable portions with the extended December pheasant hunting

82

season (MDNR 2008). The other nest, called the Extended_Season nest, contains the other
counties in Michigan’s Lower Peninsula. See Figure 3.2 for the delineation of nest membership.
County attributes used as independent variables in the nested logit model include the
travel cost (time cost + vehicle operating cost) associated with traveling to and from a county,
the amount of publicly accessible hunting land in the county, the size of the county, and the
pheasant sightings estimate from the Nielson Model summed across all landscape parcels in the
county. The variable Price is computed to reflect the round-trip travel costs from each hunter’s
residence to each of the 68 counties in Michigan’s Lower Peninsula. PC*Miler® software was
used to calculate the driving distance from each individual’s zip code of residence to the zip
code that is closest to the geographic center of each hunting site. Vehicle operating cost was
computed by multiplying the per-mile vehicle operating cost by the distance traveled. Per-mile
vehicle operating costs, which includes average per-mile costs for gas, maintenance and tires,
was 17.0 cents in 2008 (AAA 2008). Per-mile depreciation rates were calculated using 15,000
miles driven annually as a reference point, with 10,000 and 20,000 miles driven resulting in
decreased deprecation and increased depreciation of about 3.5 cents per mile and 3.8 cents per
mile (AAA 2008), respectively. The midpoint of these two estimates (3.65 cents) is added to
per-mile operating costs of 17.0 cents to obtain a per-mile vehicle operating cost of 20.65
cents. The other component of travel costs is time cost, which is the opportunity cost of the
time required for round-trip travel to the site. To estimate an opportunity cost of time for each
individual, the individual’s address is used in conjunction with ArcGIS® software to assign each
individual the United States Census Bureau median household income for their census tract of

83

residence.5 Individuals whose addresses were not recoverable using ArcGIS software (4.3% of
the population) were assigned the median household income for their zip code of residency. To
value the time component of travel cost, I follow the many studies in the recreation literature
by using one-third of the wage rate (Parsons 2003). Wage rate was computed by dividing
median income by the number of work hours per year (2080). An estimate of travel time was
computed using an average trip speed of approximately 64 km (40 miles) per hour.
Other county attributes in the model include the variable Public_Access, consisting of
the total number of acres of publicly accessible hunting land in each county. This includes
federally-owned land, state-owned land, and privately-owned land that is publicly accessible to
hunters through Michigan’s Commercial Forest Act program. Finally, the variable Size reflects
the geographic area of each county. In models using aggregate sites such as the 68 counties in
this choice set, many studies include a measure of group size to mitigate potential differences
in site heterogeneity (Haener et al. 2004). However, unlike some recreation such as fishing at
lakes (Lupi and Feather 1998), pheasant hunting takes place at myriad places across the
landscape, and therefore does not have easily defined access points. Never the less,
differences in the amount of potential hunting locations across counties is controlled for by
including the variables for publicly accessible land and geographic size of a county.

5

When lacking survey data on individual income, recreation demand modelers often assign
individuals a proxy for income via aggregate measures which characterize income at a
geographic scale, such as median household income for census areas or zip codes (see, e.g.,
Heberling and Templeton (2009), Whitehead et al. (2009), Henderson et al. (1999), and Jakus et
al. (1997)). This approach is based on the tendency of individuals to cluster geographically by
socioeconomic status. If membership in the sub-group of interest (in this case, pheasant
hunters) is correlated with higher (or lower) median incomes compared to others in their
census tract, then welfare estimates could be biased downwards (upwards).
84

3.3.1 Computing Pheasant Sightings using the Nielson Model
For the Pheasant_Sightings variable, the Nielson Model ecological production function is
used which predicts pheasant sightings obtained from the BBS as a function of landscape
characteristics obtained from the National Land Cover Dataset (NLCD) and the USDA. To
develop this model, Nielson et al. (2008) estimated pheasant sightings associated with
landscape characteristics of 3,888 BBS routes. Equation 8 is Nielson et al.’s (2008) study areawide model (which I refer to as the Nielson Model), which estimates the number of pheasants
counted along a survey route in year i.

8)

𝑇𝑖 = exp[1.5451 − 0.0059(year𝑖 − 1996) + 0.2748(NLCD Woody Vegetation)
+ 0.7040(NLCD Herbaceous Vegetation)
+ 1.4949(NLCD Agricultural Field) − 0.6584(𝑁LCD Agricultural Field)2
+ 0.1991(𝐶RP Herbaceous Vegetation) − 0.0526(Mean Patch Size)
− 0.1702(Interspersion and Juxtaposition Index)]

The Nielson Model was originally fit using landscape characteristic data from the 1992
NLCD and USDA CRP data, and landscape metrics Mean Patch Size (MPS) and Interspersion and
Juxtaposition Index (IJI). Four NLCD landscape characteristic categories are included in the
model: NLCD Woody Vegetation (includes NLCD classifications of shrubland and
orchards/vineyards), NLCD Herbaceous Vegetation (grassland/herbaceous and pasture/hay),
NLCD Agricultural Field (row crops, small grains, and fallow) and CRP Herbaceous Vegetation
(14 types of herbaceous CRP enrollment classifications). Finally, two landscape patch metrics –
85

MPS and IJI – are included in the model as well and were computed using FRAGSTATS v4
(McGarigal et al. 2012). Variables in equation 8 were calculated using a 1000m buffer around
the BBS route. Assuming a straight BBS route, this corresponds to an area of 19,487 acres.
Nielson et al. (2008) also estimated the model using buffers of 400m and 700m, however,
1000m was used for the final model as this approach yielded the lowest deviance information
criterion and the fewest variables.
In replicating the Nielson Model to estimate site-specific pheasant sightings estimates in
Michigan, I closely followed the approach described in Nielson et al. (2008) while taking
advantage of recently available landscape characteristic data. Using 2006 NLCD and Michigan
USDA CRP geolocation data6, a fishnet procedure was employed using ArcGIS software to
construct a grid consisting of 19,487 acre landscape parcels which overlays Michigan’s Lower
Peninsula (see Figure 3.1). Landscape characteristic data for each of these parcels was used in
equation 8 to produce a pheasant sightings estimate for that parcel. Pheasant sightings
estimates for each parcel were used to construct total pheasant sightings for each county. For
landscape parcels which overlap two or more counties, the pheasant sightings estimate for that
landscape parcel was proportionally allocated to each county.

6

Access to otherwise protected USDA CRP data was granted through a USDA Section 1619
Cooperator Memorandum of Understanding issued by the Farm Service Agency. This
agreement prohibits the disclosure of the geospatial USDA CRP data, and as such, we do not
map Michigan-level USDA CRP data in our figures or otherwise provide information on USDA
CRP location within this paper.
86

Figure 3.1. Michigan lower peninsula counties and landscape parcels used to estimate pheasant
sightings

Figure 3.2 compares the county-level pheasant sightings predicted by the Nielson Model
to county-level pheasant brood observations through MDNR mail carrier surveys. The MDNR
mail carrier survey is an annual survey conducted during the summer months in which
cooperating rural mail carriers within 43 southern Michigan counties record the number of
pheasant broods observed during the survey period. The 43 counties were selected for the
mail carrier survey based on the potential of habitat in that county to support wild pheasant
populations - 41 of these 43 counties were identified in the 2013 National Wild Pheasant
Conservation Plan (Association of Fish and Wildlife Agencies 2013) as having wild pheasant
range. Further, MDNR has found the mail carrier data to be a good predictor of pheasant

87

abundance and fall harvest (Luukkonen 1998). In Figure 3.2, the MDNR mail carrier survey
estimates reflect the total brood observations in each county for the previous ten surveys (2004
– 2013). To facilitate comparison of MDNR mail carrier survey brood observations and the
pheasant sightings estimates from the Nielson Model, only Nielson Model estimates for the
counties surveyed via the MDNR mail carrier survey are included in Figure 3.2. County-level
estimates for the MDNR mail carrier survey and the Nielson Model are segmented into
quartiles to further aid comparison.

88

Figure 3.2. County-level estimates of pheasant abundance from the Nielson model and a
MDNR mail carrier pheasant brood survey and the delineation between regional nests in nested
logit model

Figure 3.2 illustrates similarities between the Nielson Model pheasant sightings
estimates and pheasant sightings estimated via the MDNR mail carrier surveys. The pearson
correlation coefficient is used to characterize the relationship between the Nielson Model
sightings predictions and MDNR mail carrier brood observations. The estimated pearson
correlation coefficient of + 0.62 indicates a positive correlation between MDNR mail carrier
pheasant brood observations over a ten year period and the county-level pheasant sightings
obtained from application of the Nielson Model.

89

3.4 Results
The nested logit model results are displayed in Table 3.1. Price is statistically significant
and negative, meaning that as the cost of traveling to a hunting site increases, the utility
associated with that site (and the probability of selecting that hunting site) decreases. As the
number of pheasant sightings and the amount of publicly accessible hunting land increases at a
hunting site, the utility associated with that hunting site increases. Dissimilarity parameters
represent the interdependence of unobserved utility in each nest; a dissimilarity parameter of
one indicates that there is no correlation in unobserved utility between sites in a given nest. In
the model, Extended_Season and No_Extended_Season are statistically signficant and different
than one, meaning that counties within each respective nest are similar in unobserved utility to
other sites in that nest.

90

Table 3.1. Nested logit model of pheasant hunting in Michigan
Coefficient

Clustered
Robust SE

P Value

Price

-0.053

0.005

P<0.01

Pheasant_Sighting

0.013

0.002

P<0.01

Public_Access (10,000 acres)

0.040

0.017

P<0.01

Size (100 sq. miles)

-0.022

0.031

P = 0.48

Extended_Season

0.603

0.063

P<0.01

No_Extended_Season

0.621

0.069

P<0.01

Independent Variables

Dissimilarity Parametersa

Model Statistics
-5966.6

Log-Likelihood
Wald chi2(4)

103.51 (P<0.0001)

Observations

682 individuals; 3070 choice occasions

a

P values for the dissimilarity coefficients are estimated against the null hypothesis that the
coefficient is equal to one.

3.4.1 Pheasant Hunter Benefits of CRP Land
Economic benefits of CRP land to Michigan pheasant hunters are estimated by
examining how hunter welfare changes when all 213,674 acres of CRP land is converted to the
2006 NLCD cultivated crops landscape characteristic4. First, the per-trip economic benefits of

4

We use recently available 2006 NLCD data to estimate pheasant sightings using the Nielson
Model, while Nielson et al. (2008) used 1992 NLCD data. For our purposes, the primary
difference between 1992 NLCD agricultural field category and the 2006 NLCD cultivated crops
category is that 2006 NLCD cultivated crops includes land used for orchards and vineyards,
while the 1992 NLCD data separately accounts for orchards and vineyards. The Nielson et al.
(2008) model includes orchards/vineyards in the NLCD woody vegetation category. We felt the
benefits from using 2006 NLCD data vastly outweighed this minor discrepancy in how
orchards/vineyards were categorized within the 1992 and 2006 NLCD datasets.
91

CRP land for pheasant hunters in Michigan are estimated (equation 7). The conversion of all
CRP land to NLCD cultivated crops results in a per-trip welfare loss of about $1.82. To obtain an
estimate of the reduction of seasonal economic welfare from this conversion, this per-trip
welfare loss was multiplied by the estimated number of trips taken in the season for all
pheasant hunters represented in the model. The estimate of the total number of trips
(183,081) is obtained by multiplying the total number of pheasant hunters (43,144) by the
percentage of these hunters both living in the Lower Peninsula and not hunting in the Upper
Peninsula (94.3%) and by the estimated number of hunting trips per person (4.5). Thus, the
total seasonal welfare loss to hunters from this conversion is estimated to be about $332,600
which yields a mean per acre CRP value for pheasant hunters of $1.56.

3.4.2 Michigan Pheasant Restoration Initiative (MPRI) Policy Scenario
The MDNR has partnered with the U.S. Fish and Wildlife Service, Michigan Department
of Agriculture, and non-governmental organizations such as Pheasants Forever and Michigan
United Conservation Clubs to implement the Michigan Pheasant Restoration Initiative (MPRI).
The stated objective of the MPRI is to establish 10 cooperatives (roughly 10,000 acres in size) in
Michigan throughout 3 pilot areas (see Figure 3) and restore 1,200 – 2,000 acres of pheasant
habitat in each of these cooperatives.

92

Figure 3.3 MPRI priority pheasant restoration areas within Michigan’s lower peninsula

To estimate the economic benefits to pheasant hunters associated with achieving the
MPRI restoration goal, it is assumed that the high-quality pheasant habitat restored via the
MPRI will have the same characteristics as land enrolled under the USDA herbaceous CRP
category. To ensure that the policy analysis within is consistent with the Nielson Model, it is
assumed that each of the nine counties in the MPRI pilot focus area will have one cooperative
consisting of a single 19,487 acre landscape parcel. The midpoint of the MPRI cooperative level
restoration goal is selected, thereby assuming that 1,600 acres of herbaceous CRP land is
restored within a single 19,487 acre landscape parcel in each county, for a total restoration of
14,400 herbaceous CRP acres across the nine counties. For each landscape parcel which is
completely within the borders of an MPRI pilot focus area county, a new pheasant sightings
estimate for that landscape is generated by converting 1,600 acres of cultivated crops within
93

that landscape to herbaceous CRP land. These new pheasant sightings estimates are used to
estimate the economic benefits associated with multiple policy scenarios in which different
landscape parcels are selected for location of the MPRI cooperatives. These policy scenarios
provide important insights regarding the extent to which economic benefits accruing to hunters
vary depending on ecological restoration siting decisions.
For the previous analysis regarding the seasonal economic benefits of CRP land to
pheasant hunters in Michigan, information on the size and location of CRP patches enabled the
accounting of changes in landscape metrics MPS and IJI resulting from the conversion of these
patches to cultivated crops. However, such precision with MPS and IJI is not possible in the
MPRI analysis due to uncertainties regarding the size and location of the cultivated crop
patches converted to herbaceous CRP land. For insight into how this conversion might impact
MPS (and hence impact pheasant sightings measures and hunter welfare), three assumptions
are used regarding the patch size of the cultivated crops land area converted to CRP. First, the
change in hunter welfare is estimated using the assumption that the 1,600 acres of restored
CRP land in each landscape parcel has a mean patch size equal to the mean patch size of all
herbaceous CRP patches throughout Michigan (6.3 acres), and that these patches are not
directly adjacent to other CRP patches. This results in 254 patches of herbaceous CRP added to
the landscape parcel. Second, given the potential of the cultivated crop conversion to occur on
land currently adjacent to herbaceous CRP land, hunter welfare is estimated using the
assumption that 50% of the herbaceous CRP patches are placed adjacent to existing patches.
This results in an addition of 127 patches to the landscape. Finally, the welfare implications of
restoration was examined under the assumption that all restored herbaceous CRP is adjacent to
94

existing herbaceous CRP, thus creating zero new patches and resulting in no change in MPS in
that landscape parcel.
It is more challenging, however, to develop plausible scenarios for how the conversion
of cultivated crops to herbaceous CRP land would affect the IJI variable. As IJI depends on the
relative proportion of patch adjacencies, the direction and the extent that IJI would be
impacted as a result of the conversion of cultivated crops is ambiguous. Welfare estimates in
Table 3.2 can thus be taken to be reflective of an “average” landscape parcel in which IJI does
not change with the conversion of the cultivated crops to herbaceous CRP. Additonal
examination revealed that a +/- one standard deviation change in IJI results in a change to
hunter welfare (relative to the results in Table 3.2) of approximately +/- 20%.
In Table 3.2, three assumptions on MPS and the assumption of constant IJI are used to
examine how economic benefits accruing to pheasant hunters from the MPRI vary across three
different restoration siting scenarios. The change in pheasant sightings in each landscape
(across the nine MPRI pilot focus areas) resulting from the conversion of 1600 acres of
agricultural land to herbaceous CRP was computed, and the landscape with the minimum,
median, and maximum pheasant sightings increases in each county was identified. In the first
scenario, “Minimum”, it is assumed that for each of the nine counties, the conversion occurs
within the landscape parcel in each county that realizes the smallest increase in pheasant
sightings as a result of the conversion. Essentially, the policy-relevant analogy is that the MPRI
managers selected the worst possible landscape parcel to locate the cooperative in each of the
nine counties. In the second scenario, “Median” it is assumed that the conversion occurs within

95

the landscape parcel in each county that realizes the median increase in pheasant sightings in
that county, and in the third scenario, “Maximum” it is assumed that conversion occurs within
the landscape parcel in each county that realizes the largest increase in pheasant sightings in
that county. In this “Maximum” scenario, the policy analogy is that MPRI managers have
complete freedom to locate the additional 1600 acres of CRP land in the landscape that
provides the greatest increase in pheasant sightings. The seasonal welfare per acre in Table 2 is
the seasonal welfare estimate in Table 3.2 divided by the total number of acres of agricultural
land converted to herbaceous CRP (14,400 acres) in the MPRI restoration siting scenario.

96

Table 3.2. Seasonal welfare and per-acre welfare of MPRI restoration siting scenarios
% Change in
Pheasant
Sightings in
Restoration Area

Seasonal
Welfare

Seasonal
Welfare per
acre

Welfare
Difference
(Median as
Baseline)

Minimum

-2.9%

$-900

-$0.06

-107%

Median

+24.9%

$12,700

$0.88

Baseline

Maximum

+52.9%

$30,100

$2.09

+ 137%

Minimum

-1.5%

-$400

-$0.03

-103%

Median

+27.4%

$13,900

$0.97

Baseline

Maximum

+63.0%

$34,200

$2.38

+145%

Minimum

-0.4%

-$100

-$0.01

-101%

Median

+33.0%

$14,900

$1.04

Baseline

Maximum

+69.7%

$37,000

$2.57

+148%

Restoration Siting
Scenarios

No Patch
Increase

50% New
Patches

All New
Patches

Table 3.2 illustrates the critical role restoration siting decisions play in generating
economic benefits to pheasant hunters. Choosing the least productive landscape in each
county for restoration of 1,600 acres of herbaceous CRP essentially results in zero economic
benefits to hunters from the 14,400 acres of restored land. For a different perspective,
interpret the “Median” scenario as being a situation in which a manager, facing no constraints
on restoration site selection but also possessing no information on the benefits of restoration
on different landscapes, selects the landscape with the median increase in economic benefit
from restoration in each county. Relative to this scenario, a manager able to identify and select
for restoration landscapes with the largest increase in economic benefits (the “Maximum”

97

scenario) in each county are able to achieve an increase in economic benefits to hunters of
about 140%.
Figure 3.4. Parcel-level economic benefit estimates from converting 1,600 acres of agricultural
land to CRP herbaceous land

The above analysis demonstrates the importance of restoration site selection in
generating economic benefits. To further illustrate, Figure 3.4 shows the economic benefits for
each landscape parcel within the primary and secondary MPRI restoration areas in Michigan’s
Souther Lower Peninsula. For each parcel 1,600 acres of the agricultural land within that parcel
was converted to CRP hebaceous land, under the assumption that 50% of the converted
patches are adjacent to existing CRP patches and 50% are new patches (holding all other
landscape characteristics within and across parcels constant). Restoration parcel hot spots
which produce economic benefits are scaled using shades of red – the darker the color, the
98

greater the economic benefit of restoring that parcel. Parcels which lacked sufficient
agricultural land for the restoration scenario (less than 1,600 acres) are indicated in grey. Some
parcels, shaded in blue, had small negative economic benefit estimates (mean losses for those
parcels = -$126; median = -$108) due to the non-linearities and spatial juxtiposition of
agricultural lands within the pheasant sightings equation (equation 8).

3.5 Discussion
The objective of this paper is to explore the extent to which restoration site selection
affects economic benefits for a user group whose outdoor recreational experience is directly
affected by ecosystem services generated through ecological restoration efforts. Examining the
linkages between CRP restoration, pheasant abundance, and willingness to pay for increases in
pheasant abundance by Michigan pheasant hunters, it is clear that the level of economic
benefits realized by pheasant hunters depend dramatically on restoration site selection. An
important factor affecting the levels of economic benefits across different possible restoration
sites is that the ecosystem service output desired by pheasant hunters (i.e., increases in
pheasant) is affected by a highly non-linear ecological production function. In particular,
diminishing returns to agricultural land results in the economic benefits from CRP restoration
(via the conversion of agricultural land to CRP land) being highly sensitive to the amount of
agricultural land within the landscape selected for restoration. In the extreme, the worst
restoration site selection scenarios generate effectively zero economic benefits. In contrast, an
informed resource manager can select restoration sites in a manner that increases the returns

99

on economic benefits by about 140% relative to a baseline scenario in which the manager
possesses no spatially explicit information. Such non-linear relationships between landscapes
and human well-being have also been identified in Kopmann and Rehdanz (2013). Additionally,
this result conforms with the finding of Newbold and Massey (2010), who showed that poorly
targeted environmental improvements can lead to welfare decreases. These analyses suggest
that managers should be particularly concerned with landscape heterogeneity and spatial
dependencies when the landscape characteristic targeted for restoration provides positive but
diminishing returns within the ecological production function.
The parcel-level economic benefits estimates illustrated in Figure 4 produce key
information for governmental and non-governmental organizations seeking to maximize return
on restoration dollars when a key objective is to increase benefits for pheasant hunters.
Restoration hotspots as indicated in Figure 3.4 align closely with the nine MPRI pilot focus area
counties (see Figure 3.3) targeted for initial restoration efforts, suggesting a well-targeted
restoration focus. Even within these nine counties there is evidence of hotspots which provide
the most economic benefits. Moreover, if restoration activities could target conversion of
agricultural lands to habitat types and locations that are even more beneficial to pheasants
than generic CRP herbaceous lands, then the economic benefits would be even greater.
While site selection for CRP restoration plays an important role in the relative amount of
economic benefits accruing to pheasant hunters, both the Michigan CRP as a whole as well as
the MPRI generate modest aggregate economic benefits to these hunters. A limiting and
conservative assumption within the MPRI restoration scenario analysis is that restoration

100

occurring through this program is through the conversion of agricultural land to generic
herbaceous CRP land. In contrast, restoration actions via the MPRI program would likely focus
on restoring distinct herbaceous CRP classification subcategories which are the most beneficial
to pheasant (and are not represented in the sightings model), such as upland bird habitat
buffers. Restoration efforts targeted at the most beneficial habitat types would result in
greater increases in pheasant abundance, and thus greater economic benefits to pheasant
hunters. An additional limitation of the survey data and analysis is that the total number of
pheasant hunter trips is held constant despite increases in the pheasant. This assumption
becomes less valid with large changes in pheasant populations. If such quality improvements
resulted in latent pheasant hunters rejoining the active hunter population or if active pheasant
hunters increased their number of hunting trips, the estimates within would understate
economic benefits.
Although methodological differences complicate direct comparison to the Hansen et al.
(1999) study, the temporal and regional differences between these studies help explain the
comparatively smaller economic benefits estimated in this analysis. For example, the number
of Michigan pheasant hunters has declined by 63% over the previous two decades (Moritz,
1992; Frawley, 2012) - scaling up the aggregate economic benefits estimates within this paper
under the assumption of a 1991 level of pheasant hunters yields a per-acre CRP value of about
$3.93/acre/year. Finally, opportunities and preferences for different types of hunting in
Michigan appear to have changed over the previous two decades, with increasing interest in
and availablility of deer hunting and sharply declining participation small game hunting
(Frawley, 2006). This substitution away from pheasant hunting and towards deer hunting
101

(especially archery hunting, for which the season is open throughout the Michigan pheasant
hunting season) likely results in lower willingness to pay for increases in pheasant hunting
quality relative to the past.
In summary, while this analysis reveals relatively modest economic benefits to pheasant
hunters from CRP land, the level of economic benefits from habitat restoration efforts varies
considerably depending on restoration site selection. These results illustrate how managers
can leverage spatially explicit ecological data and the site choices of outdoor recreationists to
enhance the return on conservation investments.

102

REFERENCES

103

REFERENCES

American Automobile Association (AAA). 2008. Your Driving Costs. AAA Association
Communication, Heathrow, Florida, USA.
Arbuckle, K.E., and J.A. Downing. 2002. Freshwater Mussel Abundance and Species Richness:
GIS Relationships with Watershed Land Use and Geology. Canadian Journal of Fisheries
and Aquatic Sciences 59: 310-316.
Association of Fish and Wildlife Agencies Resident Game Bird Working Group. 2013. National
Wild Pheasant Conservation Plan, ed Veverka, N.B., Indiana Division of Fish and
Wildlife, Bloomington, IN.
Barbier, E.B., 2012. A Spatial Model of Coastal Ecosystem Services. Ecological Economics 78:
70-79.
Barbier, E.B., E.W. Koch, B.R. Silliman, S.D. Hacker, E. Wolanski, J. Primavera, E.F. Granek et al.
2008. Coastal Ecosystem-Based Management with Nonlinear Ecological Functions and
Values. Science 319: 321-323.
Bernhardt, E.S., M.A. Palmer, J.D. Allan, G. Alexander, K. Barnas, S. Brooks et al. 2005.
Synthesizing US River Restoration Efforts. Science 308, 636–37.
Bestard, A.B., and A.R. Font. 2009. Environmental Diversity in Recreational Choice
Modelling. Ecological Economics 68: 2743-2750.
Carpenter, S.R., R. De Fries, T. Dietz, H.A. Mooney, S. Polasky, W.V. Reid, and R.J. Scholes. 2006.
Millennium ecosystem assessment: research needs. Science 314, 257-258.
Carpenter, S.R., H.A. Mooney, J. Agard, D. Capistrano, R.S. DeFries, S. Diaz, et al. 2009. Science
for Managing Ecosystem Services: Beyond the Millennium Ecosystem
Assessment. Proceedings of the National Academy of Sciences, USA 106: 1305-1312.
Claassen, R., A. Cattaneo, R. Johansson. 2008. Cost-Effective Design of Agri-Environmental
Payment Programs: US experience in Theory and Practice. Ecological Economics 65: 737752.
De Groot, R.S., M.A. Wilson, R.M. Boumans. 2002. A Typology for the Classification, Description
and Valuation of Ecosystem Functions, Goods and Services. Ecological Economics. 41:
393-408.
Fontana, V., A. Radtke, V.B. Fedrigotti, U. Tappeiner, E. Tasser, S. Zerbe, T. Buchholz. 2013.
Comparing Land-Use Alternatives: Using the Ecosystem Services Concept to Define a
Multi-Criteria Decision Analysis. Ecological Economics 93: 128-136.
104

Frawley, B. J., 2012. 2009 Small Game Harvest Survey. Michigan Department of Natural
Resources, Wildlife Division.
Frawley, B. J., 2006. Demographics, recruitment, and retention of Michigan hunters: 2005
update. Michigan Department of Natural Resources, Wildlife Division Report No. 3462.
Gibson, L.A., B.A. Wilson, and J.G. Aberton. 2004. Landscape Characteristics Associated with
Species Richness and Occurrence of Small Native Mammals Inhabiting a Coastal
Heathland: A Spatial Modelling Approach. Biological Conservation. 120: 75-89.
Gould, J.H., and K.J. Jenkins. 1993. Seasonal Use of Conservation Reserve Program Lands by
White-Tailed Deer in East-Central South Dakota. Wildlife Society Bulletin 21: 250-255.
Haab, T.C., and K.E. McConnell. 2002. Valuing Environmental and Natural Resources: The
Econometrics of Non-Market Valuation. Edward Elgar Publishing.
Hansen, L., P. Feather, and D. Shank. 1999. Valuation of Agriculture's Multi-Site Environmental
Impacts: An Application to Pheasant Hunting. Agricultural and Resource Economics
Review 28: 199–207.
Haines-Young, R, and M. Potschin. 2010. The Links Between Biodiversity, Ecosystem Services
and Human Well-Being. IN Ecosystem ecology: a new synthesis, eds Raffaelli D., and C.
Frid. British Ecological Society Ecological Review Series, pp.110–139. Cambridge
University Press, Cambridge.
Haroldson, K.J., R.O. Kimmel, M.R. Riggs, and A.H. Berner. 2006. Association of Ring-Necked
Pheasant, Gray Partridge, and Meadowlark Abundance to Conservation Reserve
Program Grasslands. Journal of Wildlife Management. 70: 1276-1284.
Heberling, M. T., and J.J. Templeton. 2009. Estimating the Economic Value of National Parks
with Count Data Models Using On-Site, Secondary Data: The Case of the Great Sand
Dunes National Park and Preserve. Environmental Management 43: 619–627.
Henderson, M.M., K.R. Criddle, and S.T. Lee. 1999. The Economic Value of Alaska's Copper River
Personal Use and Subsistence Fisheries." Alaska Fishery Resource Bulletin 6: 63-69.
Hicks, R., T.C. Haab, and D. Lipton. 2004. Recreational Fishing and the Benefits of Oyster Reef
Restoration in the Chesapeake Bay. Final Report Prepared for the Chesapeake Bay
Foundation.
Huang, B., C. Langpap, and R.M. Adams. 2012. The Value of In‐Stream Water Temperature
Forecasts for Fisheries Management. Contemporary Economic Policy 30: 247-261.

105

Jakus, P.M., M. Downing, M.S. Bevelimer, and J.M. Fly. 1997. Do Sportfish Consumption
Advisories Affect Reservoir Anglers' Site Choice? Agricultural and Resoure Economics
Review 26:196-204.
Johnson, N.S., and R.M. Adams. 1988. Benefits of Increased Streamflow: The Case of the John
Day River Steelhead Fishery. Water Resources Research 24: 1839-1846.
Johnson, T.E., J.N. McNair, P. Srivastava, and D.D. Hart. 2007. Stream Ecosystem Responses to
Spatially Variable Land Cover: An Empirically Based Model for Developing Riparian
Restoration Strategies. Freshwater Biology 52: 680-695.
Kantrud, H.A. 1993. Duck Nest success on Conservation Reserve Program Land in the Prairie
Pothole Region. Journal of Soil and Water Conservation 48: 238-242.
King, J.W., and J.A. Savidge. 1995. Effects of the Conservation Reserve Program on Wildlife in
Southeast Nebraska. Wildlife Society Bulletin 23: 377-385.
Knoche, S., and F. Lupi. 2013. Economic Benefits of Publicly Accessible Land for Ruffed Grouse
Hunters. Journal of Wildlife Management 77: 1294-1300.
Knoche, S., and F. Lupi. 2007. Valuing Deer Hunting Ecosystem Services from Farm
Landscapes. Ecological Economics 64: 313-320.
Kopmann, A., and K. Rehdanz. 2013. A Human Well-Being Approach for Assessing the Value of
Natural Land Areas. Ecological Economics 93: 20-33.
Leslie, H., M. Ruckelshaus, I.R. Ball, S. Andelman, and H.P. Possingha. 2003. Using Siting
Algorithms in the Design of Marine Reserve Networks. Ecological Applications 13: 185198.
Lipton, D.W., and R. Hicks. 1999. Linking Water Quality Improvements to Recreational Fishing
Values: The Case of Chesapeake Bay Striped bass. Fisheries Centre Research Reports
7:105-110, Vancouver, BC.
Lipton, D.W., and R. Hicks. 2003. The Cost of Stress: Low Dissolved Oxygen and Economic
Benefits of Recreational Striped Bass (Morone saxatilis) Fishing in the Patuxent
River. Estuaries 26: 310 315.
Maier, J.A., J.M. Ver Hoef, A.D. McGuire, R.T. Bowyer, L. Saperstein, and H.A. Maier. 2005.
Distribution and Density of Moose in Relation to Landscape Characteristics: Effects of
Scale. Canadian Journal of Forest Research 35: 2233-2243.
Massey, D.M., S.C. Newbold, and B. Gentner. 2006. Valuing Water Quality Changes Using a
Bioeconomic Model of a Coastal Recreational Fishery. Journal of Environmetal
Economics and Management 52: 482-500.

106

McGarigal, K., S.A. Cushman, and E. Ene. 2012. FRAGSTATS v4: Spatial Pattern Analysis Program
for Categorical and Continuous Maps. Computer software program produced by the
authors at the University of Massachusetts, Amherst. Available at the following website:
http://www.umass.edu/landeco/research/fragstats/fragstats.html
Lesta, M. 2009. Evaluation of Regulation Functions of Rural Landscapes for the Optimal Siting of
Treatment Wetlands and Mitigation of Greenhouse Gas Emissions (Doctoral
dissertation).
Luukkonen, D.R. 1998. Pheasant Population Status in Michigan, 1998. Wildlife Division Report
No. 3284. Michigan Department of Natural Resources, Lansing, Michigan, USA.
Michigan Department of Natural Resources (MDNR). 2009. 2008–2009 Michigan Upland Game
Harvest Survey.
Michigan Department of Natural Resources (MDNR). 2008. 2008 Michigan Hunting and
Trapping Digest.
Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-being: Synthesis. Island
Press, Washington, DC.
Moreno-Mateos, D., and F.A. Comín. 2010. Integrating Objectives and Scales for Planning and
Implementing Wetland Restoration and Creation in Agricultural Landscapes. Journal of
Environmental Management 91: 2087-2095.
Moritz, W.E. 1992. Hunting Results, Michigan Small Game Seasons, 1991. Michigan
Department of Natural Resources. Wildlife Division Report No. 3202.
Newbold, S.C., and D.M. Massey. 2010. Recreation Demand Estimation and Valuation in
Spatially Connected Systems. Resource and Energy Economics 32: 222-240.
Nielson, R.M., L.L. McDonald, J.P. Sullivan, C. Burgess, D.S. Johnson, D.H. Johnson, S. Bucholtz,
S. Hyberg, and S. Howlin. 2008. Estimating the Response of Ring-Necked Pheasants
(Phasianus colchicus) to the Conservation Reserve Program. The Auk 125: 434-444.
Oetter, D.R., L.R. Ashkenas, S.V. Gregory, and P.J. Minear. 2004. GIS Methodology for
Characterizing Historical Conditions of the Willamette River Flood Plain,
Oregon. Transactions in GIS 8: 367-383.
Parsons, G.R. 2003. The Travel Cost Model. In P. A. Champ, K. J. Boyle, and T. C. Brown, eds. A
primer on nonmarket valuation, pp 427–440. Kluwer Academic Publishers, Dordrecht,
The Netherlands.

107

Pess, G.R., D.R. Montgomery, E.A. Steel, R.E. Bilby, B.E. Feist, and H.M. Greenberg. 2002.
Landscape Characteristics, Land Use, and Coho Salmon (Oncorhynchus kisutch)
Abundance, Snohomish River, Wash., USA. Canadian Journal of Fisheries and Aquatic
Sciences 59: 613-623.
Radford, J.Q., A.F. Bennett, and G.J. Cheers. 2005. Landscape-Level Thresholds of Habitat Cover
for Woodland-Dependent Birds. Biological Conservation 124: 317-337.
Reynolds, R.E., T.L. Shaffer, R.W. Renner, W.E. Newton, and B.D. Batt. 2001. Impact of the
Conservation Reserve Program on Duck Recruitment in the US Prairie Pothole Region.
Journal of Wildlife Management 65: 765-780.
Reynolds, R.E., T.L. Shaffer, J.R. Sauer, and B.G. Peterjohn. 1994. Conservation Reserve
Program: Benefit for Grassland Birds in the Northern Plains. Transactions of the North
American Wildlife and Natural Resources Conference. 59: 328-336.
Riffell, S., D. Scognamillo, and L.W. Burger. 2008. Effects of the Conservation Reserve Program
on Northern Bobwhite and Grassland Birds. Environmental Monitoring and
Assessment 146: 309-323.
Riley, T.Z. 1995. Association of the Conservation Reserve Program with Ring-Necked Pheasant
Survey Counts in Iowa. Wildlife Society Bulletin 23: 386-390.
Sanchirico, J.N., and P. Mumby. 2009. Mapping Ecosystem Functions to the Valuation of
Ecosystem Services: Implications of Species–Habitat Associations for Coastal Land-Use
Decisions. Theoretical Ecology 2: 67-77.
Sauer, J.R., J.E. Hines, J.E. Fallon, K.L. Pardieck, D.J. Ziolkowski, Jr., and W.A. Link. 2011. The
North American Breeding Bird Survey, Results and Analysis 1966-2010. Version
12.07.2011, USGS Patuxent Wildlife Research Center, Laurel, MD.
Schroeder, M.A., and W.M. Vander Haegen. 2011. Response of Greater Sage-Grouse to the
Conservation Reserve Program in Washington State. Studies in Avian Biology 38: 517530.
StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP.
Suding, K.N. 2011. Toward an Era of Restoration in Ecology: Successes, Failures and
Opportunities Ahead. Annual Review of Ecology, Evolution, and Systematics 42: 465–
487.
Swinton, S.M., F. Lupi, G.P. Robertson, and D.A. Landis. 2006. Ecosystem Services and
Agriculture: Cultivating Agricultural Ecosystems for Diverse Benefits. Ecological
Economics 64: 245-252.

108

United States Department of Agriculture (USDA) Farm Service Agency. 2014. Conservation
Reserve Program – Rental payments by Fiscal Year. Retrieved from:
http://www.fsa.usda.gov/FSA/webapp?area=home&subject=copr&topic=rns-css
White, D., and S. Fennessy. 2005. Modeling the Suitability of Wetland Restoration Potential at
the Watershed Scale. Ecological Engineering. 24: 359-377.
Whitehead, J.C., B. Poulter, C.F. Dumas, and O. Bin. 2009. Measuring the Economic Effects of
Sea Level Rise on Shore Fishing. Mitigation and Adaptation Strategies for Global
Change. 14: 777-792.
Young, T.P. 2000. Restoration Ecology and Conservation Biology. Biological Conservation. 92:
73-83.

109

CHAPTER 4: A DISCRETE CHOICE EXPERIMENT APPROACH EXAMINING TROUT ANGLER
PREFERENCES FOR FISHING SITE ATTRIBUTES

4.1 Introduction
Trout fishing is a popular outdoor recreational activity in the U.S., with about 7.2 million
anglers spending 76 million days pursuing species such as brown trout (Salmo Trutta), brook
trout (Salvelinus fontinalis), rainbow trout (Onchorhynchus mykiss), and cutthroat trout
(Onchorhynchus mykiss) in U.S. waters (excluding the Great Lakes) in 2011 (USDOI 2013). To
generate and sustain high-quality trout fishing experiences, fisheries managers undertake costly
management actions such as trout stocking, pollution mitigation, and stream and riparian
restoration. Further, fisheries managers seek to prevent the over-exploitation of trout fisheries
through the implementation of site and season-specific regulations which restrict trout harvest
and permissible fishing gear. Achieving the improvements in the fishery desired by anglers
while both staying within budget and being cognizant of angler tolerance for regulations
requires an in-depth understanding of the types of trout fishing experiences preferred by
anglers. Characterizing mean angler preferences for trout fishing site attributes in terms of the
willingness to incur increased costs (such as increased travel distance to reach a fishing
destination) to obtain fishing site quality improvements provides a convenient and relevant
metric for comparing angler preferences for different site attributes. Moreover, management
decision-making can be greatly aided by information on how these preferences for fishing site
attributes vary across the population of anglers. For example, fisheries managers seeking to
achieve catch-related improvements through the implementation of more restrictive
110

regulations such as higher minimum size, lower daily harvest limit, and restrictions on certain
types of fishing gear would do well to understand how heterogeneous angler preferences for
regulations affect whether and to what extent anglers are made better off via regulationinduced catch-related improvements.
To examine angler preferences and preference heterogeneity for fishing site attributes, I
use the survey-based stated preference discrete choice experiment approach. This approach
allows for the identification of angler preferences through observing the choice of a preferred
fishing site from a set of two hypothetical fishing sites, with each site consisting of different
catch rates, regulations, and travel distance. Angler choice of a fishing site is statistically
analyzed via the mixed logit model, which produces both mean and standard deviation
parameter estimates that characterize the distribution of preferences for fishing site attributes
across the angling population. This provides insight into angler preference heterogeneity for
fishing site attributes, thus providing an additional dimension of analysis relative to the widely
used conditional logit model which only provides information on mean preferences. Mean
parameter estimates are used to examine the willingness of anglers to accept greater travel
distances in exchange for changes in catch and regulations at trout fishing locations. Standard
deviation parameter estimates are used in conjunction with mean parameter estimates to
characterize the distribution of angler preferences with respect to fishing regulations.
Subsequently, I examine the extent to which trout anglers switch from being negatively
impacted by trout regulations to being positively impacted when these regulations generate
catch-related improvements in the trout fishery. Finally, I describe the potential use of these
results in guiding fisheries management decisions.
111

4.2 Study Area
There are approximately 12,500 miles of coldwater trout streams in Michigan, with the
majority of these waterways distributed throughout Michigan’s Upper Peninsula and Northern
Lower Peninsula (see Figure 4.3). Trout fishing opportunities are more limited in the southcentral and southeastern portions of the Lower Peninsula. However, this region is the most
populous, with the nine-county Detroit-Ann Arbor-Flint combined statistical area constituting
more than half of the state’s total population. Brown Trout and Brook Trout comprise the
majority of the wild, non-anadromous trout population, though there are some localized, nonanadromous populations of wild Rainbow Trout (Bryan Burroughs, pers. comm.).
Throughout these coldwater trout fishing streams, fisheries managers utilize a variety of
regulations to protect and enhance trout fishing experiences. The Michigan Department of
Natural Resources (MDNR) classifies the majority of Michigan trout streams (~1,600 streams)
into four different regulations categories, with each category having a regulations regime which
can vary by fishing season, possession season, and minimum size lengths.

112

Figure 4.1. Major categories of Michigan trout fishing regulations

Source: 2013 Michigan Fishing Guide (Michigan Department of Natural Resources)

Per Figure 4.1, trout fishing and trout possession seasons are either a) the last Saturday
in April – September 30, or b) open all year. Minimum size regulations for brook trout (brown
trout) range from 7” to 15” (8” to 15”). At all locations within the possession season, the daily
harvest limit is five trout and salmon with only three trout greater than 15”.
On some sections of trout streams, however, the MDNR has implemented special
regulations restricting the use of natural bait/artificial lures and the number of trout that may
be harvested. On three stream sections totaling 21.8 miles, catch and release is required during
the standard trout fishing season (i.e., last Saturday in April through September 30). A two
113

trout harvest limit is imposed on ten sections of stream totaling 105.4 miles. Michigan trout
anglers also face restrictions on the types of lures/baits that may be used on certain portions of
streams. Trout anglers fishing in eight stream sections totaling 78.2 miles are restricted to the
use of artificial flies, while anglers fishing in seven stream sections totaling 53.4 miles are
restricted to using artificial lures (including artificial flies). Additionally, five stream sections
totaling 34.6 miles have restrictions on lures/bait which vary throughout the year.
Special harvest and gear restrictions during the standard fishing season occur on about
260 miles of trout streams, only about 2% of the 12,500 miles of coldwater trout streams in
Michigan. The most widespread regulation, mandatory catch and release for brown trout and
brook trout outside of the standard trout fishing season, occurs on about 8% of trout streams in
Michigan. However, these regulations exist on highly regarded trout streams which receive
substantial fishing pressure, such as the Au Sable, Manistee, and Pere Marquette rivers. This
includes an 8.7 mile stretch of the Au Sable referred to many fly anglers as the “Holy Waters”
(Bain 1987, Gigliotti 1989), a highly regulated (artificial flies only and catch & release only)
portion of the Au Sable renowned for its trout fishing.
Privately and publicly funded stocking and habitat restoration efforts are used to
enhance trout fisheries in Michigan. Excluding Great Lakes and Great Lakes connecting
waterways (i.e., Detroit River and St. Clair River), about 1.5 million brown trout and about
100,000 brook trout were stocked in rivers/streams in Michigan in 2012 (MDNR 2012). The
MDNR Fisheries Division undertakes numerous habitat restoration projects that benefit
coldwater species such as trout – in 2011 the MDNR Fisheries Division reported carrying out 17

114

dam removal projects resulting in 167 miles of rehabilitated stream habitat, and ten additional
restoration projects which reconnected and improved 161 miles of streams and rivers (MDNR
2011). The MDNR Fisheries Division also manages a habitat restoration and research fund
which is funded from penalties levied on hydropower operators due to damages from dam
operation. This fund supports habitat restoration and research efforts on major trout fishing
rivers/streams in Michigan such as the Au Sable and Manistee rivers. In 2011, these mitigation
funds helped fund the removal of six dams, facilitated three woody habitat improvement
structures, and helped reduce sedimentation via two road-stream crossing projects (MDNR
2011).

4.3 Material and Methods
4.3.1 Discrete Choice Method
Discrete choice methods employed within a recreational fisheries context utilize
information on an angler’s choice of a fishing site from a set of potential fishing sites to identify
angler preferences for fishing site attributes. This information on angler preferences may be
obtained via revealed preference approaches in which actual or reported fishing site choices of
anglers are used, or through stated preference survey-based approaches in which individuals
are presented with multiple hypothetical fishing sites within a choice scenario and asked to
choose a preferred alternative. Early discrete choice modeling efforts focusing on recreational
fishing used revealed preference methods (Milon 1988a, Milon 1988b, Bockstael et al. 1989),
and until recently, a large majority of peer-reviewed research consisted of revealed preference
115

approaches (Hunt 2005). There have been numerous revealed preference discrete choice
studies which examine aspects of coldwater trout and salmon fishing experiences, with several
studies specifically focusing on the site choice of stream trout anglers (see, e.g., Ahn et al. 2000,
Morey et al. 2002, and Timmins and Murdock 2007).
The reliance on individuals’ observed or reported behavior is an attractive feature of
revealed preference approaches for many researchers. However, it may be difficult and costly
to obtain reasonably precise information on the location of trout angler fishing trips. Further,
even if information on trout angler site choice exists or can readily be obtained, site attribute
tradeoff analysis requires the measurement of site-specific attributes such as angler catch rate.
In many cases insufficient data exists to quantify angler catch rate or other attributes at
potential fishing locations. Such measurement problems can be mitigated through surveybased stated preference approaches which elicit angler preferences through choice scenarios
which explicitly and quantitatively define attributes levels available at fishing site alternatives.
Stated preference discrete choice approaches may also be advantageous if fishing site
attributes levels are correlated across fishing sites, which can make it difficult to identify
relationships between attribute levels and site choice, or if there is an interest in examining
angler preferences for attributes (or attribute levels) that currently do not exist at fishing sites.
There has been substantial growth in the use of stated preference discrete choice modeling
efforts to examine angler preferences for recreational fishing attributes. Whereas the Hunt
(2005) study identified three stated preference approaches which examine aspects of
recreational fishing, I identified over 20 peer-reviewed research papers using the stated
preference approach since Hunt’s 2005 publication. Three research efforts have focused
116

specifically on the stream trout fishing experience. Ahn et al. (2000) examined angler
preferences for catching brown trout and grayling in Norway, while Beville and Kerr (2008) and
Beville and Kerr (2009) examined angler preferences and preference heterogeneity for New
Zealand stream trout fishing attributes.

4.3.2 Stated Preference Discrete Choice Survey
A stated preference discrete choice experiment of trout anglers in Michigan was
developed and implemented to examine angler preferences for aspects of a stream trout
fishing experience. The choice experiment was a component of a larger survey which examined
Michigan coldwater angler opinions, preferences, and participation, as well as associated
economic impacts of coldwater fishing in Michigan (see Appendix B for survey). To filter survey
respondents with a history of trout fishing in Michigan to the choice experiment section of the
survey, the following skip question was used (also see Appendix B, Figure B.2.):

“During the previous 5 years, have you fished for Brown Trout and/or Brook Trout in a Michigan
river or stream?”

Individuals who responded “No” to this question were skipped past all choice experimentsupporting questions, including warm up questions (Appendix B, Figure B.3.), information
treatments (Appendix B, Figure B.4.), and trout fishing-related supporting questions (Appendix
B, Figure B.9.). Respondents who answered “Yes” to this question were directed to the choice
experiment (and the aforementioned choice experiment-related questions), which consists of
117

four choice scenarios with each scenario consisting of two stream trout fishing locations
described by the attribute levels available at each location. These fishing site locations were
described by the same seven attributes listed in Figure 4.2, while the attribute levels varied
across different fishing site options. For each choice scenario, the survey respondent was asked
to first compare the different fishing site options by the attribute levels associated with each
option, and then identify the location he/she would prefer to go fishing at. Figure 4.2 is an
example of a choice scenario that was included in the survey7.

7

The survey is a split-sample survey which presented one-half of the sample with the choice
experiment binary format seen in Figure 4.2., and one-half of the sample with a choice
experiment ranking format seen in Appendix D, Figure D.1. Each format presents two trout
fishing site options described by fishing site attributes and varying attribute levels; however,
the ranking format includes a third option “go fishing at another location”, and asks individuals
to rank these three options within this format. For this analysis, we only consider angler
preferences with respect to the two described trout fishing sites for the binary and ranking
formats.
118

Figure 4.2. Example Michigan stream trout fishing site choice scenario

The process of identifying management-relevant attributes for use in describing
Michigan stream trout fishing locations began with a comprehensive review of peer-reviewed
literature, grey literature, and additional materials which could provide insight into important
aspects of the recreational fishing experience. Reviewing 50 discrete choice analyses of
recreational fishing, Hunt (2005) found the following site attributes to influence an individual’s
fishing site selection decision; a) costs relating to travel or access, b) fishing quality (e.g., size,
catch rate, and species availability), c) environmental quality, d) facility development, e)
119

encounter levels (with anglers and other individuals), and f) regulations. These attributes were
used as starting points in the search for management-relevant attributes which might influence
stream trout fishing site choice in Michigan. The 2013 Michigan Fishing Guide (MDNR 2013)
was useful in understanding the different types of regulations faced by Michigan trout anglers.
Discussions with personnel from the MDNR and Trout Unlimited (a coldwater fisheries NGO)
helped identify fishing site attributes which might influence a trout angler’s selection of a
fishing site and also provided guidance regarding realistic attribute levels for different fishing
site attributes. Ultimately, seven stream trout fishing site attributes were identified as being
management-relevant and potentially important drivers of fishing site choice. These attributes
and the range of potential attribute levels are detailed in Table 4.1 below.

120

Table 4.1. Fishing site attribute levels for choice experiment
# OF
CATEGORY ATTRIBUTE
ATTRIBUTE LEVELS
LEVELS
Regulations
Catch and Release
Possession
3
2 Trout
Limit
5 Trout
None
Lure/Bait
3
Artificial Lures & Flies Only
Restrictions
Flies Only
Brown
25% under 7”,50% 7” to 9”, 25% over 9”
Trout Catch
25% under 7”, 50% 7” to 11”, 25% over 11”
Attributes
25% under 7”, 50% 7” to 15”, 25% over 15”
Typical
25% under 8”, 50% 8” to 10”, 25% over 10”
Angler
9
25% under 8”, 50% 8” to 12”, 25% over 12”
Catch Size
25% under 8”, 50% 8” to 16”, 25% over 16”
25% under 10”, 50% 10” to 12”, 25% over 12”
25% under 10”, 50% 10” to 14”, 25% over 14”
25% under 10”, 50% 10” to 18”, 25% over 18”
Typical
1 trout per 15 minutes
1 trout per 2 hours
Angler
6
1 trout per 30 minutes
1 trout per 4 hours
Catch Rate
1 trout per hour
1 trout per 8 hours
Brook Trout
25% under 6”, 50% 6” to 7”, 25% over 7”
Catch
25% under 6”, 50% 6” to 8”, 25% over 8”
Attributes
25% under 6”, 50% 6” to 11”, 25% over 11”
Typical
25% under 7”, 50% 7” to 8”, 25% over 8”
Angler
9
25% under 7”, 50% 7” to 9”, 25% over 9”
Catch Size
25% under 7”, 50% 7” to 12”, 25% over 12”
25% under 8”, 50% 8” to 9”, 25% over 9”
25% under 8”, 50% 8” to 10”, 25% over 10”
25% under 8”, 50% 8” to 13”, 25% over 13”
Typical
1 trout per 15 minutes
1 trout per 2 hours
Angler
6
1 trout per 30 minutes
1 trout per 4 hours
Catch Rate
1 trout per hour
1 trout per 8 hours
Travel
10
60
100
20
70
110
Travel
Region 1
35
Region 2 85
Region 3 125
6
Distance
(miles)
50
(miles)
100 (miles)
140
75
125
165
100
150
190

121

Most anglers living in the south-central or southeastern portion of Michigan’s Lower
Peninsula do not have stream trout fishing locations within close proximity to their residence.
To ensure that travel distance levels in the choice experiment reflected realistic travel distances
an angler might face on an actual trout fishing trip, anglers were separated into three distinct
geographic regions by region of residence (see Figure 4.3.). Trout anglers residing in the
Northern Lower Peninsula or the Upper Peninsula of Michigan likely have nearby stream trout
fishing options, and thus might see travel distances as low as 10 miles in choice scenarios,
whereas anglers living in southeastern Michigan likely have few or no nearby stream trout
fishing options, and thus do not see travel distances less than 100 miles.

122

Figure 4.3. Coldwater trout streams in Michigan

Experimental design procedures allow for the control of factors within a choice
experiment which affect parameter estimation, model flexibility, and the statistical efficiency of
estimated parameters (Johnson et al. 2007). The software package NGENE 1.0 (Choice Metrics)
was used to generate stream fishing site choice sets, with each choice set consisting of two
fishing site alternatives which vary by attribute levels. To obtain an experimental design
consisting of trout fishing site choice sets which enable the estimation of parameters with the
lowest possible standard errors, an algorithm was used to search for an experimental design
which would minimize D-error. Insights into likely signs of attribute parameters were sought
from the published literature and directional priors were incorporated into the design search to
improve the efficiency of the experimental design. Hunt et al. (2005) found that parameters on
cost and gear restrictions within a discrete choice recreational fishing context were mostly

123

negative, while parameters on catch attributes such as catch size and catch rate were generally
positive. These findings were followed with respect to assigning priors for use in the design
search. A total of seventy-two choice scenarios were identified by NGENE for use in the
surveys.8
To evaluate angler comprehension of the survey prior to actual survey implementation,
individuals with a history of fishing in Michigan were recruited to participate in one-hour survey
pretesting sessions which incorporated internet meeting software along with phone discussion
(following the approach of Weicksel [2012]). GatherPlace® screen sharing software was used
which allowed individuals to take the survey while progress was monitored visually from a
remote location. A phone connection was maintained throughout the process to address
immediate comments, questions or concerns an individual might have regarding specific
aspects of the survey instrument. A thorough assessment of respondent comprehension
occurred after the survey was completed, with each individual asked a series of questions
designed to identify potential issues with survey instrument design or content. Survey
instrument pretesting consisted of 15 anglers associated with the Trout Unlimited organization
and seven individuals recruited for a previous Michigan fishing survey - a total of 22 people
completed the survey pretesting sessions. Individuals from Trout Unlimited were not
compensated for their time, whereas the 7 other individuals were each paid $20. Though no
major issues were identified in the pretesting process, helpful comments and suggestions were
received which facilitated the improvement of various aspects of the survey layout and design.
8

See Appendix F for attribute levels for each choice scenario, Appendix G for Pearson Product
Moment Correlations for attributes within and across choice scenarios, and Appendix H for the
choice scenarios in the different 120 mail survey versions and 108 internet survey versions.
124

The Coldwater Resources Steering Committee of the MDNR, which includes various
representatives of the agency and the angling public, also reviewed a draft survey and their
feedback was incorporated into the final version of the questionnaire.
Survey recipients consisted of a randomly drawn sample of 4,099 Michigan residents
who purchased an “Unrestricted” fishing license (which allows for the holder to target trout and
salmon) for the 2012-2013 Michigan fishing season. A mixed mode internet and mail survey
approach was used with up to four contacts for each potential respondent (see Appendix A for
description of mailings timeline). The first three contacts consisted of an envelope with letter
(first contact) and two postcards (second and third contacts) which explained the nature and
purpose of the survey and provided a website address which the individual could type into their
web browser to access the survey (see Appendix D). Individuals who did not attempt to
complete the online version of the survey were sent a fourth contact, which included both the
website address as well as a printed 12-page questionnaire (see Appendix B for survey and
Appendix D for 4th mailing contact letter). A total of 1038 individuals completed the internet
survey and 743 individuals completed the mail survey, for a total response of 1781 individuals
(response rate of 44.6%; see Appendix A, Table A.1. for disposition table further detailing
survey response). The chi-square procedure was used to test for statistically significant
differences (P<0.01) in response rates by age, gender, and region of residence. Statistically
significant (P<0.01) differences in response rate were found for age, with the trend of response
rate increasing as age increased. In model estimation, survey results were weighted
accordingly (see Appendix A, Table A.2. for weights).

125

4.3.3 Random Utility Theory and Model Estimation
Random utility theory (McFadden 1974) is the underlying behavioral theory supporting
this discrete choice analysis of stream trout fishing site selection. According to random utility
theory, an individual selects the fishing location which maximizes his/her indirect utility. In
equation 1, 𝑉𝑖𝑗 is the indirect utility associated with individual i at site j, 𝑥𝑖𝑗 is a vector of site
attributes for individual i at site j, 𝛽 is a vector of parameters associated with these site
attributes, and 𝜀𝑖𝑗 is the stochastic error term unobservable to the researcher.
1)

𝑉𝑖𝑗 = 𝑉̅ (𝑥𝑖𝑗 , 𝛽) + 𝜀𝑖𝑗

The indirect utility function in equation 1 consists of the fishing site attribute variables defined
and described in Table 4.2 below.

126

Table 4.2. Fishing site attributes variables in indirect utility function
Fishing Site
Fishing Site Attribute
Variable
Attribute Variables
Variable Definition
Type

Min, Max

Distance

One-way distance from
individual’s residence (in miles)

Continuous

10, 190

Catch&Release

All trout caught must be released

Categorical

0,1

Harvest_2

Harvest limit of two trout

Categorical

0,1

Lures&Flies_Only

Only lures and flies may be used;
no natural bait

Categorical

0,1

Flies_Only

Only flies may be used; no other
lures or natural bait

Categorical

0,1

Brown_Rate

# of Brown Trout caught per hour
by a Typical Angler

Continuous

0.125, 4

Brook_Rate

# of Brook Trout caught per hour
by a Typical Angler

Continuous

0.125, 4

Brown_25

25th percentile of Brown Trout
catch length (in inches)

Continuous

7,10

Brown_75

75th percentile of Brown Trout
catch length (in inches)

Continuous

9,18

Brook_25

25th percentile of Brook Trout
catch length (in inches)

Continuous

6,8

Brook_75

75th percentile of Brook Trout
catch length (in inches)

Continuous

7,13

4.3.4 Mixed Logit Model
The mixed logit modeling approach accounts for preference heterogeneity of individuals
via the estimation of parameter distributions for specified random variables (Train 2009).
While the mixed logit has been applied to discrete choice analysis since the early 1980s (see
Boyd and Mellman 1980, Cardell and Dunbar 1980), the first application to recreational fishing
127

was by Train (1998), who examined angler preferences and estimated economic benefits
associated with trout fishing in Montana rivers. Other mixed logit modeling approaches
examining preferences for angling site attributes include McConnell and Tseng (1999), Breffle
and Morey (2000), Provencher and Bishop (2004), Massey et al. (2006), Oh and Ditton (2006)
and Wielgus et al. (2009). Kerkvliet and Nowell (2000) use a mixed logit model to examine
angler preferences and demand for stream trout fishing experiences in Yellowstone National
Park.
In the mixed logit model, the probability that an angler 𝑖 chooses site 𝑗 is estimated by
integrating standard logit probabilities (McFadden 1974) over a density of 𝛽 parameters (Train
2009).

2)

𝑃𝑖𝑗 = ∫

exp(𝑉𝑖𝑗 )
𝑓(𝛽)𝑑(𝛽)
exp(𝑉𝑖𝐴 ) + exp(𝑉𝑖𝐵 )

Mixed logit estimation proceeds with the objective of estimating mean and standard
deviation parameters for the parameters specified to have a random distribution. To aid in the
calculation of willingness to pay the parameter on distance is assumed to be fixed (Revelt and
Train 2000, Eggert and Olsson 2009); parameters on all other attributes are assumed to be
random. The researcher makes distributional assumptions for random parameters – commonly
assumed distributions include normal and lognormal distributions. Random parameters in this
model are assumed to be normally distributed with mean b and standard deviation s
characterizing the probability distribution𝑓(𝛽). Parameters are estimated via maximum
128

simulated likelihood estimation, in which mixed logit probabilities in equation 2 are
approximated by repeating draws of a value 𝛽 from 𝑓(𝛽|𝜃), where 𝜃 refers to the mean and
covariance of the distribution (Train 2009).

4.3.5 Estimating Willingness to Pay and Preference Heterogeneity
Once the mixed logit model has been used to estimate mean and standard deviation
parameters for fishing site attributes, I examine the willingness of stream trout anglers to make
tradeoffs between the distance traveled to reach a fishing site and the attributes available at
that fishing site. To do this, equation 3 below is used to estimate the mean willingness to pay
of trout anglers to obtain different levels of fishing site attributes.

3)

WTPmiles = ─ 𝑏̂𝑞 /𝑏̂𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒

The estimated mean parameter 𝑏̂𝑞 in equation 3 refers to a fishing site attribute other
than distance, and 𝑏̂𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 is the estimated parameter associated with the distance attribute.
This WTPmiles metric reflects the willingness of trout angler to drive further distances to obtain
quality improvements at a trout fishing site. For example, if 𝑏̂𝑞 = hourly catch rate = 5 and
𝑏̂𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = ─ 1, an individual would be willing to drive up to an additional 5 miles to obtain an
increase in the hourly catch rate by one fish per hour.

129

Finally, when the mean and standard deviation on the normally distributed random
variables are both statistically significant (P<0.05), the proportion of individuals who derive
positive and negative utility from these site attributes is calculated. Subsequently, the change
in the proportion of individuals deriving positive and negative utility from a regulation is
estimated given a regulation-induced improvement in catch-related attributes such as catch
size and catch rate.

4.4 Results
Table 4.3 provides mixed logit model results for the stream trout fishing choice model.
As expected, the Distance coefficient is negative and statistically significant, meaning that the
utility an individual receives from a site decreases as the distance to the site increases. As the
parameter on Distance is assumed to be fixed, there is a single mean parameter estimate for
this variable. For the other variables, normal distributions are assumed with mean and
standard deviation coefficients estimated. Mean coefficient estimates for the most restrictive
regulation levels in the gear and harvest regulation categories were also negative and highly
significant. All else equal, trout anglers are less likely to select fishing sites where trout cannot
be harvested (Catch&Release) relative to sites where there is a harvest limit of five trout.
Similarly, sites where only artificial flies may be used (Flies_Only) are less likely to be chosen
than sites where there are not restrictions on natural bait or artificial lures. Restricting anglers
to harvesting two trout (Harvest_2) or prohibiting the use of natural bait (Lures&Flies_Only) are
both negative - Harvest_2 is statistically significant at the 5% level, whereas Lures&Flies_Only is
not statistically significant at conventional levels of measurement. As expected, anglers prefer
130

to fish at locations where catch rates are higher - this result is statistically significant (P<0.01)
for both brown and brook trout catch rates. While anglers prefer higher catch rates of both
trout species, the mean coefficient on Brown_Rate is close to double that of the mean
coefficient on Brook_Rate, indicating a stronger preference for catching brown trout. On
average, anglers do not appear to have strong preferences for increasing the first quartile of
trout size for either species; mean coefficients on Brown_25 and Brook_25 are negative but not
statistically significant at conventional measurement levels. Finally, results indicate a
preference for larger brown trout size relative to larger brook trout size.
The standard deviation coefficients in Table 4.3 identify angler preference heterogeneity
associated with a fishing site attribute. The estimated standard deviations for Catch&Release
and Flies_Only regulations are statistically significant (P<0.01), indicating heterogeneous
preferences for these regulations across the population of stream trout anglers in Michigan.
Statistically significant (P<0.01) preference heterogeneity is identified for both Brown_25 and
Brook_25 variables, while the null hypothesis that mean coefficients for Brown_25 and
Brook_25 are equal to zero cannot be rejected. The combination of a statistically significant
standard deviation and the inability to reject the null hypothesis that the mean is different than
zero suggests that the preferences for changes in the 25 th percentile size vary throughout the
angler population.

131

Table 4.3. Mixed logit model results for Michigan trout fishing choice experiment
Robust
Coefficient
P-Values
Standard Error
Mean
Distance (Fixed)

-0.013

0.002

P<0.01

Catch&Release

-1.759

0.192

P<0.01

Harvest_2

-0.235

0.112

P=0.04

Lures&Flies_Only

-0.170

0.105

P=0.11

Flies_Only

-1.064

0.155

P<0.01

Brown_Rate

0.224

0.034

P<0.01

Brown_25

-0.020

0.044

P=0.65

Brown_75

0.060

0.017

P<0.01

Brook_Rate

0.120

0.034

P<0.01

Brook_25

0.024

0.067

P=0.73

Brook_75

0.040

0.028

P=0.15

Catch&Release

1.570

0.261

P<0.01

Harvest_2

0.014

0.071

P=0.84

Lures&Flies_Only

0.078

0.084

P=0.35

Flies_Only

1.409

0.264

P<0.01

Brown_Rate

0.074

0.137

P=0.59

Brown_25

0.321

0.099

P<0.01

Brown_75

0.002

0.008

P=0.83

Brook_Rate

0.149

0.136

P=0.28

Brook_25

0.521

0.161

P<0.01

Brook_75

0.176

0.073

P=0.02

Standard Deviation

Model Statistics
Log-Likelihood
Wald chi2(21)
Observations

1572.899
140.84
634 individuals; 2535 choice occasions

132

The WTP figures in Table 4.4 are interpreted as the number of miles an individual would
be willing to travel to obtain a change in fishing site quality attribute. For the two most
restrictive regulations, there are very large negative estimates of mean WTP, implying that
individuals would travel large distances to avoid these regulations. With respect to brown trout
catch-related attributes, an angler would be willing to travel an additional 16.6 miles to
increase trout catch rate by one brown trout per hour, and travel an additional 4.5 miles to
increase the length of each brown trout caught by 1”. Together, the WTP for Brown_Rate and
Brown_75 indicate that an angler would be roughly indifferent to a change in brown trout catch
of one per hour and a change in the length of each brown trout caught of 3.5”. Regarding
brook trout catch attributes, Table 4.4 shows that the average angler would be willing to drive
an additional 8.9 miles to catch one more brook trout every hour. Comparing WTP for both
brown trout and brook trout attributes, it appears that anglers have a stronger preference for
increases in brown trout attributes relative to brook trout attributes.

133

Table 4.4. Per-trip willingness to pay for statistically significant (P<0.05) fishing site attribute
variables
Fishing Site Attribute Variables

WTP with respect to
Distance (in miles)

Catch&Release

-130.4

Harvest_2

-17.5

Flies_Only

-78.9

Brown_Rate
(Catch increase of 1 Brown
Trout/hour)

16.6

Brown_75
(Increase in 4th quartile Brown
Trout catch size by 1”)

4.5

Brook_Rate
(Catch increase of 1 Brook
Trout/hour)

8.9

4.4.1 Management Regulation Policy Scenarios
The proportion of the angling population supporting restrictive regulations such as
mandatory catch & release and artificial flies only may be influenced by the effectiveness of
these regulations in improving catch-related angling outcomes. To understand the potential for
changes in angler support for regulations resulting from regulation-induced catch
improvements at fishing sites, I examine the change in the proportion of anglers who benefit
from a regulation when the regulation occurs along with a catch-related quality change. Take,
for example, the mandatory catch and release regulation. Some anglers who realize low to

134

moderate disutility with respect to mandatory catch and release may derive positive utility
from a scenario in which mandatory catch and release is accompanied by catch-related quality
improvements. That is, these anglers go from being worse off with the regulation to being
better off with both the regulation and the catch improvements.
Table 4.5 examines regulation-induced fishing site quality change scenarios involving the
three catch-related variables that have mean coefficients which are statistically significant
(P<0.01); Brown_75, Brown_Rate, and Brook_Rate. In Table 4.5, “better for” refer to the
proportion of anglers for whom the scenario makes them better off (i.e., generates positive
utility for the angler) while “worse for” refers to the proportion of anglers for whom the
scenario makes them worse off (i.e., generates negative utility for the angler). The scenarios
described in Table 4.5 show that regulation-induced catch-quality improvements increase the
proportion of anglers who receive positive utility, relative to the scenario in which regulations
are not accompanied by catch-quality improvements. These results also indicate that catchrelated improvements occurring from the artificial flies only gear restriction makes more
anglers better off than compared to when catch improvements result from the mandatory
catch and release restriction.

135

Table 4.5. Current proportion of anglers receiving positive utility from regulations (scenario 1)
and the proportion of anglers receiving positive utility when the regulation is accompanied by a
catch-related site quality improvement (scenarios 2-4).
REGULATIONS
Mandatory Artificial
Catch and
Flies Only
Release
Scenario 1:
Add regulations to a site without catch-related site
quality improvements

better for
worse for

13.2%
86.8%

22.5%
77.5%

Scenario 2: size increase:
Add regulations to a site with an increase in 4th
quartile length of 3” for brown trout

better for
worse for

15.8%
84.2%

26.5%
73.5%

Scenario 3: catch rate increase:
Add regulations to a site with an increase in catch
rate of 1 brown trout/hour and 1 brook trout/hour

better for
worse for

18.4%
81.6%

30.5%
69.5%

Scenario 4: size & catch rate increase:
Add regulations to a site with increase in 3rd
quartile length of 3” for brown trout and increase in
catch rate of 1 brown trout/hour and 1 brook
trout/hour

better for
worse for

21.6%
78.4%

35.1%
64.9%

4.5 Discussion
Results of this discrete choice study of stream trout fishing in Michigan generally
conform with results identified in other discrete choice studies of recreational fishing. Similar
to the review of recreational fishing discrete choice studies in Hunt et al. (2005), this analysis
reveals that on average, anglers prefer larger fish, higher catch rates, less restrictive
regulations, and lower travel distances. In particular, the average angler is substantially and
negatively impacted by the most highly restrictive regulations, with the average angler willing
to drive many miles to avoid mandatory catch and release and artificial flies only fishing
136

regulations at trout fishing sites. Importantly, results show that the less strict harvest
regulations (two trout per day) and gear restrictions (artificial lures and flies permitted; no
natural bait) have less impact on trout angler site choice. Managers desiring to mitigate
negative impacts on the fishery while minimizing impacts to anglers should consider
implementing such less strict regulations. Further, the literature on hooking mortality has
largely confirmed that the use of natural baits leads to greater fish mortality than the use of
artificial lures (see Taylor 1992 meta-analysis), while more limited and equivocal evidence exists
regarding a reduction in trout mortality through the use of artificial flies as opposed to artificial
lures (Wydoski 1977, Schisler and Bergersen 1996, and Risley and Zydlewski 2010) find evidence
for such a mortality reduction whereas Mongillo (1984) does not]. This suggests that the
greatest reductions in fish mortality with least impact to anglers can be achieved by banning
the use of natural baits, whereas mortality reductions achieved by mandating the use of
artificial flies are more uncertain and would have negative impacts on a large percentage of
trout anglers.
Despite these strongly negative mean preferences for the most restrictive regulations,
the results show that even while holding all other site characteristics (such as catch-related
attributes) constant, 22.5% of anglers (13.2% of anglers) prefer fishing in areas with flies only
gear restrictions (mandatory catch and release). Table 4.5 shows that regulation-induced
quality changes have the potential to further increase the number of anglers better off with
such regulations. In effect, this table illustrates the proportion of anglers who “switch” from
being made worse off through the implementation of strict regulations, to being made better
off through regulations accompanied by catch improvements. Though the comparison is not
137

perfect, it is worth noting that the proportions of anglers preferring these regulations (with or
without catch-related quality improvements) are far greater than the proportion of coldwater
trout streams in Michigan having mandatory catch and release (0.2%) and artificial flies only
restrictions (0.6%) throughout the trout fishing season. Given this imbalance, increasing the
number of trout fishing sites with mandatory catch & release and artificial flies only regulations
to provide additional opportunities to these anglers may be advisable from an equity
standpoint.
Despite the potential to make some anglers better off as a result of regulation-induced
improvements in catch quality, the mean angler is likely sufficiently negatively impacted by
regulations such that regulations are unlikely to generate the level of catch-related
improvements necessary to fully compensate these anglers. For example, Table 4.3 shows that
an increase in brown trout catch rate (brook trout catch rate) of 7.9 brown trout (14.7 brook
trout) per hour would be necessary to fully compensate for the mean angler for implementing
catch and release harvest regulations, whereas an increase in brown trout catch rate (brook
trout catch rate) of 4.8 brown trout (8.9 brook trout) per hour would be necessary to fully
compensate for the implementation of flies only gear restrictions. These increases in catch
rates are beyond the highest attribute level of hourly trout catch rates in the choice experiment
(four trout per hour) and may not be a realistic fisheries outcome from implementing
regulations. Such strong negative mean angler preferences for strict regulations will likely
ensure that fisheries managers are confronted with disproportionate angler opposition to
proposed new regulations. However, such regulation changes are possible and may require
operating within the state political structure. For example, gear-restricted streams in Michigan
138

were legislatively capped at 212 miles in 2002 (PA 434 2002), an increase from the 100 miles
limit in 1994 (PA 451 1994).
Our research also sheds light on angler preferences with respect to catch rate and catch
size for brown trout and brook trout. While anglers’ benefit from an improvement in catch
rates for both species, stronger mean preferences are evident for brown trout. This result
suggests greater benefits for anglers from stocking and habitat management programs which
improve brown trout catch rate. Results of this analysis indicate that Michigan trout anglers
have stronger positive preferences for increases in the 75th percentile of brown trout size
distribution, while there is less evidence that the mean angler is concerned with upper end
brook trout size. This difference may be due to angler preferences for the challenge of the fight
associated with catching a large brown trout (which can reach lengths of up to 30 inches)
relative to brook trout (for which 13 inches would be considered particularly large). However,
standard deviation is statistically significant at the 5% level for upper end brook trout size,
indicating varying preferences for changes in this variable throughout the angler population.
Finally, preference heterogeneity at the 1% significance level has been identified for the 25 th
percentile of the size distribution for both brook trout and brown trout. To speculate, this may
have to do with varying angler preferences for trout reaching the minimum size for harvest. In
the majority of trout river/streams in Michigan, minimum size limit for brook trout is 7”,
whereas the minimum size limit for brown trout is 8”. These minimum size limits are also the
median of the 1st quartile size distributions for both brook trout (6”, 7”, and 8”) and brown
trout (7”, 8”, and 10”) attribute levels that an individual could encounter in the choice
experiment. This preference heterogeneity may result from harvest-oriented anglers selecting
139

fishing sites perceived to have a greater number of legal-sized trout, while other non-harvestoriented anglers may be unconcerned with trout at the lower end of the distribution reaching a
minimum size.
Ultimately, effective management of stream trout fisheries in Michigan involves
allocating scarce budgetary funds to management actions which provide the greatest benefit to
anglers. The results of this analysis provide important insights into angler preferences for
stream trout fishing attributes such as catch size, catch rate, and gear and harvest regulations.
Further, this analysis provides managers with information on how angler support for
regulations changes when regulations are accompanied by catch-related site quality
improvements. The research within can lead to improved resource allocation by allowing
managers to direct resources to the improvement of attributes most preferred by anglers and
thus improve the quality of stream trout fishing experiences.

140

APPENDICES

141

APPENDIX A

142

APPENDIX A: SURVEY MAILING TIMELINE AND RESPONSE RATE

ASAP Printing company (Okemos, Michigan) was used for the printing and mailing of
surveys and contact materials. The following bullets illustrate the timeline for survey mailing.


First Wave Mailing: May 7, 2013
This wave consisted of a 2-sided 8.5” by 11” document (see Figure D.1 and Figure D.2 in
Appendix D). The front of the document contained information about the purposes of
the survey and a website address to access the survey. The back of the document
contained answers to common questions individuals often have about the nature and
purpose of the survey. The mailing envelope contained a survey logo we developed
which consisted of an outline of the state of Michigan overlayed with an outline of a
trout (see Figure D.3 in Appendix D).



Second Wave Mailing: May 16, 2013
This wave consisted of a 2-sided 5.5” by 4.25” postcard (see Figure D.4 and Figure D.5 in
Appendix D). The front of the postcard consisted of a brief request to complete the
survey, the link to the survey website, and a color image of the survey logo. The back of
the postcard contained rationale for the survey and contact information.



Third Wave Mailing: May 29, 2013
This wave consisted of a 2-sided 8.5” by 5” postcard (see Figure D.6 and Figure D.7 in
Appendix D). The front of the postcard consisted of a brief request to complete the
survey, the link to the survey website, and a color image of the survey logo. The back of
the postcard contained rationale for the survey and contact information.



Fourth Wave Mailing: July 2, 2013 –
This wave contained a 2-sided 8.5” by 11” document (see Figure D.8 and Figure D.9 in
Appendix D). This wave also contained a 12 page survey (see Figures B.1 – B.12 in
Appendix C) consisting of three 17” by 11” pages folded over to create a booklet. The
page containing the front and back of the survey was of slightly heavier weight. Finally,
this wave contained a 9” by 12” business reply mail envelope. These materials were
mailed in a 9” by 12” envelope which contained the same image and text as the first
outgoing envelope in Figure D.3 in Appendix D.

143

To reduce undeliverable mail, ASAP Printing company cross-checked the individuals’
mailing addresses with the National Change of Address list. A total of 192 individuals were
identified as having changed addresses, and the addresses were updated accordingly. This list
of changed addresses included out-of state addresses for 13 individuals. Throughout the survey
mailing process, a total of 92 surveys were returned as undeliverable. For calculating the
effective response rate, the 13 individuals who moved out of state and the 92 undeliverable
surveys were removed from the calculation, yielding an effective sample size of 3,994.
The internet nature of the survey, combined with the uncertainty associated with the date
that a respondent received a mailing, complicates calculating the precise survey response by
wave. However, we provide an estimate of response rate by mailing below.





First Wave Response
Second Wave Response
Third Wave Response
Fourth Wave Response

– 420 internet surveys
– 348 internet surveys
– 218 internet surveys
– 743 valid mail surveys and 87 internet surveys

For the internet survey, there were 4 people who logged on but did not answer a question,
and 10 people have 2 survey entries. These individuals were eliminated from our analysis and
subtracted from total internet survey responses listed above, leaving a total of 1038 valid
internet survey respondents. Below is the equation for the effective response rate.

𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑅𝑎𝑡𝑒 =

𝑉𝑎𝑙𝑖𝑑 𝐼𝑛𝑡𝑒𝑟𝑛𝑒𝑡 𝑆𝑢𝑟𝑣𝑒𝑦𝑠+𝑉𝑎𝑙𝑖𝑑 𝑀𝑎𝑖𝑙 𝑆𝑢𝑟𝑣𝑒𝑦𝑠
𝑆𝑢𝑟𝑣𝑒𝑦 𝑆𝑎𝑚𝑝𝑙𝑒 𝑆𝑖𝑧𝑒−𝑈𝑛𝑑𝑒𝑙𝑖𝑣𝑒𝑟𝑎𝑏𝑙𝑒𝑠−𝑁𝐶𝑂𝐴 𝑜𝑢𝑡 𝑜𝑓 𝑆𝑡𝑎𝑡𝑒

=

1038 + 743
= 44.6%
4099 − 92 − 13

Table A.1. Summary of trout angler survey response and disposition
Initial Sample Size
4,099
Undeliverable
92
Total Responses
1781
Wave 1 Responses (internet survey)
420
Wave 2 Responses (internet survey)
313
Wave 3 Responses (internet survey)
218
Wave 4 Responses (mail and internet survey)
830
Wave 4 Mail Survey
743
Wave 4 Internet Survey
87
Out of State
13
Refusals
2
144

Table A.2. Survey weighting
Birth_year
Birth year ≥ 1979
Birth year ≥ 1969 and ≥ 1978
Birth year ≥ 1959 and ≥ 1968
Birth year ≥ 1949 and ≥ 1958
Birth year ≥ 1939 and ≥ 1948
Birth year < 1939

Survey Weighting
1.44222
1.25939
1.02139
0.87668
0.73865
0.79386

145

APPENDIX B

146

Figure B.1. Mail survey page 1

147

Figure B.2. Mail survey page 2

148

Figure B.3. Mail survey page 3

149

Figure B.4. Mail survey page 4

150

Figure B.5. Mail survey page 5

151

Figure B.6. Mail survey page 6

152

Figure B.7. Mail survey page 7

153

Figure B.8. Mail survey page 8

154

Figure B.9. Mail survey page 9

155

Figure B.10. Mail survey page 10

156

Figure B.11. Mail survey page 11

157

Figure B.12. Mail survey page 12

158

APPENDIX C

159

Figure C.1. Choice experiment format #2

160

APPENDIX D

161

Figure D.1. First mail contact (front)

162

Figure D.2. First mail contact (back)

163

Figure D.3. Envelope for first mail contact

164

Figure D.4. Second mail contact (postcard front)

165

Figure D.5. Second mail contact (postcard back)

166

Figure D.6. Third mail contact (postcard front example)

167

Figure D.7. Third mail contact (postcard back)

168

Figure D.8. Fourth mail contact (letter front)

169

Figure D.9. Fourth mail contact (letter back)

170

APPENDIX E

171

Table E.1. Paired choice alternatives and attribute levels
Choice
Scenarios
1 a
b
2 a
b
a
3 b
a
4 b
5 a
b
6 a
b
7 a
b
8 a
b
9 a
b
10 a
b
11 a
b
a
12
b
13 a
b
14 a
b
15 a
b
16 a
b
17 a
b
a
18 b
a
19 b

Dist

Hvst

Gear

75
35
100
10
100
35
75
10
35
10
50
20
50
50
35
75
35
35
50
10
50
100
35
50
75
50
100
75
100
50
75
75
75
35
100
10
100
100

0
0
0
5
2
0
2
5
0
2
0
5
2
2
2
5
5
2
5
5
0
2
0
5
5
5
5
0
0
5
0
0
5
5
5
2
2
5

1
2
1
0
0
2
0
0
1
2
1
1
0
2
0
1
0
2
0
1
2
2
2
1
0
2
0
0
2
2
2
0
2
1
2
2
1
1

Brwn
catch
2
2
8
8
2
0.25
8
1
0.5
4
0.25
2
4
4
2
2
1
0.25
0.25
1
8
0.25
2
1
4
2
2
4
0.5
0.25
0.25
0.5
1
2
0.25
8
1
2

Brwn
25
7
10
7
7
8
10
8
7
8
8
8
7
7
8
7
7
10
7
10
8
8
7
8
8
8
10
8
7
10
10
10
7
10
8
10
7
7
8

172

Brwn
75
15
14
9
11
16
18
10
15
16
12
12
11
15
16
11
15
14
9
18
10
12
15
16
16
10
12
16
9
12
18
18
15
18
12
14
11
15
10

Brk
catch
2
0.25
0.25
1
1
2
2
8
0.25
1
0.5
0.5
0.5
8
0.25
4
2
8
1
2
1
1
2
0.25
0.25
4
0.5
8
0.5
4
0.25
8
0.5
8
0.25
2
0.25
1

Brk
25
7
8
7
8
8
7
8
7
8
7
8
7
7
8
7
8
7
7
7
7
6
6
6
6
6
6
6
6
7
7
7
7
6
8
6
8
8
6

Brk
75
9
13
9
13
13
9
13
9
13
12
13
12
9
10
9
10
8
12
8
12
8
7
8
7
8
11
8
11
8
8
8
8
11
10
11
10
9
7

Table E.1. (cont’d)
Choice
Scenarios
a
20 b
a
21 b
a
22 b
a
23 b
a
24 b
a
25 b
a
26 b
a
27 b
a
28 b
a
29 b
a
30 b
a
31 b
a
32 b
a
33 b
a
34 b
a
35 b
a
36 b
a
37 b
a
38 b
a
39 b

Dist

Hvst

Gear

75
50
35
50
50
75
50
10
35
20
10
10
20
20
20
10
10
20
75
20
100
35
100
75
75
100
75
10
100
20
100
50
75
75
10
75
20
100
20
75

2
2
5
0
5
5
2
0
2
5
5
5
5
2
0
5
0
2
5
0
5
2
0
0
0
2
2
0
2
2
5
0
5
2
2
2
2
5
5
2

1
2
2
0
2
2
1
0
1
2
0
1
0
2
2
1
2
2
0
1
0
0
2
1
2
0
2
1
2
0
1
1
1
0
2
1
2
2
1
1

Brwn
catch
0.25
8
0.5
0.25
0.25
0.5
4
0.25
2
0.5
4
4
2
2
4
0.5
2
0.25
1
8
0.5
4
8
8
4
4
0.25
0.5
0.5
0.25
2
0.5
4
0.25
8
4
4
8
1
0.5

Brwn
25
7
7
7
7
7
10
10
7
10
10
10
7
10
8
7
7
7
8
7
10
7
8
10
10
10
8
8
8
8
10
7
8
7
10
7
7
7
8
8
7
173

Brwn
75
11
9
15
11
9
14
18
9
12
12
14
15
18
16
11
9
15
10
11
18
9
16
14
12
12
10
10
12
12
14
9
10
11
12
15
11
11
12
16
9

Brk
catch
0.5
0.25
1
0.5
2
1
2
0.5
1
1
8
1
4
0.5
4
8
8
4
1
0.5
2
0.25
2
4
1
2
8
4
4
8
4
0.5
8
1
1
2
2
4
2
2

Brk
25
8
6
8
6
8
6
6
8
6
8
8
6
8
6
6
8
6
8
6
8
6
8
8
6
8
6
8
8
8
8
7
7
7
7
7
7
7
7
8
8

Brk
75
9
7
9
8
9
8
11
9
11
9
13
7
13
7
7
13
7
13
7
9
7
9
13
11
13
11
10
9
10
9
12
9
12
9
12
8
12
8
10
10

Table E.1. (cont’d)

Choice
Scenario
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59

a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b
a
b

Dist

Hvst

Gear

10
100
10
10
20
20
20
50
10
75
75
75
100
100
100
20
75
35
35
20
50
35
50
20
35
35
10
35
20
10
20
100
10
50
10
20
20
35
20
75

5
5
2
2
2
0
0
2
0
0
2
5
2
2
0
5
0
2
2
2
2
0
5
2
5
0
2
5
2
0
5
5
5
0
0
5
0
0
2
5

1
0
1
0
1
1
0
0
0
1
1
2
1
1
0
2
0
1
2
0
2
1
1
0
1
1
2
0
2
2
1
0
1
2
1
0
1
2
0
0

Brwn
catch
0.5
1
0.25
4
0.5
2
0.25
4
0.5
2
1
0.5
0.5
1
8
0.5
4
1
2
8
4
4
8
1
4
0.5
0.25
2
1
8
2
2
8
8
0.5
1
1
0.5
4
1

Brwn
25
8
8
8
10
8
8
10
10
10
8
10
8
10
7
8
8
8
7
8
8
8
10
10
8
10
10
10
7
10
10
8
7
8
10
7
10
7
7
10
10
174

Brwn
75
12
10
12
18
10
16
14
14
12
12
14
16
18
15
12
12
16
11
10
10
12
12
12
12
14
14
12
9
18
12
10
11
16
14
15
18
9
15
18
14

Brk
catch
1
4
2
4
1
8
1
0.5
2
1
4
0.25
8
0.5
8
0.25
4
0.5
0.5
0.5
0.25
0.25
0.25
4
0.5
2
4
0.25
8
1
8
2
4
8
0.5
2
0.25
4
0.25
0.25

Brk
25
8
8
7
6
7
6
6
7
6
7
6
7
6
7
7
6
7
6
6
7
6
7
7
6
7
6
7
6
7
6
6
7
6
7
6
6
6
6
8
8

Brk
75
10
10
8
11
8
11
8
9
8
9
8
12
8
12
8
8
8
8
7
9
7
9
9
7
9
7
9
8
9
8
7
8
7
8
11
8
11
8
9
13

Table E.1. (cont’d)
Choice
Scenario
a
60 b
a
61 b
a
62 b
a
63 b
a
64 b
a
65 b
a
66 b
a
67 b
a
68 b
a
69 b
a
70 b
a
71 b
a
72 b

Dist

Hvst

Gear

10
100
35
100
50
50
50
100
35
50
35
20
50
35
50
75
35
100
10
100
20
50
20
35
20
10

2
0
0
0
0
2
2
0
2
2
0
0
0
5
5
0
5
5
0
2
0
0
5
2
5
0

0
2
1
0
1
1
0
0
0
1
0
2
0
0
2
2
2
0
0
1
0
0
2
1
2
0

Brwn
catch
8
0.5
2
8
0.25
2
0.25
1
1
0.25
0.5
8
1
4
0.5
8
1
4
0.25
1
1
0.25
2
1
2
0.25

Brwn
25
10
7
10
8
10
10
7
8
7
10
7
7
7
10
8
7
8
10
8
10
8
8
7
10
7
8

175

Brwn
75
12
11
14
16
12
18
11
12
9
14
9
9
15
12
10
15
16
18
10
12
12
10
9
18
9
16

Brk
catch
0.5
0.5
4
8
8
2
8
8
4
2
8
2
4
4
4
0.25
8
0.5
0.25
1
0.5
0.25
0.5
1
0.5
0.25

Brk
25
8
8
8
8
8
8
6
6
6
6
8
7
8
7
7
8
7
8
7
8
7
8
8
7
8
7

Brk
75
9
13
9
10
9
10
11
11
11
11
10
8
10
8
12
13
12
13
12
9
12
9
10
12
10
12

APPENDIX F

176

Table F.1. Pearson product moment correlations for attributes within and across choice alternatives in choice pairs
Site A

Site B

Dist

Brwn Brwn Brwn Brk Brk Brk
Brwn Brwn Brwn Brk Brk Brk
Dist Hvst Gear
catch 25 75 catch 25 75
catch 25 75 catch 25
75
1.00 0.00 0.00 0.02 0.00 0.00 -0.01 0.00 0.00 -0.01 0.00 0.04 0.05 0.00 0.00 0.05 0.00 0.00

Hvst

0.00 1.00 0.00 -0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.04 0.00 0.00 0.00 0.00 0.00 0.00

Gear

0.00 0.00 1.00 0.02 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.11 0.00 0.00 -0.01 0.00 0.00

Dist Hvst Gear

Brwn catch 0.02 -0.01 0.02 1.00 -0.01 -0.04 -0.10 0.05 0.02 -0.01 -0.01 0.02 0.06 -0.10 -0.02 0.04 -0.10 0.03
Site A Brwn 25
0.00 0.00 0.00 -0.01 1.00 0.00 0.00 0.00 0.00 0.02 0.00 0.01 -0.04 0.00 0.00 0.00 0.00 0.00
Brwn 75

0.00 0.00 0.00 -0.04 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.01 -0.04 0.00 0.00 0.05 0.00 0.00

Brk catch

-0.01 0.00 0.00 -0.10 0.00 0.00 1.00 0.00 0.00 0.03 0.00 0.02 0.02 0.00 0.00 -0.02 0.00 -0.01

Brk 25

0.00 0.00 0.00 0.05 0.00 0.00 0.00 1.00 0.00 0.02 0.00 -0.04 0.00 0.00 0.00 -0.05 0.00 0.00

Brk 75

0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 1.00 0.02 0.00 0.01 0.01 0.00 0.00 -0.09 0.00 0.00

Dist

-0.01 0.00 0.02 -0.01 0.02 0.00 0.03 0.02 0.02 1.00 0.02 -0.08 -0.03 -0.05 0.00 0.00 0.02 0.03

Hvst

0.00 0.00 0.00 -0.01 0.00 0.00 0.00 0.00 0.00 0.02 1.00 -0.04 -0.15 0.00 0.00 -0.06 0.00 0.00

Gear

0.04 -0.04 0.00 0.02 0.01 0.01 0.02 -0.04 0.01 -0.08 -0.04 1.00 0.10 0.01 0.04 -0.08 -0.04 0.01

Brwn catch 0.05 0.00 0.11 0.06 -0.04 -0.04 0.02 0.00 0.01 -0.03 -0.15 0.10 1.00 0.00 -0.05 0.00 0.05 -0.05
Site B Brwn 25
0.00 0.00 0.00 -0.10 0.00 0.00 0.00 0.00 0.00 -0.05 0.00 0.01 0.00 1.00 0.00 -0.08 0.00 0.00
Brwn 75

0.00 0.00 0.00 -0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 -0.05 0.00 1.00 0.06 0.00 0.00

Brk catch

0.05 0.00 -0.01 0.04 0.00 0.05 -0.02 -0.05 -0.09 0.00 -0.06 -0.08 0.00 -0.08 0.06 1.00 0.05 0.06

Brk 25

0.00 0.00 0.00 -0.10 0.00 0.00 0.00 0.00 0.00 0.02 0.00 -0.04 0.05 0.00 0.00 0.05 1.00 0.00

Brk 75

0.00 0.00 0.00 0.03 0.00 0.00 -0.01 0.00 0.00 0.03 0.00 0.01 -0.05 0.00 0.00 0.06 0.00 1.00

177

APPENDIX G

178

Table G.1. Mail survey choice scenario groupings by survey number
Region 2

Region 1
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6
17-29-10-64
14-68-20-58

#
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

Ranking
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6
17-29-10-64
14-68-20-58

#
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6
17-29-10-64
14-68-20-58

#
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80

179

Region 3
Ranking
#
Order
30-27-58-35 81
9-16-28-67
82
41-50-12-55 83
15-68-62-38 84
59-2-65-57
85
42-3-20-14
86
1-10-8-29
87
70-21-44-66 88
36-47-40-4
89
18-61-60-23 90
53-24-51-71 91
37-13-56-69 92
17-52-25-34 93
32-45-48-63 94
5-19-54-26
95
72-43-11-39 96
46-49-7-64
97
31-22-33-6
98
17-29-10-64 99
14-68-20-58 100

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6
17-29-10-64
14-68-20-58

#
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120

Ranking
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6
17-29-10-64
14-68-20-58

Table G.2. Internet survey choice scenario groupings by survey number

Region 1
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

#
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36

Region 2
Ranking
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

#
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

#
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72

180

Region 3
Ranking
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

#
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90

Binary
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

#
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108

Ranking
Order
30-27-58-35
9-16-28-67
41-50-12-55
15-68-62-38
59-2-65-57
42-3-20-14
1-10-8-29
70-21-44-66
36-47-40-4
18-61-60-23
53-24-51-71
37-13-56-69
17-52-25-34
32-45-48-63
5-19-54-26
72-43-11-39
46-49-7-64
31-22-33-6

REFERENCES

181

REFERENCES

Aas, O., W. Haider, and L.M. Hunt. 2000. Angler Responses to Potential Harvest Regulations in
Engerdal, Norway: A Conjoint Based Choice Modeling Approach. North American Journal
of Fisheries Management 20: 940–950.
Ahn, S., J.E. Steiguer, R.B. Palmquist, and T.P. Holmes. 2000. Economic Analysis of the Potential
Impact of Climate Change on Recreational Trout Fishing in the Southern Appalachian
Mountains: An Application of a Nested Multinomial Logit Model. Climatic Change 45:
493–509.
American Automobile Association (AAA). 2013. Your Driving Costs. AAA Association
Communication, Heathrow, Florida, USA.
Anderson, R.M., and Nehring, R.B. 1984. Effects of a Catch-and-Release Regulation on a Wild
Trout Population in Colorado and its Acceptance by Anglers. North American Journal of
Fisheries Management 4.3: 257-265.
Bain, M. B. 1987. Structured Decision Making in Fisheries Management: Trout Fishing
Regulations on the Au Sable River, Michigan. North American Journal of Fisheries
Management 7.4: 475-481.
Bartholomew, A., and J.A. Bohnsack. 2005. A Review of Catch-and-Release Angling Mortality
with Implications for No-Take Reserves. Reviews in Fish Biology and Fisheries 15: 129154.
Beville S. and G. Kerr. 2008. Fishing for Understanding: A Mixed Logit Model of Freshwater
Angler Preferences. Paper presented to the New Zealand Agricultural and Resource
Economics Society annual conference.
Beville, S. and G. Kerr. 2009. Fishing for more understanding: a mixed logit-error component
model of freshwater angler site choice. Paper to the 53rd annual Australian Agricultural
& Resource Economics Society conference. Cairns, Queensland, 10-13 February 2009.
Bockstael, N., K. McConnell, and L. Strand. 1989. A Random Utility Model for Sportsfishing:
Some Preliminary Results for Florida. Marine Resource Economics 6: 245-260.
Boyd, J.H., and R.E. Mellman. 1980. The Effect of Fuel Economy Standards on the U.S.
Automotive Market: A Hedonic Demand Analysis. Transportation Research 14: 367-378.
Breffle, W., and E. Morey. 2000. Investigating Preference Heterogenity in a Repeated-Choice
Recreation Model for Atlantic Salmon Fishing. Marine Resource Economics 15.1:1-20.

182

Cardell, N., and F. Dunbar. 1980. Measuring the Societal Impacts of Automobile Downsizing.
Transportation Research 14: 423-434.
Cooke, S. J., and H.L. Schramm. 2007. Catch‐and‐Release Science and its Application to
Conservation and Management of Recreational Fisheries. Fisheries Management and
Ecology 14.2: 73-79.
Cooke, S. J., and C.D. Suski. 2005. Do We Need Species-Specific Guidelines for Catch-andRelease Recreational Angling to Effectively Conserve Diverse Fishery
Resources? Biodiversity & Conservation 14.5: 1195-1209.
Eggert, H., and B. Olsson 2009. Valuing Multi-Attribute Marine Water Quality. Marine
Policy 33.2: 201-206.
Gigliotti, L. M. 1989. No-kill fishing regulations: an assessment of the social and recreational
characteristics and behaviors of Michigan stream trout anglers with special
consideration of anglers on selected sections of the Au Sable River (Doctoral
dissertation, Michigan State University. Department of Fisheries and Wildlife).
Hunt, L.M. 2005. Recreational Fishing Site Choice Models: Insights and Future Opportunities.
Human Dimensions of Wildlife 10: 153-172.
Johnson, F.R., B. Kanninen, M. Bingham, and S. Özdemir, S. 2007. Experimental Design for
Stated Choice Studies. IN Valuing Environmental Amenities Using Stated Choice Studies,
ed B.J. Kanninen. Springer, Dordrecht, Netherlands.
Kerkvliet, J., and C. Nowell. 2000. Tool for Recreation Management in Parks: The Case of the
Greater Yellowstone’s Blue-Ribbon Fishery. Ecological Economics 34: 89–100.
Mongillo, P.E. 1984. A Summary of Salmonid Hooking Mortality. Washington Department of
Game, Seattle.
Massey, D.M., S.C. Newbold, and B. Gentner. 2006. Valuing Water Quality Changes Using a
Bioeconomic Model of a Coastal Recreational Fishery. Journal of Environmental
Economics and Management 52.1: 482-500.
McConnell, K., and Wei-Chun Tseng. 1999. Some Preliminary Evidence on Sampling of
Alternatives with the Random Parameters Logit. Marine Resource Economics 14.4: 317332.
McFadden, D. 1974. Conditional Logit Analysis of Qualitative Choice Behaviour. IN Frontiers in
Econometrics, ed P. Zarembka Academic Press, New York.
Michigan Department of Natural Resources [MDNR]. 2013. Michigan Fishing Guide

183

Michigan Department of Natural Resources [MDNR]. 2012.
http://www.michigandnr.com/FISHSTOCK/
Michigan Department of Natural Resources [MDNR]. 2011. Fisheries Division 2011
Accomplishments Report.
Milon, J. W. 1988a. A Nested Demand Shares Model of Artificial Marine Habitat Choice by Sport
Anglers. Marine Resource Economics 5: 191–213.
Milon, J. W. 1988b. Travel Cost Methods for Estimating the Recreational Use Benefits of
Artificial Marine Habitat. Southern Journal of Agricultural Economics 20.1: 87–101.
Morey, E. R., W.S. Breffle, R.D. Rowe, and D.M. Waldman. 2002. Estimating Recreational Trout
Fishing Damages in Montana’s Clark Fork River Basin: Summary of a Natural Resource
Damage Assessment. Journal of Environmental Management 66: 159–170
Oh, C., and R.B. Ditton. 2006. Using Recreation Specialization to Understand Multi-Attribute
Management Preferences. Leisure Sciences 28: 369–384.
Provencher, B., and R. Bishop. 2004. Does Accounting for Preference Heterogeneity Improve
the Forecasting of a Random Utility Model? Journal of Environmental Economics and
Management 48: 793–810.
Revelt, D., and K. Train. 2000. Customer-Specific Taste Parameters and Mixed Logit:
Households' Choice of Electricity Supplier.
Risley, C. A., and J. Zydlewski. 2010. Assessing the Effects of Catch-and-Release Regulations on a
Brook Trout Population Using an Age-Structured Model. North American Journal of
Fisheries Management 30.6: 1434-1444.
Schisler, G., and E. P. Bergersen. 1996. Post Release Hooking Mortality of Rainbow Trout Caught
on Scented Artificial Bait. North American Journal of Fisheries Management 16: 570–
578.
Taylor, M. J., and K.R. White. 1992. A Meta-Analysis of Hooking Mortality of Nonanadromous
Trout. North American Journal of Fisheries Management 12.4: 760-767.
Timmins, C. and J. Murdoch. 2007. A Revealed Preference Approach to the Measurement of
Congestion in Travel Cost Models. Journal of Environmental Economics and
Management 53.2: 230-249.
Train, K. 2009. Discrete Choice Methods with Simulation, Second Edition. Cambridge University
Press, Cambridge, MA.
Train K. 1998. Recreation Demand Models with Taste Differences Over People. Land Economics
74: 230-239.
184

U.S. Department of the Interior [USDOI]. 2013. 2011 National Survey of Fishing, Hunting, and
Wildlife-Associated Recreation—Michigan. U.S. Fish and Wildlife Service, Washington,
D.C., USA.
Weicksel, S.A. 2012. Measuring preferences for changes in water quality at great lakes beaches
using a choice experiment (M.S. Thesis, Michigan State University).
Wielgus, J., L.R. Gerber, E. Sala, and J. Bennett. 2009. Including Risk in Stated-Preference
Economic Valuations: Experiments on Choices for Marine Recreation. Journal of
Environmental Management 90: 3401–3409.
Wydoski, R.S. 1977. Relation of Hooking Mortality and Sublethal Hooking Stress to Quality
Fishery Management. Pages 43-87 IN Catch-and-Release Fishing as a Management Tool,
eds R. A. Barnhart and T. D. Roelofs. Humboldt University, Arcata, California, USA.

185