a. I. P «gnu

‘0‘. t1.

t‘r OI

lt‘ih..‘. ‘.| p 0....

 

wit...” a x. x

 

 

  
 

.4

’vtmfinfié’

.. b.04‘

2—; :né 3*.

" ‘m‘ r-...... -
"4 ‘

 

i
,1
24
:1!

 

 

{g4 {5.

a.
u..-
1 . a,-

. 1,
E313?

,

111, i
lv'iiij {6:11.11

1:1- :1

, A. 2;.
1- “Vi“

1 L,11111‘1 1Hjl11_ 71 I
1‘3” ‘

| . :UI

:11,,:. . 1.
162"" w '«v'..'2il‘,.- »;
‘32: h1w"if:'
1.1.1.111; 1"111 ‘12,;1‘11 1. ”Fl.
1.1.“ r: ”11111.1.

{1‘52}
{(3,2‘1" 111:5‘1'1:,,M11

6 "131’11‘3" '
'3'
10.1, 3" 1111121221.“,
1' "1' 111 ..I. 11;:
.91.!,-‘Li1.1.1.11q,131‘1‘h‘3“
"5:, 211'111i'i'2’13.‘

#1'1‘éé‘
Vg'lzb- 1%;92‘11

$3111.11“; 3 £111
:3“? “5:: {‘Eéifiiti‘! Mandi
.-l‘!"f;zgx . Ei‘fi- “£92.“
1. J , A . ,

v_1‘1
[111‘ "$1?
”12152
1.1,11 .31
. H

E‘”w

2113! 2:11,“ -1
“I M; I 1 5:1

V113,!!111f111fiil :

-"', 11,,1‘1311,‘

1;":131’J'h ubmvi
;, 1..

' ‘2 71",‘1111111111

h
s” "2

1;!

u

‘11,
3:1

...
"1-: an
.
Wm..-

. w...

311‘} 5'7»-

m

. ‘1: XL '1 $1"
1?;!i}.ft%%;if.2112111:1 1) 2:47
1' , 11”
1121:1321 1111.11,. 1.;
1211-1i1‘l11211111n1

1‘1 finhx‘fl"
12 .211

113???

‘v..“._
vv—u

l l 1
I}; { 213133321121?"

~21
2,5

.--—"

III-1'
‘25:“

3...; “~—
A-“
M. . .

. a 799:7—

.
:3?“ w
,. -....._ W... .— g
- ;_ ~0:.. _ _- ,.
“3.- . ,- w;
_ ‘ ""41
1} W I

V‘.

o
_,_. _ V. .—
-.'o‘.—‘4 .—o -.—--'

”dw"""
m
32:23. v

......—.~o-—-—-
‘ ;

‘ W

-.,.. ..
-...~:‘ F"

. -’
-fi, ,:- A.
..._... - ~—-
4...,- .. ..
.. - ..
._.--_,_m_,
J.
..

flu...
L' :1 M
- a-n.

._ M
.—

zcu a

.a—u

 

.—

IllmlIliil’llllllllllllfilllul

31293 018

This is to certify that the

dissertation entitled

Birder Preferences for Attributes of Birding Sites:
A Binary-Choice Experiment

presented by
Karin M.J. Steffens

has been accepted towards fulfillment
of the requirements for

Ph. D . degree in Agricultural Economics

 

' v on

/ Mor- professor

 

Date (28-20-1999

MSU is an Affirmative Anion/Equal Opportunity Institution 042771

 

 

LIBRARY

M'Chlgan State
University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

0(

fig? 0 L 2334;

‘4 2:4

#:1149215 6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 c/CIHCIDmOuopes-gu

BIRDER PREFERENCES FOR ATTRIBUTES OF BIRDING SITES:
A BINARY-CHOICE EXPERIMENT

By

Karin M.J. Steffens

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Agricultural Economics

1 999

ABSTRACT

BIRDER PREFERENCES FOR ATTRIBUTES OF BIRDING SITES:
A BINARY-CHOICE EXPERIMENT

By

Karin M.J. Steffens

This research uses a newly developed optimal experimental design for
binary-choice experiments to explore birder preferences for selected attributes of
birding sites. It also explores how these preferences are related to birder
preferences for biodiversity, as measured by seven different biodiversity
indicators. Qualitative interviews with birders generated information about
birders and their birding activities. This information led to the development of a
questionnaire, the selection of birding site attributes to be evaluated by survey
respondents, and contributed to the creation of a birder profile.

Study participants, who were members of the American Birding
Association (ABA) in Central Michigan were similar in socioeconomic
characteristics to ABA birders in general. ABA birders tend to differ, however,
from birders more broadly defined.

The birder preference data were analyzed using a binary-choice
experiment. The new optimal experimental design differs from a traditional
approach with pre-selected attribute levels in that a variable is updated in the

process of data collection to try and achieve optimal choice probabilities. The

number of variables in the design solely determines the optimal choice
probabilities. This study is the first to show that the model has practical
applications. Using D-optimality as a criterion, the “updating” approach of the
model is a more efficient design than a traditional approach. The limitations of
the model include the trade-off between realism in the choice scenarios and the
large differentials in the updating variable that are required for optimality, the
potentially high cost of conducting the interviews, and the time and effort
involved in implementing the design.

The binary-choice experiment was estimated as a binary-logit model. The
model with six attributes of birding sites correctly predicted 64% of the
responses and all coefficients were significant. Models with alternative
biodiversity-indicators revealed that the biodiversity indicators explained site
choice well and all seven indicators were significant. - .

Tests indicated that the preference model for six site attributes was not
equivalent to the biodiversity-indicator models. The user-preference model is
preferable. Thus, birders care about the composition of biodiversity but where
birder preference information is not available biodiversity measures were a good

indicator of birder preferences for birding sites.

Copyright by
KARIN M.J. STEFFENS
1999

To John, my mother and in memory of my father

ACKNOWLEDGMENTS

I would like to express my gratitude to John Hoehn, my dissertation
supervisor, for his support throughout this research process, particularly at
critical stages of the research when I needed his support the most. Thanks to
Frank Lupi, one of my committee members. His professional expertise,
readiness to discuss issues when they arose, and willingness to review draft
after draft were invaluable to the completion of this research. Frank, I could not
have done it without you. I also wish to thank of Eileen Van Ravenswaay, Jeff
Wooldridge, and Don Holecek, the other members of my dissertation committee.
Their support and professional insights at important junctures during the
research were very much appreciated.

I am grateful for the support from the College of Agriculture and Natural
Resources that was extended in the form of a Dissertation Completion
Fellowship. This research also benefited greatly from the support of the
American Birding Association (ABA). I would like to thank Greg Butcher (former
ABA executive director) for permission to use the ABA membership directory to
contact ABA members for participation in the study, for a letter of support to
accompany survey materials, and for a complimentary 1-year membership to the
ABA Thanks also to Sharon Bartels for providing a summary of the 1994 ABA

membership survey.

vi

Discussions with Barbara Kanninen about the application of her model
were critically important at various stages of the study. I would like to thank her
for her help.

I am particularly grateful to Sharon Johnson (president of the Genessee
Audubon Society) for inspiring me to use birders in the research application, for
many hours spent looking at bird charts and recounting birding experiences, and
for assistance in contacting ABA members in Genessee County. Her extensive
knowledge of birds and birding was an invaluable resource for this research.

This project could not have been completed without the help from many
birders who were willing to answer my questions. I would like to give special
thanks to Ted Black (Capital Area Audubon Society), Kip Miller (Love Creek
Nature Center), and Dave Ewert (The Nature Conservancy) for spending time
talking about birds and for providing relevant information about birds.

Thank you to all my friends who helped me along the way. Jennifer Wohl
is not only my close friend and confidante but contributed to this research project
as my skilled editor. Her knowledge of agricultural economics allowed her to go
beyond basic editing tasks. Having gone through the dissertation process
herself and having known me for a long time, she was able to encourage and
help me when I felt stuck. I would like to thank Julie Stepanek and Susan
Rozanski Jones for lightening the burden by joining me for Latte when we
needed a break from the “D.” Thanks, Julie, for the running times to help

balance body and mind.

vii

I thank my family in Germany for their encouragement along the way.
Most importantly, thank you, John, for always believing in me, for your
unwavering support, and for taking care of our dogs, horses, and home so i

could pursue my scholarly endeavors.

viii

TABLE OF CONTENTS

List of Tables ....................................................................................................... xi
Chapter 1 .............................................................................................................. 1
Introduction to the Research Project and Overview .......................................... 1
Purpose of the Study and Justification .......................................................... 1
Objectives ...................................................................................................... 2
Research Steps ............................................................................................. 2
Organization of Thesis .................................................................................. 5
Important Findings of the Research .............................................................. 5
Chapter 2 .............................................................................................................. 9
Birder Preferences for Attributes of Birding Services: Survey Development,
and Descriptive Results .................................................................................... 9
Introduction .................................................................................................... 9
Survey ......................................................................................................... 10
Survey Development ................................................................................ 10
Structure of Questionnaire ....................................................................... 11
Sample ..................................................................................................... 12
Birder Profile ................................................................................................ 14
Socioeconomic Characteristics ................................................................ 14
Birding ...................................................................................................... 16
Summary and Conclusion ............................................................................ 27
Chapter 3 ............................................................................................................ 30
Optimal Experimental Design For Binary-Choice Experiments ....................... 30
Introduction .................................................................................................. 30
Optimal Experimental Design ...................................................................... 34
Random-Utility Model .................................................................................. 36
Empirical Model ........................................................................................... 38
Application ................................................................................................... 42
Setting initial attribute levels .................................................................... 43
Survey ...................................................................................................... 50
Updating ................................................................................................... 51
Findings ....................................................................................................... 54
Design Comparison .................................................................................. 55
Design Efficiency ..................................................................................... 58
Summary and Conclusion ............................................................................ 62
Chapter 4 ............................................................................................................ 66
Birder Preferences and Biodiversity indicators ............................................... 66
introduction .................................................................................................. 66
Biodiversity Indicators ................................................................................. 71
Empirical Model ........................................................................................... 72
Application ................................................................................................... 76
Selection of Attributes .............................................................................. 76

ix

TABLE OF CONTENTS (continued)

Chapter 4 (continued)
Information for Biodiversity Indicators ...................................................... 77
Computation of Biodiversity Indicators ..................................................... 81
Results ......................................................................................................... 86
Summary and Conclusion ............................................................................ 95
Appendix A ......................................................................................................... 99
Questionnaire .................................................................................................. 99
Appendix B ....................................................................................................... 123
Binomial Logit Regressions ........................................................................... 123
Bibliography ...................................................................................................... 127

LIST or TABLES

Table 1-1: Binary-Choice Scenario ....................................................................... 4
Table 2-2: Socioeconomic Characteristics ......................................................... 15
Table 2-3: Comparison of Birding Expertise ....................................................... 18
Table 2-4: Birding Characteristics ...................................................................... 20
Table 2-5: Comparison of Means between Female and Male Blrders ................ 22
Table 2-6: Birding Partner .................................................................................. 24
Table 2-7: Considerations for Birding Ranked by Importance ............................ 26
Table 3-8: Experimental Design Plan ................................................................. 41
Table 3-9: Sequential Updating of Entrance Fee ............................................... 52
Table 3-10: Sequential Updating of Entrance Fee (continued) .......................... 53
Table 3-11: Estimated Model Results ................................................................. 55
Table 3-12: Difference in Entrance Fees for Alternative Designs ....................... 58
Table 3-13: Normalized D-optimality Scores ...................................................... 59
Table 3-14: Coefficient Variances ...................................................................... 60
Table 3-15: D-Optimality Scores For Selected Fee Differences ......................... 60
Table 4-16: Equivalencies .................................................................................. 81
Table 4-17: Biodiversity indicators for Scenario Alternatives (Scenarios 1- 4) ..85
Table 4-18: Biodiversity Indicators for Scenario Alternatives (Scenarios 5 - 8) .85
Table 4-19: Speannan Rank Correlation Coefficients ........................................ 85
Table 4-20: Binomial Logit Results for Disaggregated Case .............................. 87
Table 4-21: Binomial Logit Results for Alternative Model Specifications ........... 91
Table 4-22: Vuong Statistics for Alternative Model Specifications ..................... 93
Table 8-23: Biodiversity Indicators And Other Variables .................................. 124
Table B-24: Biodiversity Indicators and Entrance Fee ..................................... 125

xi

Chapter 1

INTRODUCTION TO THE RESEARCH PROJECT AND OVERVIEW

Purpose of the Study and Justification

This research explores birdwatchers preferences and attitudes about
birdwatching in an attempt to understand the value of biodiversity to
recreationists.1 The main goal of this research was to develop procedures for
examining the relationship between birder preferences for attributes of birding
sites and measures of biodiversity. A secondary goal was to generate a birder
profile that highlights some of the main characteristics of birders.

The literature on biodiversity is extensive, reflecting the concern over loss
of species as more and more natural areas are converted for development
purposes (Polasky et al. 1999). Natural areas are increasingly managed for
biodiversity and they are also a source of benefit to outdoor recreationists who
enjoy the natural environment for hiking, fishing, camping, and other outdoor
activities (Barbier et al. 1994, Pearce 1993, Burton et al. 1992, McNeely 1988,
Wilson 1988, Margules and Usher, 1981 ). Yet very little is known about what it is
that recreationists value about the natural areas. If part of the goal of preserving
natural areas is to afford recreation opportunities to users, it is important to know
just what it is that recreationists value since the management of natural areas

can and does alter natural area characteristics. Uninformed management may

 

‘ The terms ‘birder' and ‘birding,’ rather than ‘birdwatcher‘ and ‘birdwatching,’ are used in the
remainder of the text.

unnecessarily alter characteristics that are highly valued by birders. This
research explores birdwatchers’ preferences and attitudes about birdwatching in

an attempt to understand the value of biodiversity to recreationists.

Objectives
The objectives of this research are summarized as follows.

1. To develop procedures for examining the relationship between birder
preferences for selected attributes of birding sites and measures of
biodiversity.

2. To compare model efficiency of a new approach to experimental design, an
approach based on optimal “updating” (Kanninen, 1998), relative to a
traditional experimental design that remains fixed throughout a preference
experiment. I

3. To develop a birder profile that describes birders in terms of their socio-

demographic characteristics and their birding activities.

Research Steps

Data were collected through a survey to address the research objectives.
Survey development was a major step in the research process as not much is
known about the relationship between birder preferences for attributes of birding
sites and biodiversity. To learn about birding activities, qualitative pre-survey
interviews were conducted in 1996 and 1997 with birders in the Lansing and

Flint areas. The information was used to identify the salient activities of birders

and to develop an understanding of the attributes of site choice that birders care
most about. The survey data for this application came from personal interviews
that were conducted in 1998 with members of the American Birding Association
(ABA) in the Lansing area.

Once relevant attributes of birding sites were identified, a utility-theoretic
approach was taken to study the relationship between birders’ site choice and
site attributes. Under this approach, birders choose a site from alternative sites
based on the combination of attributes that provides them with the highest level
of utility. Table 1-1 shows a stylized binary-choice scenario. Survey respondents
are asked to choose one of the two alternatives given in the choice scenario.
Alternatives, A and B, are described by m attributes and distinguished by the
attribute levels, the x’s. in this study’s choice experiment, respondents evaluated
eight choice scenarios with two alternatives each.

Site attributes were previously identified through qualitative research.
Attribute levels, the x’s, are typically assigned with experimental designs for all
of the attributes, 1... m, across alternatives, A and B. A recently developed
optimal experimental design for binary-choice experiments was applied in this
study to assign attribute levels across all alternatives.2 The relationship between
site choice, a dichotomous variable, and site attributes was then estimated using
a logit model. The new approach, the “updating” approach, was compared to a
traditional approach to binary choice experiments in terms of their relative

design efficiency (Objective 2).

Table 1-1: Binary-Choice Scenario‘

Question: Which alternative do you prefer, A or B?

 

 

 

 

 

 

 

 

Alternative A Alternative B
Attribute 1 x1“ x13
AttrIbUtO 2 X2A X2a
Attribute rn x...A x"?

 

 

* Two alternatives described by their attributes (1 ...m); attribute levels ()3,
x°)differ across alternatives A and B

The logit model also served to estimate the relationship between site
choice and alternative measures of biodiversity. All but one of the biodiversity
measures are non-linear combinations of some of the x’s that appear in the site-
choice experiment. The choice model was estimated using the full set of x’s and
compared to a choice model based on the biodiversity measures. One of the
biodiversity models is nested in the birder preference model. It was tested
against the disaggregated model using a nested test. The biodiversity models
that are not nested versions of the model for birder preferences of birding sites
were compared to the disaggregated birder preference model by conducting a
non-nested test. The tests were used to determine whether the two models are
equivalent (Objective 1).

Respondents to the survey were asked about their birding activities and

their socio-demographic characteristics to generate the birder profile. Qualitative

 

2 Implementation of the new approach requires the levels of one of the attributes to be updated
and only the starting values were assigned using the experimental design.

4

interviews and evidence from the literature served to supplement the survey data

for the birder profile (Objective 3).

Organization of Thesis

The remainder of this thesis is organized as follows. Chapter 2 describes
the design of the questionnaire as input for a binary-choice experiment of birder
preferences and a birder profile.

The utility-theoretic foundation for the binary-choice experiment of birder
preferences for attributes of birding sites, the random-utility model, and the
binary—choice experiment are explained in chapter 3. The chapter describes the
optimal experimental design, recently developed by Kanninen (1998), that was
used for this research and compares it to a traditional approach.

Chapter 4 estimates a logit model of birder preferences for six attributes
of birding sites: number of warblers, abundance category for warblers, rare or
unusual species, number of other species, habitat, and entrance fee. The model
is then compared with a model that uses aitemative biodiversity indicators as
explanatory variable. The biodiversity indicators were: Shannon’s H, Simpson,

McIntosh, Menhinick, species richness, Margaief, and Berger-Parker.

Important Findings of the Research
This section describes some of the salient findings of the study. These
findings are discussed more fully in the body of the thesis. The birder profile

revealed that the study population of birders is made up of more male birders

than female birders, tends to be married, middle-aged, has relatively high
household incomes, and high levels of education. The study population is similar
in socio-demographic characteristics to the general ABA population of birders
(ABA 1994).

VVIthin the literature, birders are defined in different ways (McFarlane
1996, Kerlinger 1995b, ABA 1994, Boxail and McFarlane 1993, Butler and
Fenton 1986). Birders that include but are not limited to ABA members tend to
have less birding expertise, as measured by the number of birds on a Iifeiist and
by their bird identification skills, than do the ABA birders. The study found that
male birders did not differ significantly from female birders in selected
characteristics. There was a significant difference in the number of species that
can be identified by sight without a field guide and their reported birding
expertise in their home area. The study found significant differences for groups
of different birding expertise in North America in terms of the number of species
on a lifelist, the number of species that can be identified by sight, the number of
bird books owned, and the farthest distance traveled in the 12 months prior to
the survey.

The qualitative research results indicated that birding trips were the most
important component of respondents’ birding activities. Birders weigh many
factors in their decision of which birding site to visit, but bird-related attributes
play a prominent role. Bird-related attributes emerged as important

characteristics for birding activities and the selection of a birding site.

Comparison of model designs revealed that the “updating” approach
resulted in improved statistical efficiency over a traditional approach. The choice
model application to estimating birder preferences indicated that there is a trade-
off between realism of the choice scenarios and design efficiency. in deciding
which approach to use, the time, effort, and cost of tracking the choice
probabilities and updating the continuous variable have to be weighed against
the gains in efficiency.

The estimation of the binary-logit model revealed that all six attribute
coefficients in the birder preference model and all of the biodiversity indicators in
the aitemative biodiversity models were significant and had the expected sign.
No clear preference can be given to any of the seven biodiversity indicators as
all of them were significant and explain birders' site choice about equally well.

A nested and non-nested test was conducted to test the hypothesis that
the model examining birders’ site choice as a function of six site attributes is as
close to the true data as the model specifying site choice as a function of
biodiversity. The tests revealed that the two models are not equivalent. The test
results indicated that the model using six site attributes as explanatory variables
are preferable to the model using biodiversity as an explanatory variable. This
result confirms evidence from other measures of model fit. Overall the results
suggest that the biodiversity indicators used in the study application explain
birders’ site choice well and are a good indicator of birder preference for birding
sites. However, birders care about the composition of birding sites and where

birder preference information for individual attributes of birding sites is

available, it is preferable to an aggregate indicator of biodiversity. The
aggregate measures used in this research do not fully capture the complex

relationship that determines site choice.

Chapter 2

BIRDER PREFERENCES FOR ATTRIBUTES OF BIRDING SERVICES:
SURVEY DEVELOPMENT, AND DESCRIPTIVE RESULTS

Introduction

The main goal of this research was to develop a survey-based approach
to assess the overlap between birder preferences for site attributes and
measures of biodiversity. This chapter reports on the collection of the data on
birder preferences for attributes of birding sites. Data was obtained from a
questionnaire administered to a small sample of birders. The questionnaire
included a sequence of questions that asked individual birders to make pairwise
choices among birding sites with different site characteristics. These questions
support the choice experiment that was conducted to estimate birder
preferences for site attributes and biodiversity (described in chapter 3). The
questionnaire also contained questions regarding the personal characteristics
and experience of the respondent to provide data for the birder profile, a
secondary goal of this research (characterized below).

This chapter describes the questionnaire, the methods used to develop
and administer the questionnaire, and a descriptive profile of the birders
included in the sample. The first subsection explains how the survey
questionnaire was developed from pre-survey qualitative interviews to pretesting

of the questionnaire. The second subsection details the structure of the

questionnaire, and the third subsection describes the sample population. The

description of birders in the birder profile concludes the chapter.

Survey

Survey Development

In the fall of 1996 and 1997, qualitative interviews were conducted with 20
individual birders. The interviews drew on open-ended ethnographic questions
and closed-ended survey-type questions (Fetterman 1998, Sudman et al. 1996,
Johnston et al. 1995, Weiss 1994, Spradley 1979). Participant responses
revealed the terms and concepts birders use when talking about their birding
activities and helped to determine what is important to birders when they pursue
their birding activities. The interviews generated information for the birder profile
and provided a list of attributes to be used in the binary-choice experiment. Bird-
related characteristics were mentioned by the majority of respondents as
positive characteristics of birding activities and birding sites.

The questionnaire was pretested to ensure that survey questions
consistently produce answers that provide comparable information across the
sample and answers generate reliable and valid measures of the question
objectives (Fowler Jr. 1995, Mitchell and Carson 1989). The pretesting took
place in several stages. The survey questions were tested for wording and
understanding as well as clarity in its lay-out, its visual appeal and ease with

which respondents could follow the flow of questions (Sudman et al. 1996,

10

Fowler Jr. 1995). Twenty-one birders were asked to test the choice scenarios.
Responses to willingness-to-pay questions were used to set starting values for
the entrance fee attribute (see Table 1-1 for a stylized choice scenario, setting of

the other attribute levels is described in Chapter 4).

Structure of Questionnaire

The full text of the questionnaire is listed in Appendix A. The
questionnaire contained four data collection sections. The first section obtained
data on respondents’ birding activities. Respondents were asked about their
birding trips, birding skills, birdlists, and participation in birding activities. The
information was used for the birder profile.

The second section elicited data for the choice experiment. Respondents
were provided with detailed information, including definitions, (in writing and
orally) about the choice experiment. Qualitative interviews had revealed the
importance of bird-related attributes of birding sites. The bird-related variables
selected for the choice experiment were: number of warblers, abundance of
warblers, rare or unusual species, and other species. Diversity of habitat was
included as a variable to test for an effect on site choice independent of bird
diversity. An entrance fee served as a payment vehicle. Several feedback
questions were asked to learn how familiar respondents were with the choice
setting and to give them an opportunity to ask clarification questions. Most of the

information served to draw up a consistent choice setting for all respondents

11

(Sudman et al. 1996, Fowler Jr. 1995). The table with equivalent terms was
included in the information sheet to be able to model site choice as a function of
biodiversity measures (see Chapter 4). The choice experiment itself consisted of
eight different choice scenarios that were administered in random order (see
Chapter 3 for a detailed description of the choice experiment).

In the third section, respondents were asked to indicate how important the
six selected site attributes were in their choice of a birding site. This information
and importance ratings of listed statements about a variety of considerations for
birding served as a check of survey respondents’ priorities in birding against
pre-survey data. The importance ratings of listed statements and questions
about respondents’ socioeconomic characteristics were self-administered by the
respondents. The socioeconomic information became part of the birder profile.

The questionnaire concluded with a set of debriefing questions and
interviewer observations. These responses were not analyzed statistically but
served to give the researcher some feedback as to respondents’ degree of
familiarity with the terminology used, degree of perceived difficulty of the site

choices, and their level of experience with entrance fees for birding.

Sam is

Respondents were selected from a list of 85 ABA members in regions 488
and 489 in Michigan. The regions were selected because of their proximity to the

Michigan State University campus in East Lansing. The area extends from South

12

of Clare to North of Jackson and East of Grand Rapids to West of Saginaw and
is referred to below as the central Michigan sample (CMS). The ABA list was
used since ABA members were thought to be experienced birders with a good
understanding of site characteristics (Butcher 1995). The sample size was kept
. small in order to keep interview costs to a minimum while assuring adequate
quality control.

The ABA list provided addresses for 85 members. These 85 potential
respondents were sent a contact letter on official University letterhead (von
Kammen and Stouthamer-Loeber 1998, Salant and Diliman 1994, Diliman 1977).
This letter contained information on the objectives of the study, why the potential
participant was selected, and how his or her name was obtained. The letter
mentioned that the researcher would follow up with a telephone call to set up a
time for an interview. The approximate length of the interview was also indicated.
The contact letter included a written endorsement from the ABA in order to
increase the likelihood that potential respondents would participate and to
reduce any fears associated with admitting a stranger to their home or
workplace.

Of the 85 birders that were contacted by letter, 12 respondents could not
be interviewed because either the contact letter was not deliverable, the person
had moved outside the study area, or an interview could not be successfully
scheduled during the interviewing phase. Three respondents refused to be

interviewed when they were contacted by telephone. Sixty personal interviews

13

were conducted to yield a response rate of 82%. Two interviews were not usable

for the choice experiment because of item non-response.

Birder Profile
Socioeconomic Sharacteristics

Table 2-2 reports the results of the socioeconomic characteristics of the
study population (CMS) and provides results from other birding studies for
comparison. Seventy percent of the birders in the CMS were male. The median
birder age was between 50 and 59 years old. This was the median age category
for male as well as female birders. Thirty-five percent of the CMS birders were
retired. Most birders were married and 46 percent indicated that a family
member was a birding partner with whom they birded most of the time in the past
12 months. The birders tended to be highly educated, with 80 to 85 percent
holding a Bachelor’s degree or higher. Generally speaking, the male birders had
somewhat higher levels of education than female birders, for example, 31
percent of male respondents in the CMS held a doctoral degree as compared to
six percent of female respondents. Approximately 50 percent of surveyed birder

households had annual incomes of $60,000 or more in 1997.

14

...aEmcozaE. Soto» a c. .uoEaEca. as 82852. 35:25 $5 t
.3382 escoEum 26... new £25.83 «.6305 $828 .8 a?! acacia E30 gm muEutgo

.33....on E... 3 .35 29: >552 2 can 32.: 8658323 BEEES 2 2:6 $.88 23..

 

 

 

 

 

$3 5:» A $8 888.83 A
$2 52.8» $8 88.8.6888
$2 88 $2 8-8» $8 88.8.8858
$8 86.8» $8 8.8» $8 88.8.8988
$3 88 $8 8.8....» $8 80.8-8853
8.8» e85 $8 ”88» $8 88.8888»
88 $2 8? 823-8: $888M $8.858m $. 53 v $.. 80.8» v
888.... .83 68» 85 88... 88 £209 68.1209 88.” s. ”058:.
$8 $8A $8 2.8: $8 87 $8-8 $8 $8 9.2888 N
$2 $2 $8 2888
$3 $8 $8 eases.
$8 $8 $8 82058
”.26. 5.833
$8. $3 8 A
$8 $8 8.8
$8 $8 8.8
$8 8 A $8-8 a} $8 $8 3.8
$9 $~. 8.8
$8 $8 8.8
$8 . $8 8 v
t. we 8 8 8.8 =86
. 8a<
$8 ...$8 $8 298
$8 $8 $8 8E8:
832m .282
$8 $8 $8 $8 $8 .88 $8.88 $88 $2. 222
$8 $2. $8 $2 $8 .88 $8.: $8 $8 228m
Bow—30
82uz =82 8uz ...8Tz 882
38: 88: 8th .. 88: 83uz 8»
838x :8: 38: 8.8“. $8: 888%:
8822, ..a 8 Es. 82.8. 8:8 .8888: 288 $8: <m< wzo

 

 

 

 

 

 

 

 

85.82890 Co ..ooEzc uzv 3.3.3.0220 o_Eo:oooo.oom "N8. 63s...

15

mg

Comparisons of research results from the CMS and the ABA membership
survey with results from studies in which a sample of birders was drawn from a
larger population reveal that ABA birders differ from birders more broadly
defined (Table 2-4). Kim et al. (1997) found that ABA birders have more
personal and behavioral investments associated with birding than do visitors to a
bird festival.

The average ABA birder has birded for over 20 years, mostly by himself
or herself but also with friends or family. Surveys that sampled birders from a
wider birder population found that the average number of years of experience
ranged from 5-28 years, depending on the level of birdingexpertise (Boxail and
McFarlane 1993). Seventy eight percent of visitors to Pt. Pelee National Park,
Ontario, had been birding for five years or more and the average visitor had
been birding for 14 years (Butler and Fenton 1986). Most CMS respondents
(95%) kept a lifelist (Wauer [1991] reports 98%). Eighty seven percent of
respondents kept a list of birds they had ever seen anywhere (Wauer [1991]
world list: 40%). Regarding specific locational lists, 53% of respondents
recorded birds seen in the United States, and the same percentage kept lists for
their backyard, 48% of respondents (Wauer [1991]: 57%) had an ABA area list
(which essentially covers North America). A large percentage (60%) of
respondents kept a dayiist where they recorded the bird species they had seen

in the course of a day of birding.

16

CMS participants took an average of 33 daytrips in the 12 months
preceding the survey. Sixty percent of respondents typically take birding
vacations 2-3 times per year. The average number of bird books owned was 61.
But 50% of respondents owned more than 34 bird books. This high average
figure was biased by the large number of books owned by one of the
respondents. The mean number of subscriptions held by respondents was 2.8.
This is slightly higher than the mean of 2.3 subscriptions held by advanced
birders in McFarlane’s (1996) sample. 95% of the CMS respondents kept a
lifelist with a total of 590 species entered on average on a respondent's lifelist.
This number is higher than the median number of species on a lifelist (435)
because of the very high number given by some respondents. Only 21% of
respondents in McFarlane’s sample (1999) kept a lifelist. The mean number of
species on the lifelist across her sample was 48 but it was 363 for advanced
birders (McFarlane 1996).

The average CMS respondent was able to identify 305 species by sight
without the use of a field guide. The median number of species was 254. Eighty
percent of respondents (75% of ABA birders) identified some or most birds by
song or call. in the qualitative interviews, several birders said that they had
hearing problems and therefore did not rely on sound for bird identification.
Other birders may not rely on sound because ‘they are still developing that
identification skill.’

Birders differ by their bird identification skills. Boxail and McFarlane

(1993) found that identification skills increased with the level of birding expertise.

17

They concluded that self-reported skills are a reasonable measure of birding
expertise. Their finding contrasts with research findings from a study of
appreciative wildlife users at Brigantine National Wildlife Refuge. Applegate et
al. (1982) found that visitor groups had a poor ability to judge their own birding
competence. A comparison of birding expertise determined by cluster analysis
(McFarlane 1996) self-rated expertise for the entire sample (McFarlane 1999)
and self-rated expertise for a split subsample (Boxail and McFarlane 1993)
reveals that there are large differences between the percentages for casual,
intermediate and advanced categories (Table 2-3). The samples were not
identical. Boxail and McFarlane (1993) birders were limited to Christmas Bird
Count participants and McFarlane (1996 and 1999) included members of natural

history societies.

Table 2-3: Comparison of Birding Expertise

 

 

McFarlane 1996 McFarlane Boxail/McFarlane"
1999 1993
(787) Grouping by (787) Self-rated (508) Self-rated
cluster analysis split sample
casual 43% 9% 0-8%
novice 38% 40% 28-49%
intermediate 12% 45% 34-51%
advanced 7% 6% 9-22%
" Sample was spit by region. Definitions: casual = rarely watches birds and is

not really interested; novice = interested beginner who wants to learn more; intermediate
and advanced were not defined

The CMS data on birding skills reveal that more birders consider
themselves advanced or expert birders in their home area (67%) than in North
America (37%) (Table 2-4). This observation is also true for ABA birders in

general. The self-reported skill levels are similar across the different categories

18

for CMS respondents and ABA birders, indicating that ABA birders participating
in the CMS are not very different from ABA birders in the rest of the country in
terms of their self-reported birding skill levels. Noticeabie differences appear in
the comparison of the CMS and the ABA survey (ABA 1994) with McFarlane’s
survey (McFarlane 1996) (Table 2-4). McFarlane used cluster analysis on past
experience, centrality-to-lifestyle, and economic commitment variables to derive
four distinct categories labeled casual, novice, intermediate, and advanced
birders. In Table 2-3, the self-reported birding expertise elicited in her survey is
reported next to the data derived from the cluster analysis (McFarlane 1999).
Her sample of Alberta (Canada) birdwatchers is drawn from a heterogeneous
birder population, members of natural history societies, a bird observatory, a
birdwatching club, and participants in the Edmonton Christmas bird count (45%
of CMS respondents had participated in a bird count or survey such as the
Christmas bird count, 20% of ABA respondents were involved in bird research
[Wauer 1991]). The smaller percentage of advanced birdwatchers suggests that
the ABA attracts more highly skilled birders rather than the broad range of
individuals who are interested in observing birds in the wild (Table 2-4).
Qualitative results failed to show a clear relationship between the number
of years a respondent had been birding and the number of species he or she
had on the lifelist. A comparison of means test confirmed the finding for the CMS
participants. The hypothesis that the mean number of species on the lifelist is
the same for birders with 21 years or more of birding experience and those with

less than 21 years of birding experience could not be rejected.

19

.mmmTwNéo cosmoEaesoo _m..o scotch—us. co comma Emu .

 

 

 

 

$o ”.288 $9. ”.288 $v $m 596
$c $5 $3 $3 “385%“
3 $9 $3. $8 BEEEBE
$3 $8 $9 $m 83835553
. «682.28 .85 8820 ”8:251 582
$3 $2 598
$8 $8 83.828
$5 $8 08625....
$v $o @5553 82¢ wEoI
”=28 858
$8 $8 ..888 .3 A25 89: a.
$8 $8 .888 E 88 2:8 o.
$v~ $9 9 Ram; L2 958 32.0“.
$N $m __a 8 958 on: $80
8:28 3 0.
$8 6mm A
non Enos 2% E o.
nonamoEfiooca>uu $8 6.9. A
av :82: our. “:aoE E85 co £58 .02
$8 ”N A
"N ”..on “68:58.8 o.~ ESE 225.5898 62
$8 ”3 A
B N=35 9.25 EB .62
$3 use» $8 mac»
.8 3E: n-~ A can no.5. n-~ A 3628.; EB .oz
3 ”cumE ”unocgua $0... ”or A
otcuoEuusEB 93v an £on 2285 up 5 3:33 62
$8 #m A
NN “:aoE mu ”cuoE 5525 Boo» .oz
nohuz owopnz oonz
83 estates. 32 <m< mzo

Amcozmaomno Co LBEEuzv 8.88.0220 955m "Ya 23a...

20

Differences in birding experience between men and women have been
noted in the literature (McFarlane 1994, ABA 1994). Women rated their bird
identification skills lower than men (CMS, ABA 1994). Of the women in the CMS,
44% rated themselves as advanced and 6% as expert birders in their home area
compared with 52% and 21%, respectively, of the men.

For North America the percentages for women were 17% advanced, 0%
expert; the percentages for men were 38% advanced and 7% expert birder. On
average, female birders could identify 193 bird species without using a field
guide, whereas male birders reported 354 species. More women (44.4%) than
men (16.7%) selected the categories “Don’t use sound at all” or “Follow sound
for visual identification” when asked how much they rely on sound when birding.
McFarlane (1994) found that women, who made up approximately 50% of her
sample, accounted for 60% of casual birders, but only 37% of advanced birders.
But the average female birder had birded for fewer years than had the average
male birder, 22 years as opposed to 26 years (CMS). This is confirmed by
McFarlane (1994), 46% of women had 10 years or less of birding experience as
compared with 36% of men, and 60% of respondents with 5 years or less of
experience were women. Female Michigan birders had also birded on fewer
occasions in a 12 months period (21 daytrips) than had the men (38 daytrips)
(CMS).

A comparison of means test (Table 2-5) shows that there is a significant

difference in means between female and male birders only for the number of

21

birds that can be identified by the respondent without the use of a field guide

and self-reported level of expertise in the home area.

Table 2-5: Comparison of Means between Female and Male Birders

 

 

t-test for Equality
of Means
t Sig. (2-tailed) Mean
Difference

No. on lifelist -1.867 .068 -273.79
Years birding -1.142 .259 -4.37
No. daytrips -1.613 .113 -16.41
No. books -1.304 .198 -21.12
No. subscriptions -1.036 .308 -.46
ID by sight -4.933 .000 -160.82
Expertise home -2.197 .034 -.40
Expertise North -1.963 .055 -.29
America

ID by sound -1.808 .082 -.40

Kellert (1985) found that there was a difference between committed
birdwatchers and casual birdwatchers in their primary reason for birdwatching.
The groups of birdwatchers were defined according to their level of birding
experience as measured by the number of bird species a birder could identify
without a field guide. Committed birdwatchers were defined as birdwatchers who
could identify 40 bird species or more, and casual birdwatchers as birdwatchers
who could identify 10 bird species or fewer. Committed birdwatchers were
primarily motivated by their personal fascination with birds, while casual
birdwatchers cited aesthetics as their main motivation for birdwatching.

These findings suggest that experience affects birder motivations for

birding. Assuming that there is a link between motivation and preferences,

22

Kellert’s results indicate that experience may explain preferences for certain
attributes of birding services.

Comparison of means analysis using independent sample t-tests for the
CMS data showed that there was a significant difference between birders with
different self-reported birding skill levels (beginning birders were grouped with
intermediate birders and advanced birders were grouped with expert birders) in
the number of species on a birder’s lifelist, the number of birds that can be
identified by sight, the number of bird books owned, and the farthest distance
traveled in the 12 months prior to the survey. Not significant were the number of
daytrips in the previous 12 months, the farthest distance ever traveled for
birding, the categories for sound identification (Table 2-4), and the number of
bird magazine subscriptions.

Advanced birders were found to have a primary achievement motivation,
causal birders were primarily motivated by appreciative experiences, novice and
intermediate birders had a primary conservation orientation (McFarlane 1994).
McFarlane (1994) suggests that wildlife management programs could be
designed to meet the needs of the different specialization levels. If primary
motivations are indicative of preferences, McFarlane’s findings suggest that
birder specialization affects preferences.

Questions in the CMS asked birders to rate a number of considerations
on a scale from 1 to 7 by importance for their birding activities. None of the
categories showed any significant differences for beginning and intermediate

birders on the one hand and advanced and expert birders on the other in the

23

 

 

 

 

 

320 $8 52
88.893 $8 85:86 3 $8 8.0 8.93 5.;
.mnEoE
$8 88 93 26.3. 5.;
$8 $ 8 $8 5.88:. 2.8.. 5.;
$8 $8 $3 .322... 5.;
.623
$8 2888.85 5.;
$8 . $8 :35: 3
$2 ”88...
.o >__Em_ M.
3 .o 0:03 8:th 053m
. 8%... awe EN. .83 as
603d E 85 50830 2 8:29. 8880 35:2: NF
89 9.2a .208. as w: s .265 7.8 c. .85 83.:
8.89.28 82 228%.). v8? <m< 220

355.. 9:98 a." as...»

comparison of means analysis. The findings from the CMS may be
interpreted to mean that ABA birders are a more homogenous group of birders
than are birders as defined by Kellert (1985) and McFarlane (1994). They thus
do not show significant differences in responses that can be regarded as
representing different motivations for birding.

Table 2-6 indicates that most birders go birding by themselves. When
they go with others it is mostly with family or friends. Similar results have been
found in other studies.

When respondents were asked in the qualitative interviews about the
positive characteristics of birding experiences, most mentioned seeing a new
bird species, followed by unusual or rare species and many species. These
three characteristics also ranked among the 5 most important considerations for
birding in the CMS. Survey respondents were asked to rate possible
considerations for their birding activities by the level of importance on a scale
from 1 to 7. Table 2-7 lists the considerations and the mean score in order of
importance from the highest to the lowest level. The highest mean score was
assigned to “getting outdoors for a chance to enjoy the natural environment.”
“Being outdoors, appreciating nature” was the most important reason for
participating in the Christmas Bird Count for 20% of surveyed participants
(Boxall and McFarlane 1993). CMS ABA birders also rank “seeing of new bird
species,” “the study of birds in their natural habitat,” “seeing rare bird species,,
and “seeing many different species” very highly. Sixty one percent of visitors to

Pt. Pelee National Park indicated that what they liked most about birding at the

25

Park was “the variety and number of bird species” (Butler and Fenton 1986: 7).
Least important for CMS participants was “competing with other birdwatchers.“
Twenty two percent of respondents indicated that they had participated in a
competitive birding event.

Forty percent of survey respondents (Wauer 1991 reports 41%) had
served as a guide for birding field trips. This is one way to “help others develop
their birdwatching skills,” which was ranked 10‘" among the list of important

considerations for birding with a mean score of 4.5 out of 7.

Table 2-7: Considerations for Birding Ranked by Importance

 

Rank How important are the following N Mean

considerations for my birding:

1 to get outdoors for a chance to enjoy the 60 6.50
natural environment

2 0 see new bird species 60 5.97

3 0 study birds in their natural habitat 60 5.82

4 to see rare bird species 60 5.80

5 0 see many different species of birds 60 5.75

6 o contribute to the conservation of birds 60 5.48

7 o challenge my birdwatching abilities 60 5.25

8 0 add species to a list 60 4.90

9 0 get away from everyday problems 60 4.65

10 0 help others develop their birdwatching 60 4.53
skills

11 o contribute to society's general knowledge 60 4.17
and understanding of birds

12 the social interactions with others 59 4.14

13 o be with family or relatives 59 3.59

14 to photograph, draw, or paint birds in their 60 2.68
natural habitat

15 o compete with other birdwatchers 60 2.12

 

 

26

Birders are typically very involved in birding and/or conservation, as the
number'of memberships in birding or conservation organizations indicates. On
average, CMS respondents belong to 5.6 such organizations. Besides
membership in the American Birding Association, 70% of respondents were
members of The Nature Conservancy and 52% were members of Michigan
Audubon Society, National Audubon Society, and Whitefish Point Bird
Observatory. Sixty percent of Pt. Pelee visitors were members of at least one

birding club (Butler and Fenton 1986).

Summary and Conclusion

The binary-choice experiment of birders’ site choice, described in Chapter
3, is based on an experimental design with six variables. The present chapter
described how the variables were selected using qualitative interviews with
birders and how they were related to biodiversity measures. Survey data was
analyzed to yield a birder profile, which was supplemented and compared with
information from the qualitative interviews and the literature.

For the birders that participated in the qualitative interviews, bird-related
factors were the most important determinants of birding activities. An entrance
fee was selected as the payment vehicle rather than travel costs in order to
avoid having to account for differences in gas mileage of vehicles and types of
accommodation. Other payment vehicles, such as contribution for bird habitat,
were considered but rejected because birders' conservation concerns and

birding activities involved different bird-related considerations. For

27

environmental conservation contributions see Yen et al. (1996), for bird
conservation see McFarlane and Boxall (1996). Abundance categories for bird
species were used to help make the connection between the bird-related
variables and biodiversity measures. Habitat was included as a variable to
determine Whether it has an effect on birders’ site choice that is distinct from the
effect of bird species.

The study revealed a birder profile that shows that birders tend to be
married, middle-aged, somewhat more likely to be male than female, well-
educated, and from a household with relatively high income. Other birder profiles
confirm this general picture. Most birders bird mostly by themselves but many
will also go birding with family or friends. Comparisons with other studies reveal
that ABA birders tend to be more knowledgeable about birds and more involved
in birding than are birders more broadly defined. The average number of species
on an ABA birder's lifelist was 590. Eighty percent of respondents identified
some or most birds by song or call and the average number of species a birder
was able to identify without a field guide was 305. Most birders consider
themselves advanced in their home area and intermediate for North America.
While there are differences between male and female birders in birding
expertise, the only significant differences are that the average female birder can
identify fewer species without a field guide and has a lower self-reported level of
expertise in the home area than the average male birder. Most ABA birders keep

at least one life-list, typically a list of all the birds they have ever seen.

28

The average birder had been birding for 25 years, took 33 daytrips in a
12-months period, and went on birding vacations more 2-3 times per year. In
terms of birder involvement in birding, the average birder belonged to 5.6 bird or
conservation organizations, owned 61 bird books, and had 2.8 bird magazine
subscriptions.

Female and male birders differed significantly only in terms of the number
of species that could be identified without a field guide and by self-reported level
of expertise in the home area. Beginning and intermediate birders differed
significantly from advanced and expert birders in the number of species they had
on their lifelist, the number of birds they can identify without a field guide, the
number of bird books they own and the farthest distance they had traveled for
birding in the previous 12 months. They did not differ significantly in terms of the
average importance rating of a variety of considerations for birding, suggesting
that the ABA birders do not differ significantly in terms of their motivation for

birding.

29

Chapter 3

OPTIMAL EXPERIMENTAL DESIGN FOR
BINARY-CHOICE EXPERIMENTS

Introduction

Valuing the attributes of non-market goods and services, rather than the
goods and services themselves, is becoming an increasingly popular valuation
method (Adamowicz et al. 1998, Freeman 1991 ). A valuation method that can be
used for attribute valuation is that of choice experiments. In a choice experiment,
respondents are presented with well-articulated attribute descriptions of two or
more sites or programs and are then asked which they prefer. In a binary-choice
experiment a scenario is presented to respondents which contains information
on attributes of a pair of sites or programs (Table 1-1). Each site or program
which is part of a scenario is called an ‘alternative’. The alternatives differ by the
levels of at least one attribute. The question arises with this method as to how to
construct the choice scenario such that the respondents’ preferences for
attributes can be accurately modeled. This chapter compares a recently
developed experimental design for binary choice experiments in terms of its
efficiency with a traditional approach.

Experimental designs have been used extensively in areas such as
product marketing (Lazari and Anderson 1994, Kuhfeld et al. 1994, Louviere and
Woodworth 1986) and transportation (Hensher, 1994). However, much of the

existing knowledge generated by this research was based on empirical models

30

that have a continuous dependent variable. Design plans for these types of
models are available in the literature (Addelman 1962, Plackett and Burman
1946, Montgomery 1991). Rubey and Lupi (1997) provide an example of a
discrete choice model application using an orthogonal design. Because
probabilistic choice models, such as the binary-choice model, have a discrete
dependent variable such design plans may not be efficient in model estimation
(Kanninen 1998, Zwerina et al. 1996).

In a traditional approach to choice experiments, the researcher pre-
selects the attributes to be valued and the levels of attributes to be presented to
survey respondents (see, for example, Rae 1985). Attributes and their levels can
be combined in a random combination scheme or more typically involve the use
of a design plan. The experimental design literature offers several methods for
combining the pre-selected attributes and attribute levels (Addelman 1962,
Plackett and Burman 1946, Abdelbasit and Plackett 1983, Montgomery 1991).
Alternative combinations can be created, for example, from the full factorial
combination of each of the levels of each attribute. However, as the number of
attributes and attribute levels increases, this approach quickly results in a very
large number of alternatives. Another approach is a fractional factorial design.
This approach uses only a subset of possible combinations of attribute levels
and thus limits the number of alternatives. Often orthogonal main-effects plans
are used. Orthogonal main-effect plans allow estimation of only the direct effects
of variables and ensure that the variables are independent (orthogonal). These

plans are applicable when the interaction effects between variables are

31

expected to be negligible (Addelman 1962). Orthogonal main-effects plans for
many different model specifications are available (Addelman 1962, Plackett and
Burman 1946, Montgomery 1991).

A traditional binary-choice experiment pre-selects both the number of
attribute levels and the attribute levels themselves prior to the design stage (by
using pretest results, for example). A design plan, eg. an orthogonal main-
effects plan, may be used as a basis for combining attribute levels or as Zwerina
et al. (1996) show an efficient design may be generated using an anticipated
parameter vector. Data are collected without changes in attribute levels during
data collection. Respondents’ actual choices between alternative 1 and
alternative 2 provide the final response probabilities. Model efficiency depends
on how accurately variable coefficients can be anticipated (eg. through pre-
testing).

The traditional approach contrasts with Kanninen’s (1998) optimal
experimental design for binary choice experiments. Kanninen’s model takes into
account a discrete dependent variable and develops criteria for determining the
efficiency of the design. Unlike the traditional approach, this approach allows the
number of attribute levels and the attribute levels themselves to be part of the
design by making them explicit design parameters.

To try and achieve optimality empirically, all variables but one are set at
their upper and lower boundaries and the remaining variable, which must be a
continuous variable, is sequentially updated during data collection in order to

approach the optimal choice probabilities. The number of attributes included in

32

the design determines the optimal response probabilities for the scenario
alternatives. Model efficiency depends on how quickly and how closely the
optimal response categories can be approximated.

In this study, bird-watchers’ preferences for selected attributes of birding
sites provide the application of an optimal binary-choice model design. Kanninen
(1998) developed theoretical criteria for an optimal experimental design in the
case of a binary-choice setting. The empirical model that is estimated with the
data uses sequential updating to approach the theoretical level of model
efficiency. Sequential updating requires the collection of data in multiple stages
and the upward or downward adjustment of, in this case, one variable in order to
approach theoretical efficiency. The goal of this research is to determine
whether the theoretical model is practically applicable and whether gains in
efficiency are possible relative to a traditional approach. The traditional
approach is based on a pre-specified number of levels of attributes and pre-
specified values for attribute levels. Data collection is then typically conducted in
a single stage. Model efficiency is compared using the criterion of D-optimality.
D-optimality is discussed in a later section of this chapter.

The research results show that, as expected, the updated design
approximates theoretical design efficiency more closely than a traditional model.
The application also points to trade-offs between increases in design efficiency
and realism in the choice scenarios. The study application shows that
Kanninen's optimal experimental design yields a higher level of efficiency than a

traditional approach when little information is available about the unknown

33

parameters. Ultimately, a researcher who is faced with the decision on whether
to adopt a traditional approach or the Kanninen approach must decide whether
possible efficiency gains are worth the time and effort that is required for the
multiple stages in updating of the continuous variable.

The remainder of the chapter contains sections on experimental design,
the theoretical and updated models, research application and findings, and the
summary and conclusion. The experimental design section reviews optimal
experimental designs in general and with specific reference to the Kanninen
approach. The theoretical section contains a description of the random-utility
model which forms the theoretical basis for this research. The section on the
empirical model describes the specification of a binary-logit model to estimate
the relationship between birders’ site choice and site attributes. The application
explains how the model was applied to birders’ preferences for attributes of
birding sites. The study findings describe the research results and the chapter

concludes with a summary and conclusion section.

Optimal Experimental Design

An optimal experimental design maximizes the statistical information that
can be obtained from the data on the basis of design criteria. An experimental
design that minimizes the covariance matrix will require fewer observations than
would be necessary in the absence of design criteria. Optimal designs are
therefore of special interest to researchers who are trying to estimate models

with small sample sets.

Standard optimality measures are “A-efficiency” and “D-efficiency" (or D-
optimality')(e.g. Zwerina et al. 1996, Alberini 1995, Kuhfeld et al. 1994). Both
measures can be derived from the covariance matrix. A-efficiency minimizes the
average variance of the parameter estimates. A-efficiency has the drawback that
its measure changes with recodings of the design matrix (Zwerina et al. 1996).
D-optimality is achieved by maximizing the determinant of the Fisher information
matrix3 (Kanninen 1998).

D-optimality has become a commonly used design criterion (Kanninen
1998, Zwerina et al. 1996, Kuhfeld et al. 1994, Alberini, 1995) and may be
generated by using computer-based algorithms (Zwerina et al. 1996, Cook and
Nachtsheim 1980). Kanninen (1998) analytically developed the requirements for
D-optimality of binary-choice experiments.‘ Unlike other approaches, the number
of attribute levels as well as the levels themselves are part of the design.

Kanninen’s solution for a D-optimal design depends only on the number
of attributes included in the design. Once the number of attributes has been
selected, maximization of the determinant of the Fisher information matrix will
yield the choice probabilities necessary for D-optimality (Kanninen 1998).

D-optimality for binary-choice models with a binary logit specification and
negligible interaction effects requires only that all attributes but one be set at
their extreme values following an orthogonal main-effects plan. The number of

attributes determines the necessary choice probabilities to meet design

 

3 The Fisher information matrix inverted and its sign reversed yields the covariance matrix for
the maximum likelihood estimators.
‘ Zwerina et al. (1996) generate Doptimal designs by minimizing D-error using an algorithm.

35

efficiency. Attribute levels of the updating or balancing variable, the variable
whose levels were not set by the main-effects plan, are selected to obtain the
required choice probabilities. The design-balancing variable must therefore be a
continuous variable and Kanninen (1998 p. 4) suggests that price “should be
able to take any value that some portion of the experimental respondents might
consider reasonable for a particular choice being offered".

The conditions Kanninen derived for D-optimality in a binary-choice
experiment require a researcher to determine the upper and lower bounds for
attribute levels to be included in the experimental design and the continuous
variable that must be sequentially updated in the course of data collection to try

and attain the choice probabilities that achieve theoretical D-optimality.

Random-Utility Model

The random-utility model forms the utility-theoretic underpinning for the
research application. The random-utility model is a model that emphasizes the
choice among a set of alternative bundles of attributes for a single choice
occasion. In the context of recreational site choice, the model is particularly
suitable when individuals choose between sites with different attribute levels
(Bockstael et al. 1991). Thus, it is possible to model resource users’ site choice
as a function of the differences in site attribute levels.

An indirect utility function, V, for a resource userj choosing site A can be

defined as:

VIA(QWMI 43»)

36

where:

Q, = a vector of specific attributes of birding sites
P“\ = the cost associated with visiting site A
M,= income of user j

All terms of the utility function that do not vary across scenarios do not
influence the individual’s choice between alternatives and are therefore not part
of the consumer’s choice calculation when the utility function is linear in the
parameters. M, therefore, drops out of the model. In the application presented
below, all costs except the entrance fee for visiting possible birding sites are
assumed to be equal across alternative sites and the difference in cost of

visiting, Pr P, is therefore equal to the difference in entrance fees between

site A and site B as faced by individual j. For the remainder of this chapter the

entrance fee will be folded into the site atttributes and thus P, and P], will
become part of the vector QIA and 0,3.

In the random-utility model, it is assumed that an individual selects the
alternative that maximizes his or her utility. Thus, in a binary-choice setting,
when a resource user faces the decision of which site to visit from a set of two
possible alternatives, a constrained utility-maximizing resource user chooses the
site that gives the higher level of utility. Individual j will choose site A over site B
if:

V,(O,,) exceeds Vamp).

37

An error term, e, is added to the model for estimation purposes. Individual j will
then choose site A if:

V”(Q,)+eu >V,(Qp)+ep. Equation1

The probability of individual j choosing site A can be expressed as:

P(A)=P{v,,(o,)+e, >v,(o,,)+e,,}. Equation2

The empirical model described in the next section translates the

theoretical model above into a model one can estimate econometrically.

Empirical Models

The model is estimated using a binomial logit model specification for a
linear utility function. The utility function is assumed to have random errors with
an extreme value distribution. Following these assumptions, the probability of

respondentj choosing alternative A is given by:

P(Al) = Pl .. fl

_ 1+ EXP(6j) Equation 3

Typically, it is assumed that the vector of attributes of the birding sites, Q,

enter the utility function linearly and thus:

9] = B1Iq1jA ' Q1jBI+ [52(9sz " QZjB)+-°'+Bm(Qn1A ’ (1ij ) Equation 4

(qw‘ - q,,f)represents the difference in the level of site attribute m between site

A and site B faced by respondent j.

 

5 The notation and much of the exposition in this section follows Kanninen (1998).

38

The model is estimated by maximizing the log-likelihood function:

InL = 2y, lnP(A,)+1- y,)ln(1—P(A,)) Equation 5
1:1
y , = 1 when respondentj prefers site A and =0 otherwise.

D—optimality is achieved by maximizing the determinant of the Fisher
information matrix. The resulting value will be referred to as the “D-optimality

score” in the remainder of the chapter. The Fisher information matrix is:

 

 

(
2w, Zwlx" 2‘”;an \
r= Zw,x,f _ Zwif‘x" Equation 6
o -. :
\ Zwlxuzl
where:
wI = P,(1— P,), x.] = q,“ - q,I3 . Equation 7

The determinant of the information matrix is then computed as:

Det(l) = z Z . . . Z Z ijl. . .Wklx.u_g ' X.IJJ| Equation 8
i=1 j=I+1 I k=I+1
The summation covers all combinations, i ...k, of the total observations, n, and

X '1, 1..-: is a matrix of those combinations of observations.

Kanninen (1998) has demonstrated that this can be rewritten as:

Det(l) = [Eda—21:?) g§1m Elk-211mwruw, '9‘;ij '9‘”; Equation 9

9, =0:1 +131x1 +Bzx2+...+j3mxm , 91:8“, forj = 2, m, and 9,“, =1.

39

9'.“ is a matrix composed of combinations of observations on the 9 '5.

Equation 10 must be maximized to achieve the optimal solution. In this

expression, the determinant alone would be maximized when thee ’5 take on

extreme values. Since the empirical model is based on attribute level differences
and an underlying utility function that is linear, these differences should be as
large as possible. This can be achieved by choosing the opposite extreme

values for the attribute levels in the choice scenarios according to a

2" orthogonal main-effects experimental design plan. Maximization of

PJ.(1- P,)alone would require P}. to be equal to 0.5. Thus the maximization must

find a balance between these opposing forces. The solution can be derived
mathematically and depends on the number of attributes used. (Kanninen 1998)
There are similarities between the maximization solution and the
2" orthogonal fractional factorial experimental designs. For all but one of the
attributes, the standard experimental design plan can be used to set attribute
levels at their extreme values. Experimental design plans may use different
notation to indicate alternative attribute levels. For a design with attributes of
only two levels pluses (+) and minuses (-) may be used or positive ones (+1) and
negative ones (-1) to distinguish between the two possible attribute levels. Thus,
the upper bound may be set when the design calls for (+) or (+1), and the lower
bound when the design calls for (-) or (-1). The remaining attribute is used to
balance the design to achieve the calculated response probabilities. Optimal

choice probabilities depend only on the number of attributes.

4O

Table 3-8 illustrates the use of an orthogonal main-effects plan for the
study application. A 2‘ design plan (Montgomery 1991) was selected for six
attributes with two levels. The design plan was used for alternative A in each of
the eight scenarios. Alternative B was created by selecting the opposite attribute
level from that given for alternative A. Each of five variables was assigned to a
column. The sixth column is labeled ‘split’ to indicate that the balancing variable
(entrance fee) will be used to try and approximate the optimal 70-30 split in
choice probabilities for a 2‘ design. A (+) corresponds to a desired 70% choice

probability and a (-) corresponds to a desired 30% choice probability.

Table 3-8: Experimental Design Plan

 

 

 

 

 

 

 

 

Scen‘ Alt‘ Warblers Abundance Other Rare Habitat Split
1 . A - - - + + +
B + + + - - -
2. A + - - - - +
B - + + + + -
3. A - + - - + -
B + - + + - +
4. A + + - + - -
B - - + - + +
5. A - - + + - -
B + + - - + +
6. A + - + - + -
B - + - + - +
7. A - + + - - +
B + - - + + -
8, A + + + + + +
3 - - - - - -

 

* Scen = scenario; Alt = alternative. Design based on Montgomery (1991)

41

This optimal binary-choice design still requires some knowledge of the
attribute levels in order to set all but one of the attributes at their boundary
values and to obtain starting levels for the balancing variable. However, this
approach to experimental design is more flexible than traditional approaches
with pre-selected variables because it uses updating of the balancing variable to

approach the optimal response probabilities.

Application

For the present research, Kanninen’s (1998) optimal experimental design
approach was applied to a binary-choice experiment in which birders were
surveyed regarding their site and program preferences. Qualitative research
(see Chapter 2) revealed the main attributes of birding sites that birders care
about. Among these, six attributes were selected for the choice setting: the
number of warbler species, the number of rare or unusual species, the number
of other species, abundance category of warblers (indicating how likely birders
are to see a species), habitat diversity, and site entrance fee as the balancing
variable.

The next section discusses how qualitative interviews were used to help

set the initial levels of each of the attributes.

42

Setting initial attribute levels

The optimal design requires boundary levels to be set for five of the six

attributes and starting values for the balancing attribute, the sixth attribute.

Bird-Related Attributes

Birders are usually familiar with bird checklists, which are available for
many popular birding sites. These checklists present the names of all the
species, often with abundance categories, that have been sighted at particular
birding locations. Entire checklists could not be used for the research as it would
make the experimental design process too complicated. However, in order to
maintain some of the complexity of information contained in a checklist, warblers
and rare or unusual species were considered separate attributes from the
general attribute of “the number of bird species.” This specification was based
on the finding that most respondents have a particular preference for warblers,
and for species that are rare or unusual for a specific area in general.

Bird lists from birding sites in Michigan and Pt. Pelee, Ontario, were used
to set attribute levels (Boettner et al. 1983, Evers and Granlund 1991, Friends of
Point Pelee and Pelee Island Winery 1994, Graham 1997, Jones 1990, Lerg
undated, McWhirter 1997, Smith et al. undated, The Nature Conservancy undated,
US. Department of Interior 1991 and 1979, Sault Naturalists undated). The number
of species in the “warbler“, “rare or unusual”, and “other species" categories

were counted for each list to learn what a realistic range of species numbers

43

would be for birding sites that survey participants might visit. The lists also
provided information about abundance categories.

Some lists gave detailed information such as the seasonal distribution of
each species (Graham 1997). The type of information varied across lists, but
most lists provided at least the names of species and their abundance category.
The abundance categories were either the set consisting of “abundant,”
“common,” “uncommon,” “occasional,” and “rare,” or the set consisting of
“common,” “uncommon,” and “rare.” Other categories used in some lists included
“accidental,” and “vagrant.”

The bird lists that were used for the present research are the lists for
Point Pelee, Ontario (Friends of Point Pelee and Pelee Island Winery 1994,
Graham 1997); Whitefish Point (Evers and Granlund 1991), Seney Wildlife
Refuge (US. Department of Interior 1991, Jones 1990), Shiawassee Wildlife
Refuge (US. Department of Interior 1979), Rose Lake (Lerg undated), Lansing
Area (McWhirter 1997, Berrien County (Smith and Witkoske undated), Sault Ste.
Marie Area (Sault Naturalists undated), Genesee County and For-Mar Nature
Preserve and Arboretum (Boettner et al. 1983), and Grand Mere. The list for
Grand Mere was constructed from a listing of species for Grand Mere and
abundance categories taken from a list for Berrien County (Grand Mere forms
part of the Berrien County list).

The total number of species listed for a given site ranged from 166
species at For-Mar to 357 species at Pt. Pelee. Five lists were analyzed in more

detail, Whitefish Point, Seney Wildlife Refuge, Shiawassee Wildlife Refuge,

44

Rose Lake, and Grand Mere. When the lists were split into “early spring” and
“late spring,’ the range was 104 to 224 species. The lowest total number of
species was set at 85 and the highest at 274 species. This range covers a
realistic range of number of bird species that may occur in one day at a site in
the spring.

The range for the number of rare or unusual species was selected based
on the five lists that were studied in more depth. An experienced birder provided
information to determine which species would be considered rare or unusual in
the study area in Michigan (Johnson). The number of rare or unusual species for
each of the five sites ranged from 7 to 38 in the early spring season. Because
warbler species appear in a separate category, no warbler species should be
included in the count of rare or unusual species. A range of 5 to 30 is therefore a
reasonable range for the number of rare or unusual species.

The number of wood-warbler species found at the five sites ranged from 4
to 32. Since rare or unusual warblers would also be included under rare or
unusual species, they were eliminated from the list. A range of 5 to 25 warbler
species (not including rare or unusual species) was selected for the scenarios.

Bird checklists may differ in the type of abundance information they
provide for bird species listed at a site. Most checklists use abundance
categories for spring, summer, fall, and winter. But they may also provide more
detailed additional information such as the number of birds observed at different
times of the year (Evers and Granlund 1991). Since the abundance categories

“rare”, “uncommon”, and “common" were available or could be constructed for all

45

sites, the boundary values for abundance category were set at “rare“ and
“common“ for the choice setting. Because “rare“ was also used to designate
species not typically found in the area, some confusion between the two
concepts was expected and respondents were alerted to the two different
meanings of “rare“. A change in terminology was contemplated but rejected
because this term is commonly used for both concepts and birders are familiar

with it.

Habitat

Habitat was designated as “forest with edge“ habitat for the low,
homogeneous habitat level and as “forest, wetlands, open water with sandy-
gravel shoreline“ for the high, diverse habitat level. The homogeneous habitat
was selected as forest with edge because warbler species are specifically
included in the scenarios and they require forest habitat.

Estimation of main effects requires that there be no significant interaction
effects between the selected variables. In order to create scenarios with a high
total number of species and with homogeneous habitat as a realistic alternative,
edge habitat, where one habitat type changes to a different type, was added to
the forest habitat. It was felt that respondents might reject a scenario in which a
high number of species dwells in a habitat of only forest. Forest habitat and a
transitional habitat from forest to some other habitat(s) increase the probability

that a diverse number of species can be found.

46

While most of the birders who were interviewed did not have a problem
conceiving of low habitat diversity and a large number of species, a few birders
commented that this combination of attribute levels was unrealistic. However,

they were still able to perform the required choice task.

Entrance fee

While public birding sites usually do not have an entrance fee, some
public sites and most private sites that birders visit have entrance fees. Entrance
fee is therefore not an unreasonable variable to include in an estimation of
birders’ choices among sites or program attributes.

Pre-tests were conducted with 20 birders to determine how the fees
should initially be set. Eight of the pre-tests included testing of the entire survey
instrument. Pre-test participants were asked which of the two sites in scenario 8
they preferred, what their maximum WTP for each of the two sites was, and how
high the fee for their preferred site would have to be set for them to switch to the
other site.

Scenario 8 was chosen because it has all of the attributes set at the high
level for one site and all of the attributes set at the low level for the other site. It
is reasonable to assume that if all attributes are “goods“ rather than “bads,” then
all respondents would prefer the site with all high levels and would provide a
higher WI'P than for any of the other scenarios. This provided an upper limit to

WTP for birding sites in this study.

47

The question about the dollar amount for the preferred site that would
make a respondent switch to another site was included in the pre-test to
determine whether respondents had tmly given their maximum WTP for the
preferred site. In almost all cases, the respondent gave a higher amount for the
switch price than they stated as their maximum WTP. Therefore, the switch price
reflects more accurately respondents’ “true“ maximum WTP.

In pre-test interviews 9-17, which were conducted by telephone,
respondents had previously been sent by mail (e-mail or regular mail) 3 or 4
scenarios with pairs of birding site descriptions. Each respondent received
scenario 8 and 2 or 3 additional randomly selected scenarios. Respondents
were asked to report their maximum WTP for each site and, given their WTP,
which site they preferred.

Respondents were also asked the entrance fee (dollar amount) for
scenario 8 site A (all high attribute levels) at which they would switch to site B
(all low attribute levels). In 6 of the 9 cases, dollar amounts higher than the
previously stated maximum WTP were indicated. In three cases, respondents
indicated that they would prefer to not go birding if the entrance fee exceeded
their maximum WTP.

In three telephone pre-test interviews, respondents were asked which site
out of each of three pairs they would prefer if both sites charged an entrance fee
of $10. The fee for the preferred site was subsequently raised and/or the fee for
the non-preferred site lowered until the respondent chose the initially non-

preferred site.

48

The above information was used to set entrance fees used in the survey
instrument. Birders in the pre-test indicated that most sites they visit for birding
do not charge an entrance fee. The birders who had paid entrance fees for
birding sites such as state parks, Pt. Pelee National Park, Detroit Metro Park,
stated that they consider the fees nominal. Most fees reported by birders do not
exceed $5 per day. The daily fees would be even lower for senior citizens and
when seasonal or annual passes are purchased and several trips are made to
the same site. For this reason, the fee for the site that the pre-tests indicated
was the less desirable site was set at $1. The adjustment to generate a 70-30
split therefore required adjusting the entrance fee for the more desirable site. If a
site was clearly preferred by pre-test respondents and had a high number of
variables set at the high level, the fee was set at a high level. The highest level
chosen was $25 for scenario 8 site A. This site has all variables set at the high
level and a targeted choice probability of 70% of the respondents. The highest
reported WTP for this site ranged from $0 (respondents were not willing to pay
an entrance fee to go birding) to $75. The other high levels that were set for
entrance fees ranged from $5 for a site that was not clearly preferred by pre-test
respondents but required a 70% choice probability, to $21 for two sites that had
more than half of the non-fee characteristics set at the high levels, and were
preferred by all respondents in the pre-test. One of the sites required a 70%
choice probability and the other a 30% choice probability.

Only a few pre-test observations were available for all scenarios, with the

exception of scenario 8. The researcher had to use some judgment to set the

49

initial entrance fees. In only one of the scenarios was a higher fee set for the site
alternative that ended up with a lower fee at the end of data collection. In all
other scenarios, the site alternative that started with a higher entrance fee also
had the higher fee after the interviews were concluded. For the single scenario
where the fees moved in the opposite direction from what was anticipated,
birders had not shown a clear preference for one or the other site alternative in
the pre-tests. This illustrates that the pre-tests were a good indicator of site

preference.

Members of the American Birding Association listed under areas 488 and
489 in the membership directories for 1996, 1997, and 1998 were sent a letter
describing the research and informing them that they would be contacted to
participate in a personal interview (American Birding Association 1996, 1997,
1998). Members whose telephone numbers were not available were asked to
return a postcard with their telephone number. Sixty members of the American
Birding Association in the Lansing area were personally interviewed. Adjusting
for birders who had moved out of the area, had not returned a postcard with their
telephone number, individuals who were unable to schedule an interview during
the interviewing period, and three individuals who refused to be interviewed

when contacted by telephone, the response rate was 82%.

50

For details about the questionnaire see chapter 2. The questionnaire was
administered the same way to each respondent. Respondents’ clarification
questions were handled in a uniform manner. Each respondent was presented
with all 8 pairs of alternatives and asked to choose between the two sites in
each pair. The order in which the scenarios were presented to the respondents

was randomly selected.

Updating

After each interview, or after several interviews that were scheduled
closely together, the entrance fees presented in the scenario alternatives and
the choice of birding sites from each pair were recorded in a table. Table 3-9
and Table 3-10 is the actual table used for the updating process. The table
headings included not only the scenarios (indicated by a number) and
alternative sites (labeled either A or B), but also the targeted choice
probabilities. The rows beginning with ‘I’ followed by a number represent the
different interviews. Cell entries correspond to the entrance fees that were
presented to the interviewee and bold figures designate the respondent’s choice
of the preferred alternative. Rows beginning with ’%’ show the actual choice
probabilities for the interviews that were conducted. The percentages (in bold)
are calculated based on the number of respondents out of the total number of
interviewed respondents who chose a given alternative. As long as the trend

was in the direction of the desired 70-30 split, the entrance fees were

51

Table 3-9: Sequential Updating of Entrance Fee‘

 

 

 

 

 

 

 

 

 

 

1A 18 2A 28 3A 38 4A 48 5A 58 6A 68 7A 78 8A 88
7O 30 7O 30 30 7O 30 70 30 70 3O 70 7O 30 7O 30
% % % % % % % % % % % % % % % %
l1 1 21 1 16 1 21 1 1 1 1 5 1 1 1 1 1 1 25 1
I2 1 21 1 16 1 21 1 1 1 1 5 1 1 1 1 1 1 25 1
I3 1 21 1 1 6 1 21 1 1 1 1 5 1 1 1 1 1 1 25 1
I4 1 21 1 1 6 1 21 1 1 1 1 5 1 1 1 1 1 1 25 1
I5 1 21 1 1 6 1 21 1 1 1 1 5 1 1 1 1 1 1 25 1
I6 1 21 1 16 1 21 11 1 1 5 11 1 1 16 25 1
I7 1 21 1 16 1 21 11 1 1 5 11 1 1 16 25 1
I8 1 - 21 2 12 2 1 7 1 1 1 3 5 1 1 1 1 20 25 1
I9 1 21 2 12 2 17 1 1 1 3 5 11 1 1 20 25 1
I10 1 21 2 20 2 1 7 20 1 1 O 5 1 1 1 1 20 25 1
I12 1 21 2 20 2 17 20 1 10 5 1 1 1 1 20 25 1
I1 3 1 21 2 20 2 1 7 20 1 1 O 5 1 1 1 1 20 25 1
I14 1 28 2 25 2 17 30 3 1 6 3 1 8 2 1 30 25 1
I15 1 28 2 25 2 17 3O 3 16 3 18 2 1 3O 25 1
I16 1 28 2 25 2 1 7 30 3 20 3 1 8 2 1 30 25 1
I1 7 1 28 2 25 2 1 7 30 3 20 3 1 8 2 1 30 25 1
. I1 8 1 28 2 25 2 1 7 30 3 20 3 1 8 2 1 30 25 1
I1 9 1 28 2 25 2 1 7 3O 3 20 3 1 8 2 1 30 25 1
I20 1 28 2 25 2 17 30 3 20 3 18 2 1 30 25 1
l21 1 28 2 25 2 17 3O 3 20 3 1 8 2 1 30 25 1
I22 1 28 2 25 2 17 30 3 28 _ 3 25 2 1 30 25 1
I23 1 28 2 25 2 17 30 3 28 3 25 2 1 30 25 1
l24 1 28 2 25 2 17 30 3 28 3 25 2 1 30 25 1
I25 1 28 2 25 2 17 30 3 28 3 25 2 1 30 25 1
I26 1 28 2 25 2 17 30 3 28 3 25 2 1 30 25 1
I27 1 28 2 25 2 17 30 3 28 3 25 2 1 30 25 1
I28 1 30 2 27 2 1 7 3O 3 28 3 25 2 1 30 28 1
|29 1 30 2 27 2 1 7 30 3 28 3 25 2 1 30 28 1
I30 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
$6 52 48 38 62 31 69 45 55 52 48 52 48 55 45 80 20

 

‘ Heading numbers designate scenario and letters the alternative within the scenario. The
number in the first column indicates the number of the interview. The second heading row
represents optimal response probabilities. Cell entries are the entrance fees that were presented
to the respondents in the scenarios. The percentages (in bold) are calculated based on the
number of respondents out of the total number of Interviewed respondents who chose a given
alternative (indicated by bold cell entry).

52

Table 3-10: Sequential Updating of Entrance Fee (continued)

 

 

1A 1 8 2A 28 3A 38 4A 48 5A 58 6A 68 7A 78 8A 88

70 30 70 30 30 70 30 70 30 70 30 70 70 30 70 30

96 96 96 96 96 96 96 96 96 96 96 96 96 96 96 9%
I31 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I32 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I33 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I34 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I35 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I36 . 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I37 1 30 2 27 2 17 30 3 28 3 25 2 1 30 28 1
I38 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I39 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I40 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I41 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I42 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I43 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I44 1 30 2 27 2 20 30 3 28 3 28 2 1 30 28 1
I46 1 30 1 30 2 20 30 3 28 1 29 1 1 30 28 1
I47 1 30 1 30 2 20 30 3 28 1 29 1 1 30 28 1
I48 1 30 1 30 2 20 30 3 28 1 29 1 1 30 28 1
I49 1 30 1 30 2 20 30 3 28 1 29 1 1 ~30 28 1
I50 1 30 1 30 2 20 30 3 28 1 29 1 1 30 28 1
l51 1 30 1 30 2 20 30 3 28 1 29 1 1 30 28 1
I52 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I53 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I54 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
l55 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I56 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I57 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I58 1 21 2 12 2 17 11 1 3 5 11 1 1 20 25 1
I59 1 30 1 30 2 24 30 1 28 1 29 1 1 30 30 1
I60 1 30 1 30 1 28 30 1 30 1 30 1 1 30 30 1
96 52 48 33 67 24 76 51 49 41 59 54 46 53 47 83 17'

 

 

 

 

 

 

 

 

maintained. If they were moving in the opposite direction, the entrance
fees were changed by increasing the fee of the site that appeared to be selected

at a higher-than-desired percentage.

53

Only two pre-test respondents indicated a willingness-to-pay of more than
$25 for the birding site that had all high levels of non-fee attributes. It was
determined that an entrance fee of $30 was close to the limit of realistic fees.
This decision was based on the pre-test results for WTP and the fact that birders
do not have to pay for most of the birding sites they visit, especially ones located
within day-trip driving distance.

The entrance fees were updated a total of 11 times for a total of 30
scenarios. Individual scenarios were updated between two and six times. By the
time all 60 of the interviews were conducted, prices for one of the alternative
sites in each scenario had reached the $30 threshold for all but one scenario
(Table 3-9). The scenario that had not reached the threshold was close to the
desired target split with choice probabilities of 76% and 24%. Three of the
scenarios had splits that were moving in the opposite direction of the desired
split. One scenario came close to achieving the opposite choice probabilities for
the alternative sites with 67% of the respondents selecting the site that was

supposed to be selected 30% of the time.

Findings

A binary logit model was estimated using LIMDEP Version 7.0 (Greene,
1995) to obtain coefficients for the site attribute differences (Table 3-11). The
binary logit model estimates the probability of choosing site A given the
differences in the independent variables. The variable differences are defined as

attribute level A minus attribute level B.

Table 3-11: Estimated Model Results

 

 

 

Variable Estimated Coefficient
(Standard Enor)
No. of warblers 0.033
(0.007)
No. of rare species 0.027
(0.005)
No. of other species 0.0057
(0.001)
Abundance of warblers 0.27
(0.10)
Habitat 0.23
(0.10)
Entrance fee -0.029
(0.0074)
-Log Likelihood Fct. 293.15
% Correctly Predicted 64%

All of the estimated coefficients were significant (95% confidence) and
had the expected sign (Table 3-11). The probability of choosing a site increases
with the number of warblers, rare or unusual species, other species, higher
abundance of warblers (as indicated by abundance categories), higher levels of
diversity of habitat, and decreases as the entrance fee to the site is increased.

[The estimation results will be discussed further in Chapter 4.]

Design Comparison

One of the goals of this study was to compare the efficiency of a
traditional binary-choice experiment with that of the approach developed by
Kanninen (1998). Model efficiency depends on the unknown parameters, the

(3's. The best and only estimates we have for the [5's are the binary-logit model

55

estimates, 8. Therefore, in the model comparisons B is used to compute model

efficiency.
Kanninen developed her optimal design for binary—choice experiments
analytically. The theoretical level of design efficiency cannot be reached with the

empirical model because the B'sare not known. However, using the updating

approach, efficiency of the empirical model will approach the theoretical model
efficiency. In the application described here, efficiency improvements were
constrained.

The theoretical optimal design assumes a linear utility function which
resulted in the requirement of setting the attribute levels at their extreme
boundaries. In one of the scenarios, respondents were asked to select a site
from two alternatives where the boundary values for one of the site’s attributes
made the site so desirable that the entrance fee would have had to be set so
high to achieve the desired response probabilities to increase the fees beyond
any respondent’s experience. The alternative sites would have required a
difference of $72 in entrance fees to achieve the optimal 70-30 split. It is likely
that many respondents would have rejected a scenario with entrance fees at
levels beyond respondents’ experience. While the entrance fee differences that
would have been required to obtain the desired choice probabilities in the other
cases were lower, they ranged from $24 to about $59 (Table 3-12), the $30 cut-
off point was selected for this study to maintain the realism of the choice

scenarios.

56

There are three design matrices to evaluate, the theoretical design matrix
with the optimal entrance fees, the updated design matrix with the actual
entrance fees presented to respondents, and the traditional design matrix with
pre-selected entrance fees. For the theoretical design matrix the fee differences
were calculated that would have been necessary to obtain the desired choice
probabilities for site A in each of the scenarios (Table 3-12). The coefficient
estimates, the attribute differences, and the desired choice probabilities for
selection of site A were entered into a spreadsheet. The fee difference was
adjusted until the desired choice probability was obtained using the following

logitequation:

(£5.11)
PM) = —°—.——.

(2 PIN)
1+eH

Prior to data collection, two entrance fees, $1 and $11, were selected as fees
that would have been used had a traditional approach been taken.7 The
selection was based on pre-test information about birders’ maximum WTP for
visiting the alternative birding sites. These entrance fees were entered into the
experimental design to calculate fee differences for each scenario (Table 3-12).
The difference in average absolute entrance fees and the difference in entrance
fees that resulted from the updating process of the empirical model are provided

in Table 3-12 for comparison.

 

7 The difference in the pre-selected entrance fees for the traditional design was 10. Therefore -10
or +10 were entered into the design matrix according to the design plan.

57

Table 3-12: Difference in Entrance Fees for Alternative Designs

 

 

Scenario Theoretical Final Fee for Average Fee for Traditional
Model Updated Model Updated Model Model

1 -54.67 -29 -26.50 10
2 -72.15 -29 -23.00 10
3 -24.15 -27 -17.33 -10
4 51.24 29 23.93 -10
5 37.05 29 18.24 -10
6 52.67 29 20.62 -10
7 -49.23 -29 -25.84 10

8 59.25 29 25.88 10
Design Efficiency

To compare the efficiency of the updated model with the traditional model
relative to the theoretical model, D-optimality scores were computed for all three
models. It was expected that the theoretical model would have the highest D-
optimality score reflecting the fact that Kanninen derived the efficiency criteria
for the theoretical model. The D-optimality score for the updated model, the
model that uses the collected data, was expected to be lower compared to the
theoretical model because the optimal entrance fees were not known. By
updating the balancing variable the model will approach the theoretical level of
optimality. The closer the final choice probabilities will be to the optimal choice
probabilities the closer the updated model will be to the theoretical D-optimality
score. It was expected that the D-optimality score for the updated model would
exceed the D—optimality score for the traditional model. The choice probabilities
were adjusted for each scenario of the updated model while the two entrance

fees that were selected for the traditional approach were assigned according to

58

the main-effects plan for all scenarios. The D-optimality score for the traditional
model was then computed using the pre-selected entrance fees. The efficiency
of the design will depend on how the entrance fee is set. More extensive pre-
testing is Iikely to improve design efficiency. The D-optimality results are

presented in Table 3-13.

Table 3-13: Normalized D-optimality Scores

 

 

 

 

 

Model D-optimality Score‘3
Theoretical model 1
Updated model .91
Traditional model .25

 

The results show that the D-optimality scores of the theoretical model and
the updated model are fairly close. The updated design matrix achieves 91%
efficiency relative to the theoretical design matrix. This suggests that while an
optimal design cannot be reached with the estimated updated model, it is
possible to come close to the optimal results by updating the balancing variable.
Comparisons of the relative D-optimality scores indicate that the updated model
with its approach of sequentially adjusting the entrance fees comes much closer
to the theoretical score than the traditional model.

D-optimality is a concept that applies to overall efficiency of the model
design. This does not imply that the model with the highest level of model
efficiency also has the lowest variance for the individual coefficient estimates

(Table 3-14). The relative variance for the entrance fee has implications for the

 

° The Doptimality scores were normalized by setting the score for the theoretical model equal to
one.

59

precision of the welfare measures. Precision of the welfare measures using the
traditional model will be lower relative to those using the updated model because
the entrance fee coefficient will be in the denominator and the coefficient

variance is more than twice the variance for the updated model.

Table 3-14: Coefficient Variances
Theoretica Updated Traditional

 

I Model Model Model
Entrance fee 0.0000120l 0.0000541 0.0001238
Warblers 0.000041 1 0.0000515 0.0000310

Rare species 0.0000269l 0.0000291 0.0000207
Other species 0.0000011 0.0000011 0.0000008

 

 

 

Abundance 0.01 12765 0.0103784 0.0126705
category
Habitat 0.01 10241 0.0101243 0.0126705

The D-optimality score for the traditional model is a function of the pre-
selected entrance fee difference. If a particularly “bad“ entrance fee differential
was selected for the traditional model, it would look worse in comparison to the
updated model. To address this concern, D-optimality scores for several

aitemative entrance fee differences were computed, 15, 20, 25, 40,50,60.

Table 3-15: D-Optimality Scores For Selected Fee Differences

 

Pro-Selected Entrance Fee D-opiimality Score"
$10 .25
$15 .51
$20 .82
$25 .72
$40 .46
$50 .31
$60 .20

 

60

The results are presented in Table 3-15. The highest D-optimality score
was achieved for a $20 fee difference but the relative efficiency was still lower
than that of the updated model, 82% versus 91%.

A possible explanation for this finding is that under the traditional
approach, preselected fees are assigned across all scenarios according to the
design plan and the updated model allows adjustments in the fee separately for
each scenario.

The updated model has several caveats. The model is based on a linear
utility function. Linearity may not be realistic for the application, however, and it
is not possible to test for non-linearity. The experimental design is based on the
existence of boundary values for the attributes. If boundary values cannot be
identified the model cannot be used and if boundary values change the study
results will lose efficiency. The extreme boundary levels may also create
unrealistic scenarios because there is no middle ground. In reality respondents
may not be faced choices between the extreme site attributes that are generated
by the optimal experimental design.

Despite these caveats, the D-optimality scores for the aitemative design
matrices suggest the usefulness of the updating approach, particularly in cases
where little information is available about the unknown parameters and where
time and/or budget constraints exist and small samples must be used. In the
end, the researcher must decide whether efficiency gains through more

extensive pre-testing in conjunction with a traditional approach is worth the time,

61

effort, and money relative to the ‘costs’ involved in using the sequential updating

approach used in this study.

Summary and Conclusion

Model efficiency may be an important criterion for model selection in
cases where sample size is limited due to budgetary or other constraints. The
results of this study show that optimal experimental designs that allow for
updating of a continuous variable provide efficiency gains relative to traditional
binary-choice experiments that use pre-determined attribute levels.

Choice experiments are typically based on an orthogonal main effects
plan where attributes and attribute levels are specified in advance. The model
that forms the basis of this application only requires that upper and lower bounds
be placed on attribute levels. The requirements for an optimal binary-choice
design then involve setting all but one attribute at the upper or lower boundary
levels according to an orthogonal main effects plan and using the remaining
attribute as a design-balancing variable.

Respondents to the survey used in this research were asked to select one
aitemative from a pair of scenarios. Each aitemative in a pair has, thus, a certain
probability of being selected by any one respondent. The split for the choice
probabilities that are required for optimality of the design are determined by the
number of attributes in the experiment. The probability to be assigned to an

alternative is based on the experimental design plan for the balancing variable.

62

The balancing variable is sequentially adjusted in the data-collection stage to
achieve the required choice probabilities for each alternative in a pair.

While theoretical design efficiency cannot be reached empirically
because model parameters are unknown, the sequential updating approach may
closely approximate theoretical design efficiency if prior information exists or can
be collected on the possible size of the coefficients and if updating is possible.
One strength of Kanninen’s approach is that it is not necessary to know the true
coefficients prior to data collection. It is sufficient to know the response
probabilities, which are determined solely by the number of attributes in the
design. Updating of the balancing variable is then used to try and approach the
optimal response probabilities.

To explore the implications of Kanninen’s approach, the optimal binary-
choice design with sequential updating was applied to bird watching. A survey
instrument was developed whose valuation portion consisted of an arrangement
of 8 pairwise comparisons of birding sites with 6 attributes. The surveys were
administered in personal interviews. The design-balancing variable was an
entrance fee to a birding site. The frequency of updating of the entrance fee
depended on how closely together the interviews were scheduled and on how
the actual choice probabilities behaved relative to the desired probabilities. A
constraining factor was the upper threshold of $30 that was placed on the
entrance fees. Any higher amount was felt to be too unrealistic for a daily

entrance fee to go birding for one day.

63

Model efficiency was measured by D-optimality scores. D-optimality
scores were calculated for the updated model, the theoretical model, and a
traditional model based on pre-selected entrance fees. Comparisons of relative
model efficiencies showed that the updated model closely approximates the
theoretical model efficiency, as compared to the traditional model.

The results of this study underscore the efficiency gains that are possible
by sequentially updating attribute levels of the design-balancing variable relative
to using a model with pre-determined attribute levels. These efficiency gains
must be balanced against the increased administrative time and money costs
required for tracking choice probabilities and adjusting attribute levels. These
costs may not weigh very highly in cases where small samples must be used
and efficiency gains are prized. The study also underscored the trade-off
between efficiency gains and realism in the choice scenarios. By setting
attribute levels at their extreme boundaries, optimal choice probabilities may
require values for the balancing variable that respondents consider unrealistic.

In this pilot study personal interviews were conducted with a small number
of respondents (60). Personal interviews are not feasible on a large-scale
because of the time, cost, and effort that would be required. A mail-survey that
could be conducted in several waves, to be able to update the balancing
variable, could save on money cost and effort involved. It would be time-
intensive, however, because of the tum-around time involved in administering
mail surveys. A telephone survey may be a good alternative when the survey is

not very long and scenarios can be related to the respondent during the

54

interview. Where the latter is not possible a combination of a mail- and
telephone-based survey may be feasible. Scenarios with all but the balancing
variable entries could be mailed prior to the telephone interview and updating of

the balancing variable could proceed between sets of interviews.

65

Chapter 4

BIRDER PREFERENCES AND BIODIVERSITY INDICATORS

Introduction

Natural and social scientists and environmental activists have identified
many reasons for preserving biodiversity. These include the contribution of
biodiversity to the proper functioning of ecosystems, nutrient cycling, watershed
protection, and recreational values (McNeely 1988). Yet the world has
witnessed rapid rates of species extinction (Lovejoy 1986). While exact numbers
are not available for the rate of extinction of species, “the number of extinctions
that might be expected by the end of the century...” range from hundreds of
thousands to over a million“ (Lovejoy 1986: 14). High rates of species extinction
thus motivate a large part of the extensive literature on biodiversity (Ryan 1992).
Habitat protection to preserve species receives the most emphasis in the
literature (Burley 1988). But in order to preserve a large variety of species, it is
necessary to know where the most critical areas for biodiversity are (Bibby et al.
1992, McNeely et al. 1990, Reid and Miller 1989). The selection of areas that
receive “protected“ status thus becomes important (Barbier et al. 1994, Jenkins Jr.
1988, Burley 1988). Protected areas have also become attractive places for
people to visit for outdoor recreation purposes, such as birding. The issue of
which areas to designate as “protected“ has been taken up by researchers and
protected-area managers, as they are dealing with questions of the size and

management of protected areas, conflicts between human uses of the protected

66

area resources (eco-tourists and local populations) and the preservation of species,
canying capacity, and sustainable use of protected areas (Reid and Miller 1989,
McNeely et al. 1990, Reid et al. 1993).

The economic significance of biodiversity is also recognized in the
literature (Pearce and Moran 1994, Pearce 1993, Swaney and Olson 1992,
McNeer 1988, Haneman 1988, Randall 1988). The preservation of species and
the required habitat have value not only because of the ecological benefits
mentioned above, but also because individuals and society enjoy the
recreational benefits provided by biodiversity (Barbier et al. 1994, Norton 1988).
The fact that biodiversity has both social and scientific value raises the question
of how to select areas for protection that will yield the “correct" level of
biodiversity. Biodiversity indicators have been suggested as a basis for
selecting habitat areas for protected status (e.g. species richness; Scott et al.
1987). Researchers have also tried to use people’s preferences for biodiversity
to devise conservation incentive schemes, payment schemes for conservation,
coordination of international conservation efforts, action plans, strategies, and
charters (Reid and Miller 1989, McNeer et al. 1990, Cooper 1991, Gray 1991,
Schi‘rcking and Anderson 1991, Bibby et al. 1992, Ryan 1992, Swingland 1993).
For a general description of how to value‘ biodiversity see Steffens and Hoehn
(1997) and for a contingent valuation study of endangered species see
Jakobsson and Dragun (1996). This study asks whether biodiversity is a
significant factor in birders’ choice of birding sites, and whether there is a

significant relationship between birder preferences and the biodiversity levels of

67

sites. Birders derive pleasure from observing birds in the wild and thus represent
a group that enjoys the recreational benefits of natural resources (Kerlinger
1995a and 1995b, Wiedner and Kerlinger 1990, Jacquemot and F ilion 1987,
Kellert 1985). For the purposes of this study, the aspect of biodiversity that is
examined is limited to bird diversity.

This research application of birds and birding is motivated by several
considerations. First, birds are considered good indicators of biodiversity
because they exist throughout the world, are sensitive to environmental
changes, and more is known about their distribution and taxonomy than is known
about other elements of ecosystems (McNeely et al. 1990). Second, bird-related
data are available for a large number of birding sites. Third, access to a
database of birders was provided the American Birding Association (ABA); such
data is not available for other potential biodiversity indicators. A group of ABA
members were selected as the resource-user group for this study.

ABA birders in Central Michigan were interviewed to obtain preference
information for six attributes of birding-sites. These site attributes are: the
number of warblers, the number of rare or unusual species, the number of other
species, the abundance of warblers, habitat, and entrance fee to the birding site.
The disaggregated attribute information was used to compute seven aggregated
biodiversity indicators.

The collected data was analyzed by specifying a binomial logit model.
The empirical models that are estimated are the user-preference model and

models using aitemative biodiversity indicators. The study investigates whether

68

biodiversity as an aggregate measure of bird-related attributes of birding sites is
a significant factor for birders’ choice of a birding site.

The study also tested the hypothesis that the models with the biodiversity
indicators are not significantly different from the user-preference model with the
site attributes. This hypothesis suggests that biodiversity indicators substitute for
the more complex information contained in the disaggregated user-preference
model.

The biodiversity measures for the birding sites can be decomposed into
measuring biodiversity of different types of bird species. The disaggregated
model distinguishes between warblers, rare species, and other species.
Comparison of a decomposed measure of biodiversity versus an aggregated
measure of biodiversity was only possible for the given dataset in the case of
species richness. The decomposed model breaks species richness (total number
of species) down into number of warblers + number of rare species + number of
other species. The aggregate and decomposed models for species richness
were compared to show whether the decomposed model has a better fit (percent
correctly predicted responses) than the aggregate model. For the other
biodiversity indicators, the dataset does not include variability in the
subcategories comparable to variability for the site to make the comparison
between an aggregated biodiversity model with a decomposed biodiversity
model meaningful. The difference in these biodiversity indicators relative to
species richness is that they are driven by considerations of abundance and

evenness of distribution in addition to the number of species. However, in this

69

dataset warblers and rare species are always in the same abundance category
at anyone site and other species are evenly divided into three groups. Thus, the
biodiversity indicators for each subcategory of birds would be dominated by
species richness.

Ecologists have suggested using biodiversity indicators to select areas for
protected status (e.g. Scott et al. 1987). In cases where resource user
preferences are to be taken into consideration in addition to an ecological
measure for biodiversity to select areas for protection, a biodiversity indicator
that fits user-preference data better than another would be preferred. The model
specifications with the different biodiversity indicators were compared in terms of
percent correctly predicted responses to determine whether one or several
indicators perforrn(s) better than others.

The biodiversity-indicator models were also estimated using site attributes
that are included in the user-preference model but are not (directly) bird-related
attributes, i.e. entrance fee, habitat. The expanded models were compared to the
biodiversity models in terms of goodness-of-fit.

Natural-resource decision makers may be interested in knowing how
much birders value changes in the levels of birding-site attributes, including
levels of biodiversity. The empirical model can be used to estimate birders’
marginal willingness to pay (WTP) for attributes of birding sites, including
biodiversity, and total WTP for a site with certain site characteristics. Meaningful

estimates for WI'P would, however, require a larger-scale study than was

70

available for this research; WTP was therefore not calculated in the present
study.

The rest of this chapter is organized as follows: biodiversity indicators are
explained in general, than the empirical model is introduced, this is followed by a
description of the model application to birding, the results of the study, and the
summary and conclusion. The section on the application explains how the
attributes were selected, how information about biodiversity indicators was
obtained and incorporated into the study, and how the biodiversity indicators that

were selected for the study' were computed.

Biodiversity Indicators

No unique and universally accepted definition of biodiversity has emerged
from the ecological literature on the meaning and measurement of biodiversity,
which dates back to the 1950's (Reid et al.1993). Biodiversity may be defined in
terms of genetic diversity (i.e., variation in genetic material), species diversity
(i.e., variability across species), and ecosystem diversity (i.e., the variability in
the ecological complexes which form habitat for species) (McNeely 1988, Pearce
and Moran 1994, Reid and Miller 1989). For the purposes of this study,
biodiversity is defined as species diversity. Species diversity can be measured in
different ways. While species diversity is synonymous with species abundance
(species richness) for some (McNeely 1988), others include distributional
considerations as well (Pearce and Moran 1994, Williams et al. 1993). For

recent approaches to selecting optimal biodiversity preservation strategies see

71

Weitzman (1992), Solow et al. (1993), and Solow and Polasky (1994). These
latter approaches focus on species extinction while this study takes a broader
approach to include species that are not threatened or endangered (see also
Scott et al. 1987).

Commonly used ecological measures of biodiversity are species richness,
the Shannon index (also called Shannon-Wiener index or Shannon's H), and the
Simpson index (Pielou 1975). Four additional biodiversity indicators, described
in Magurran (1988), were used in this study: the Margaief, Menhinick, McIntosh,
and Berger-Parker indices. The seven biodiversity indicators are described in

more detail below.

Empirical Model

A respondent’s choice of which of two birding sites to visit can be
modeled as a binary-choice problem. The respondent is asked to choose one of
two aitemative sites. The aitemative sites differ in the level of at least one of the
attributes. A utility-maximizing respondent will choose the site that gives him or
her the greater level of utility, assuming that the travel costs, including entrance
fee, across sites is constant. The utility function is assumed to have a random
error with an extreme value distribution.
Individual j will choose site A over site B if:
V,(Q,)+e,, >V,(Q,)+sp. Equation 10
where for site A or site B depending on subscript:

Vl = indirect utility of respondentj

72

Q‘ = vector of site attributes facing respondentj
s,= random error term

The probability of individual j choosing site A can be expressed as:

P(A)=P{V,(Q,)+s, >V,(O,)+ep}. Equation 11

A user-preference and several biodiversity-indicators models were
specified. The user-preference specification models utility as a function of six
attributes of a birding site. For respondentj at anyone site indirect utility is:

V‘ = f(warb, abun, rare, 0th, hab, fee) Equation 12
where:

warb = number of warblers

abun = abundance category for warbler species

rare = number of rare or unusual species

oth = number of other species (other than warblers or rare or unusual species)
hab = dummy variable for diversity of habitat (= 1 if high level of diversity, = 0
otherwise)

fee = daily entrance fee

The underlying utility function for the preference model using six attributes of a

birding site is:

Vll = warbifl, + abunfiji2 +rareij33 + othiB, +habiBs +feeiB, +6i Equation 13

where the variables for respondentj at site i are defined as above and a =

random error term.

73

Alternatively, the model was specified with differenced biodiversity
indicators (biodiversity indicator k for site A minus biodiversity indicator k for site
B)“ The biodiversity-indicator model is a nonlinear function of the bird-related
attributes of the preference model (number of warblers, number of rare species,
number of other species, and (except for species richness) abundance).1o
For any given site:

BIO, = f(warb,abun,rare,oth) Equation 14

where BIO,K is any one of the k biodiversity indicators. In the case of species
richness ‘abun’ does not enter the biodiversity indicator. Species richness, BIO,,
is computed as follows:

BIO, = warb + rare + oth . Equation 1 5
The other biodiversity indicators are a more complicated function of the four
bird-related attributes and they are described in the section “Computation of
Biodiversity Indicators“.

The following function was estimated:

VIll = BlOny. + Vui Equation 16

where BIO is defined as above andj indicates the individual respondent.

 

9 In this case: 0, = B,(Bro,,‘ - BIO.,,°).

‘° See the section on “lnfonnation for Biodiversity Indicators“ for details on how the biodiversity
indicators relate to preference information and “Computation of Biodiversity Indicators“ on how
they were computed.

74

In the study application a scenario described two aitemative birding sites
in terms of birding-site attributes. The respondents were asked to select their
preferred birding site from the pair of sites. In the event that they did not like
either of the sites, respondents were asked to choose the site they disliked less.
A binomial logit model specification with an underlying linear utility function was
selected for estimation.‘1 The random error is assumed to follow an extreme
value distribution. Following these assumptions, the probability of respondentj

choosing site A is given by:

Exp(e,)

P(A‘)=1+Exp(6,)

Equation 17

Where: 9, = B1(Q1j‘ " Q1jo)+ 32IQ211 " QZjo )+---+Bm(qfl1 . qtlilo)

P(A,) = probability of respondentj choosing site A from the pair of sites.

The right-hand side variables are the differenced levels of site attributes (level of
attribute i at site A minus level of attribute i at site B, for i = 1, m).

The log-likelihood function for the aitemative model specifications is:

InL = in, InP(A,) + (1 — yi)ln(1- P(Ai))] Equation 18
H

y ,= 1 when respondentj prefers site A and =0 otherwise.

 

" A random-effects probit model did not show significant individual effects to warrant a panel-
data approach to the multiple individual responses contained in the data set.

75

Application

The data for this study were obtained from a survey of Michigan birders
that was conducted in the fall of 1998. Sixty members of the American Birding
Association in the Lansing area were interviewed for this study; usable data
were obtained from 58 respondents. Each birder was presented with eight
different pairs of scenarios and thus eight observations were obtained from each

respondent for a total of 464 usable observations.

Selection of Attributes

Individual interviews were conducted with birders to learn more about
their birding activities, particularly their birding trips. Structured and semi-
structured interviews, using ethnographic interviewing techniques, were used to
determine attributes that may influence birding-site choice.

Six site-specific variables were selected for this study. They are: number
of warbler species, abundance of warbler species, number of rare species,
number of other species, habitat, and entrance fee. Birder interviews revealed
that the number of species that could be found at birding sites, the chance to
see rare or unusual species, and the diversity of habitat were important site-
specific determinants for site selection. The chance of seeing life-birds, that is,
bird species never seen before by a respondent, was also an important attribute
but was not included in the study because it had the chance of heavily

dominating respondents’ choice. There may also have been too many

76

respondents who had seen all of the species occurring in Michigan. These
respondents would thus have rejected a scenario that mentioned a life-bird. The
majority of the birders interviewed also particularly enjoyed seeing migratory
warblers because of their colorful plumage in the spring. An entrance fee was
selected as a site-attribute in order to model monetary trade-offs (Buchanan et
al. 1998).

Because warblers and rare species would be part of the total number of
species at the site, the three species-related attributes selected for the study
design were designated as, “warblers,“ “rare,“ and “other species.” The presence
of these three attributes, as well as habitat and the probability of having warblers
at the site was expected to increase the probability that a site would be selected.
An increase in the entrance fee for a site was expected to reduce the probability

that a site would be chosen.

Information for Biodiversity Indicators

The attributes that were presented to respondents in the survey as part of
the site descriptions are only sufficient to calculate the species richness
indicator of biodiversity (warblers+ rare species + other species). The other
biodiversity indicators require knowledge of the number of individual birds per
species. The scenario descriptions of the birding sites that were presented to
respondents contained information only about the number of species and the

abundance categories but did not include information about the number of

77

individual birds per species. Therefore, to calculate the biodiversity indicators,
other than species richness, additional information was needed.

The abundance category for a species provides a birder with information
about the likelihood of having a species occur at a site (Smith et al. undated).
This is similar information to the number of individual birds per species, from
which birders can also infer the likelihood of having a species occur at a site
(Evers and Granlund 1991). To create the link between abundance and
individual birds per species, survey respondents were presented with an
equivalency chart (Table 4-16) that presented them with information about how a
given abundance category relates to the number of birds per species.

This information was then used to calculate the biodiversity indicators.
Thus, while the number of birds per species was only mentioned by interview
respondents in connection with seeing large flocks of birds, this variable was
included in the study design to allow calculation of the biodiversity indices.

The connection between an abundance category and the probability of
observation has been documented in the literature (ABA 1995, Smith et al.
undated). An approximate correspondence between the number of birds per
species and an abundance category can also be deduced from the literature
(Evers and Granlund 1991).

Publications are available for birders who are interested in obtaining
information on the bird diversity they are likely to find at different birding sites
(Wauer 1993, Jones 1990, Kitching 1976). Some of the publications are bird charts

themselves or contain bird charts (e.g. Friends of Point Pelee 8. Pelee Island Winery

78

1994, Evers and Granlund 1991, US. Department of Interior 1991, Jones 1990).
Bird charts typically list the names of the bird species to be found at the site by
season and give an abundance category (abundant, common, uncommon,
occasional, or rare).

The range of attribute levels for the survey scenarios were selected based on
actual ranges of the attribute at Michigan birding sites for which bird charts were
available. Five charts were analyzed in more detail, Whitefish Point (Evers and
Granlund 1991), Seney Wildlife Refuge (US. Department of Interior 1991),
Shiawassee \Nildlife Refuge (U.S.Department of Interior 1979), Rose Lake (Lerg
undated), and Grand Mere. With the help of an experienced birder, the charts
were split into birds that occur at the sites in early and in late spring (Johnson).
The numbers of warblers, rare or unusual species, and other species were then
summed for the different sites and a range set for each attribute. The birding
consultant also identified rare or unusual species for each chart. A range of 530
species was set for rare or unusual species, 5-25 for warblers, and 75-201 for
other species.

The data required for the calculation of the biodiversity indicators was
generated by matching the abundance categories to numbers of birds per species
as indimted by the equivalency chart. Several of the bird charts and information
from the breeding bird survey were used to develop the equivalency chart (Sauer
1997, ABA 1995, Evers and Granlund 1991, Smith et al. undated).

The bird list for Berrien County explained abundance categories in terms

of the likelihood of seeing species a certain percent of the time (Smith et al.

79

undated). Other bird lists (Jones 1990, for example) explain abundance
categories in terms of the individual birds per species, eg. abundant: likely to be
seen in large numbers, and/or the likelihood/difficulty of observing the species, '
e.g. common: usually seen in proper habitat. A birding trip description in
‘Winging lt’ provided percentages to indicate the likelihood of observing species
on a given birding tour (ABA 1995). The Berrien County list used percentages of
trips on which one would see the species (Smith et al. undated).

The percentage ranges differed across sources. For this study a
percentage range of less than 20% was assumed to be the likelihood of
observing rare or unusual species and was assumed to be the lowest level of the
abundance categories for warblers and other species. The highest level was set
at the 80-100% range used in the Berrien County list. For the likelihood of
having other species occur at the site, the number of other species was divided
into three categories, roughly corresponding to the categories, “rare,“
“uncommon,“ and “common.“ The categories were assigned the ranges of less
than 20%, 21-79%, and 80-100%, respectively.

To construct the biodiversity indices, the abundance categories] likelihood
of having the species occur at the site was translated into the number of birds
per species. The Whitefish Point bird chart relates the abundance categories to
the number of individual birds per species recorded in a 7-10 day period. The
breeding bird survey also provides categories for individual birds per species. ‘2

The numbers used here are based on a crude calculation of dividing the

80

Whitefish Point numbers by 7 or 10. The cut-off point was randomly set at 101. It
was felt that most species are not seen in numbers greater than 101 per species
on most days. Exceptions would be certain species during peak migration days
at points where migratory birds tend to congregate. The numbers that resulted
are similar to (combined) categories used in the breeding bird survey. The
breakdown by category and the equivalencies for abundance category,
likelihood of species occurrence, and number of birds per species is given in

Table 4-16.

Table 4-16: Equivalencies

 

 

 

 

 

Abundance Explanation in words Chance of Number of
Category having species birds of the
at site same species

Common usually seen in proper 80-100% 31-101

habfiat
Uncommon present, but not 21-79% 4-30

certain to be seen
Rare rarely seen less than 20% 1 -3

 

 

 

 

Computation of Biodiversity Indicators13

The seven selected biodiversity indicators are divided into two types: 1)
species richness indicators and 2) indices based on the proportional abundance
of species. Within the latter group, biodiversity indicators can further be

distinguished as information index or dominance index.

 

‘2 The breeding bird survey data is generated on the basis of counts along a survey route with 50
points where birds are counted during 3-minute intervals.

‘3 In order to calculate the biodiversity indicators n, is defined as the midpoint of the abundance

category (Le. ’birds per species'). N is calculated by multiplying the number of species in each
abundance category by their respective abundance category midpoint and summing the results.

81

 

Species Richness Indicators:
1. Species Richness:

Species richness is the simplest of the biodiversity indicators. Its
calculation is straight forward; it is the total number of species. Species richness
in this study ranged from 85 to 275.

2. Margaiefs Index:

Margaiefs index, as well as Menhinick's index (described below), is a
diversity index that incorporates the distribution of species abundance. It has
been found that this is a “more sensitive measure of environmental disturbance
than species richness alone“ (Magurran 1988, 11). It is calculated using the
formula:

(3 - 1)lln N
where S = number of species and N = total number of individuals at the site.

3. Menhinick’s Index:

Menhinick’s index is easy to compute, as is Margaiefs index (described

above). It is calculated as: SI N

where S and N are defined as above.

indices Based On the Proportional Abundances of Species:
These indicators do not provide as much information as species
abundance models, but by providing a single indicator they allow comparisons

across communities when one model does not fit all of the selected communities.

82

The indicators take both evenness (how evenly individuals are distributed across
the different species) and species richness into account. Because they do not
assume a particular underlying abundance distribution, these indicators can be
considered non-parametric (Magurran 1988). These indicators include:
1. Shannon’s H:

Shannon’s H is a heterogeneity measure that takes evenness of

distribution into account. It is derived from information theory. The formula for

the calculation of Shannon’s H is: — Z, p,+ln p, ,

where p, = proportion of birds per species i in the total population of birds.

p, = if where n, = number of birds per species i, and N = total number of birds.

2. Simpson’s index:

Simpson's index is a dominance measure. Rather than providing a
measure of species richness, it gives more weight to the abundance of the
common species than it does to other species. The Simpson’s index is computed

as follows: 21%.

where n, and N are defined as above. The Simpson index decreases as the level

of diversity increases. The Simpson index is therefore usually expressed as 1 -
D or as 1ID. The Simpson index was computed as 1ID in this study.

3. Mclntosh’s Index:
The Mclntosh’s index is calculated as follows: U = Zn,2 , where ni is

defined as above. This index is not a dominance indicator but can be converted

83

 

. . N — U . . .
into one by calculating: N m , where U IS the McIntosh index and N IS the

same as above. The latter calculation will be used as the Mclntosh’s index in this
study.
4. Berger-Parker Index:

The Berger-Parker index reports the proportional importance of the most

abundant species and is calculated as: Nm I N, where NM = the number of

individuals in the most abundant species and N is defined as above. As with the
Simpson index, the Berger-Parker index increases as biodiversity decreases.
The inverse of the above formula will therefore be used as the Berger-Parker
index in this paper. The Berger-Parker index is also a dominance index.

The values of the different biodiversity indicators are listed in Table 4-17
and in Table 4-18 for each of the alternatives in the eight different scenarios.
The entries in the tables illustrate that all of the biodiversity indicators move in
the same direction (with the exception of the Menhinick’s index in scenario 7).
This finding is consistent with examples using indices for species richness,
Margaief, Simpson, Shannon’s H, Berger-Parker reported in Magurran (1988,
63, 67). Magurran also shows that sites ranked by their level of diversity using
different indices produce rankings that are not significantly different (using the
Spearrnan rank correlation coefficient) for indicators of the same type, either
species-richness type or dominance/evenness type. The Simpson, McIntosh,
and Berger-Parker indices are dominance indicators, Shannon’s H is a

heterogeneity index, and all others are species-richness indicators.

84

Table 4-17: Biodiversity Indicators for Scenario Alternatives (Scenarios 1- 4)

 

 

 

 

 

Alt1 a Alt1 b Alt2a Alt2b Alt3a Alt3b Alt4a Alt4b
Species 1 10 231 105 236 85 256 130 21 1
Richness
Margalef 14.17 25.83 13.53 26.97 10.76 29.42 15.63 24.28
Menhinick 2.35 2.69 2.25 3.03 1.71 3.36 2.10 2.79
Shannon's H 3.96 4.99 3.94 4.89 3.94 4.89 4.31 4.83
Simpson 42.18 130.95 41.80 113.11 44.81 110.14 66.41 106.78
McIntosh (U) 0.863 0.923 0.862 0.917 0.867 0.916 0.891 0.914
Berger-Parker 33.26 111.44 33.11 92.20 37.35 87.95 58.11 86.59

Table 4-18: Biodiversity Indicators for Scenario Alternatives (Scenarios 5 - 8)

 

 

 

 

Alt5 a Alt5 b I Alt6 a Alt6 b Alt7 a Alt7 b Alt8 a Alt8 b
Species 236 105 231 1 10 211 130 256 85
Richness
Margalef 27.14 12.62 26.57 13.92 24.12 16.73 28.62 10.95
Menhinick 3.11 1.71 3.05 2.19 2.72 2.75 2.97 1.84
Shannon's H 4.86 4.26 4.85 4.02 4.86 4.03 5.02 3.86
Simpson 108.64 64.70 108.27 46.62 111.28 43.72 132.72 40.30
McIntosh (U) 0.915 0.889 0.915 0.870 0.916 0.866 0.923 0.860
Berger-Parker 87.35 57.35 87.20 38.11 91.44 33.86 112.20 32.50

Table 4-19: Spearrnan Rank Correlation Coefficients

Species Margalef Menhinick Shannon's Simpson Mclntos Berger

Richnes
s
Species
Richness
Margalef 1 .00"
Menhinick 0.95““
Shannon's H 0.84““
Simpson 0.81“
McIntosh (U) 0.81“
Berger- 0.81“
Parker

1.00““

0.95““
0.84““
0.81“
0.81“
0.81“

H
0.95““ 0.84““ 0.81 “*
0.95““ 0.84“ 0.81“
0.70“ 0.67“
0.70“ 0.99““
0.67“ 0.99““
0.67“ 0.99““ 1.00““
0.67“ 0.99““ 1.00““

“Correlation is significant at the 99%

confidence

“Correlation is significant at the 95%

confidence

85

h (U)
0.81““
0.81“
0.67“
0.99““
1.00““

1.00““

Parker
0.81““

0.81“
0.67“
0.99““
1.00““
1.00““

The Spean'nan rank correlation coefficients for the biodiversity indicators
using the differenced biodiversity indicators for the aitemative sites and a
ranking of site differences for the eight scenarios used in this study are
presented in Table 4-19. All of the correlation coefficients are significant at
either 95 or 99% confidence. The 99% confidence level was achieved for
correlations among indicators of the same type, species richness type or

dominance type, with the exception of the Menhinick index.

Results

As previously discussed, respondents were given eight pairwise choice
scenarios and were asked to choose one site from each pair based on the
information provided for attributes of birding sites.

A binomial logit model was estimated that included the six attributes as
explanatory variables. In this disaggregated case, all of the estimated
coefficients had the expected sign. The bird-related attributes, number of
warblers, number of rare or unusual species, number of other species, and
abundance category for warblers all had positive coefficients, as did habitat. The
coefficient for the entrance fee was negative, as expected. All of the coefficient
estimates were significant. The coefficient estimates and the standard errors for
the birder preference model are reported in Table 4-20 as the disaggregated

case.

86

Table 4-20: Binomial Logit Results for Disaggregated Case

Reported are coefficient estimates with standard errors in parentheses

Variable

Disaggregated
Case

 

No. of warblers

No. of rare species
No. of other species
Abundance of warb.
Habitat

Entrance fee

Log Likelihood Fct.
Correctly Predicted
Responses (total)
(site A)
(site B)

 

0.033“
(0.007)
0.027“
(0.005)
0.0057“
(0.001)
0.27"
(0.10)
023*
(0.10)
6.029"
(0.0074)
-293.15

64%
5696
7296

“ significant at 95% confidence level; *“ significant at 99% confidence level

The table reveals that birders are more likely to choose a birding site the

more warbler species, rare species, and other species can be found at the site.

The abundance category for warblers differed across sites from which

respondents selected their preferred site. This indicates that the degree to which

birders can be sure of having warblers at the birding site is important for birding

site selection. Birders are also more likely to select a birding site that has a more

diverse habitat. Survey respondents care about habitat independent of the

diversity of species that can be found at the site. Habitat is a significant variable

but it is not correlated with the number of species, some scenarios had diverse

habitat combined with less species diversity for one aitemative while the other

aitemative had a less diverse habitat but more species.

87

As a measure of goodness of fit, the percentage of correctly predicted
responses is also reported in Table 4-20 (Greene 1993). The disaggregated
model correctly predicted 64% of the actual choices. This means that in 64% of
the total cases the model predicted that respondents choose A when they
indeed chose A and predicted that respondents choose B when they indeed
chose B. The total cases of correctly predicted responses for A and B was
broken down by correctly predicting selection of site A and correctly predicting
selection of site B.

Respondents were provided with supplemental information about the
correspondence between abundance categories, which were part of the
valuation scenario, and the number of individual birds per species for the given
categories. This information was used to compute the seven different
biodiversity indicators as described above.

Table 4-21 reports the results for the estimated models using the
aitemative biodiversity indicators. For all of the models reported in the table, all
coefficient estimates had the expected sign. All of the coefficient estimates were
significant. With the exception of the model specification using the Menhinick
index, the percentages of correctly predicted responses were 59% for the
biodiversity-indicator model specifications. The Menhinick model showed 58%
correctly predicted responses. The significance of the estimates with the
expected sign indicates that the aggregate measure of biodiversity indicators

explains site choice well. The goodness-of-fit measures are almost identical and

88

the loglikelihood values are also very similar. This suggests that preference
cannot be given to anyone of the biodiversity indicators.

When the biodiversity-indicator models were estimated with the entrance
fee and alternatively with the entrance fee and habitat (also entrance fee,
habitat, and abundance category for species richness) included as independent
variable(s), the coefficients for the added variables were not significant for any of
the models (except habitat when three variables were added to species
richness) and the entrance fee did not have the expected negative sign for most
model specifications (see Appendix B). The goodness-of-fit measures, percent
correctly predicted total responses, were identical for the different functional
forms across individual indicators. The biodiversity-indicator model specification
with only the indicators was therefore used in this study.

It cannot be concluded from this finding that entrance fee, abundance of
warblers, and habitat are not important. The disaggregated model shows they
are significant in site choice and jointly they add 6% to the percent correctly
predicted responses.

The disaggregated model performed better than the aggregated models
using alternative biodiversity indicators, 64% correctly predicted responses for
the disaggregated model versus 59% for the aggregated models (58% for
Menhinick indicator). Comparison of the aggregate biodiversity indicator models
versus decomposition of the models is only reasonable for the species richness
model. Species richness can be decomposed into number of warblers + number

of rare species + number of other species. The other biodiversity indicators are

89

computed using an equivalent to abundance categories expressed in numbers of
individual birds. In this study an abstraction from the real world involved
assigning all warblers at a site to the same abundance category. Rare species
were assumed to always be in the rare abundance category and other species
were evenly split into three abundance categories. Because of this simplification
decomposing the other biodiversity indicators by warblers, rare, and other
species would eliminate much of the depth that is contained in the aggregate
indicator by giving undue emphasis to species richness rather than reflect the
degree of evenness of distribution and abundance.

A loglikelihood-ratio test for the biodiversity-indicator model using species
richness, as the restricted model, and the decomposed species richness model,
using number of warblers, number of rare species, and number of other species,
as the unrestricted model rejects the restrictions. The loglikelihood statistic of
10.59 exceeds the critical value for a chi-squared distribution at 95% confidence
with 2 degrees of freedom. This indicates preference for the decomposed model.

The biodiversity-indicator model resUlts show that birders care about
biodiversity as measured by the indicators when making choices about birding
sites. In order to compare the user preference model, the disaggregated model,
with the aitemative biodiversity-indicator models, a likelihood ratio test was
specified for the species richness indicator and a Vuong test for the remaining
six indicators. Because the species richness indicator is nested in the

disaggregated model the likelihood ratio test was performed.

90

.65. 8:63:00 88 .6 2:62:ch : ”.06. 00:03:00 8mm .6 E3558 ..

 

 

 

$8 $8 $8 $8 $2. $8 $8 $2. m 6%
$8 $8 $8 $8 $9. $8 $8 $8 < 2?.
6206de
$8 $8 $8 $8 $8 $8 $8 $8 .22 2.0260
8.2”. 8.8”- 8.2». ~35- 8.3”. 888. 8.8a. 988. so“. 685.93 8..
. 58.9
scmN0.0i 00* 8cmhcw
ate .
.88 .536...
82:
32.10 .995 no mocmbcan<
$8.9
..nmood 36QO .050 6 .02
68.9
..smod 3.0on 88 “.o .02
£88.
..8o.o 82263 .0 .oz
$8.9 85.: 38.2 82.2 5.8.3 38.9 88.9
:83 :9; .888 .888 :83 :33 :83 66.... £22,605
sofiau I 30:30.1 U20
.398 52:32 585m 8888 5.5.55. 8.8.22 3.08m 888.0 636:;

3355.3 5 896 28:68 53 3.9586 290508 6.6 cotoaom
82.8588 .28: 2.8522 .3 8.33. .53 382.6 "a... as.»

91

The remaining models are not nested and the Vuong test as a non-nested
test is applicable. The likelihood ratio test is a test of restrictions on the
disaggregated model. The restrictions are that the number of warblers, the
number of rare species, and the number of other species can be summed and
thus have the same coefficient, and the coefficients on abundance category of
warblers, habitat and entrance fee are zero. The Vuong test allows testing of the
hypothesis that the birder preference model and the alternative biodiversity-
indicator models are not significantly different.

The likelihood ratio test statistic is computed as:

LR = —2(lnLr — InL“)

where:

InL, = the restricted loglikelihood function,

InL": the unrestricted loglikelihood function.

The statistic is distributed as a chi-square with 5 degrees of freedom and takes
on a critical value of 11.07 at the 95% confidence level. The likelihood ratio test
statistic for the species richness restrictions is 32.24 and therefore the
restrictions are rejected at the 95% confidence level. This gives preference to
the unrestricted model, the disaggregated model where all six site attributes
enter the utility function separately.

The Vuong statistic is computed as follows:
v = ME I s,"

m. = YIIInP(A1)l - INPM: )r] + (1 - Y.)['n(1- P(A1)i) - In(1- PM: )a )l

92

where:

 

 

__ expixr'fl)
P(A1)l " 1+9XP(X.'B)

_ exp(2.'v)
HA2” " 1+exp(2.'v)

x. = vector of the site differences in attributes from the disaggregated model with
its coefficient estimate B,

2, = vector of aitemative specifications of the site differences in biodiversity
indicator with its coefficient estimate 7 ,

N=number of observations,
8 =sample mean,

s,“ =standard deviation for the sample of m ls .

Table 4-22: Vuong Statistics for Alternative Model Specifications

 

Model Specification using: Vuong Statistic
Margalef Index 296
Menhinick’s Index 2.94
Shannon’s H 2.97
Simpson Index 3.05
McIntosh Index 2.94
Berger-Parker Index 3.02

 

93

The Vuong statistic has an asymptotically standard normal distribution
with a critical value of 1.96 at a 95% confidence level (Vuong 1989). The results
of the Vuong test for the alternative biodiversity-indicator model specifications
are reported below.

The Vuong-test results show that the hypothesis of no significant
differences between the birder-preference specification and the six alternative
biodiversity-indicator specifications can be rejected. As a directional test
(Greene 1995), the positive sign of the test statistic indicates that the model
specifications using the six site attributes separately in the utility function are
favored over the biodiversity specifications. This supports the findings from the
goodness-of-fit measure of correctly predicted responses which indicated that
the birder preference specification is preferable, 64% correctly predicted
responses as compared to 59% (58% for the Menhinick specification) for the
biodiversity-indicator model specifications. The log-likelihood values for the two
specifications also indicate a preference for the user-preference model (see
Table 4-21).

While ecological biodiversity indicators as an aggregate measure explain
birder site choice well, birders care about the composition of bird diversity as
indicated by the preference for the disaggregated user— preference model over
the aggregated biodiversity-indicator model and as reflected in the rejection of
the aggregate model restriction for species richness relative to the decomposed

species richness model.

Because the questionnaire asked respondents to choose one of two
alternative sites for eight different scenarios, it is possible that an individual’s
responses to the eight scenarios are correlated, in which case a panel-data
approach for estimation might be more appropriate than a cross sectional
approach. The data set was thus subsequently analyzed using a random-effects
probit model. This model takes the possible correlation across an individual’s
responses into account by adding an error term for the individual effect. As part
of the model output, rho, the correlation between the individual's responses, can
be generated. The fact that rho was not significant (95% confidence) indicated
that there were no significant random effects due to multiple individual

responses. This model is therefore not reported here.

Summary and Conclusion

Birder preferences for six attributes of birding sites were modeled as a
binary-choice experiment, using an underlying linear utility function. The data for
the application was obtained from personal interviews with birders in Michigan.
Birders were presented with eight pairs of birding site descriptions. The birders
were asked to choose one site from each pair. The site descriptions contained
information on the number of warblers, the number of rare or unusual species,
abundance category of warblers, the number of other species, habitat, and the

entrance fee.

95

A binomial logit model was estimated for this disaggregated model. All of
the coefficient estimates were significant and had the expected signs. The
goodness-of-fit measure of correctly predicted responses was 64%.

The birder survey also provided information that allowed for the
calculation of biodiversity indicators for the birding sites. The biodiversity
indicators used were species richness (number of species), Margalef, Menhinick,
Shannon’s H, Simpson, and Berger-Parker. The first three indicators are
species-richness-type indicators, the remaining four are indices of the
proportional abundance of species (heterogeneity indices, taking into account
both evenness of distribution and species richness).

Binomial-logit model specifications with the alternative biodiversity
indicators as an explanatory variable were estimated. The biodiversity-indicator
model specifications had a goodness-of-fit of 59% correctly predicted responses
(58% for the Menhinick biodiversity index).

Significant coefficients on the independent variables and coefficient signs
of the expected direction suggest that biodiversity as an aggregate measure is a
significant factor in explaining birders’ site choice and that the probability of
choosing a birding site increases with the level of biodiversity, as measured by
the selected biodiversity indicators.

Specifications for the biodiversity-indicator models that include the
entrance fee and habitat (and in addition a specification for the species richness
model that also includes abundance of warblers) yield coefficients for the

additional variables that are not significant (except habitat in the species

96

richness model that includes abundance). This does not suggest that those
variables are not important to birders’ site choice. The variables are significant in
the disaggregated model of user preferences.

Decomposition of the biodiversity indicators by warblers, rare species,
and other species is only reasonable for the given data in the case of species
richness. The decomposed model is preferred to the aggregate model for
species richness.

A nested test for the species richness indicator and a non-nested test for
the remaining six biodiversity indicators, the Vuong test, was conducted to test
the hypothesis that the alternative biodiversity-indicator models are preferable to
the user preference model. The test results lead to a rejection of the hypothesis.

No preference can be given to any one indicator or to any specific type of
biodiversity measure (species richness or heterogeneity) for selection of an area
for protected status that would consider birder preferences as a selection
criterion. All of the coefficients for the biodiversity indicators are significant and
the alternative models have the same (or almost the same for the Menhinick
model) goodness—of-fit measures.

Biodiversity is significant in explaining birders’ site choice, as indicated by
the significant coefficients on the biodiversity-indicator variables, and the
biodiversity-indicator model has a fairly good fit compared to the user-preference
model. Thus, while the biodiversity-indicator and the user-preference models are
not equivalent, as birders care about the composition of biodiversity, the

aggregate biodiversity measures indicate user preferences relatively well

97

APPENDICES

98

APPENDIX A

QUESTIONNAIRE

99

1998
Michigan Birder Study

Michigan Birders’ Preferences
for

Birding Site Characteristics

 

Department of Agricultural Economics
Michigan State University
East Lansing MI 48824

100

CONSENT FORM

The interview is held to develop a birder profile and learn more about birders'
choices of birding sites. The research will try to identify characteristics of different
types of birders and their trade-offs between characteristics of different birding sites.

The session will last about ‘A hour to 45 minutes. Participation is completely
voluntary and can be terminated at any time. Answers to any questions can be
refused. Names are not recorded and the information provided will be treated with
strict confidentiality.

Questions can be addressed to the researcher at any time, before, during or
after the session. Karin Steffens can be contacted at Michigan State University,
Department of Agricultural Economics: (517)355-8529. '

 

 

(interviewee) (date)

101

YOUR BIRDING ACTIVITIES

Here are a few questions about the kind of birding you like to do. It will be helpful to
answer them before the interview.
Thank you very much.

Karin Steffens
(Project Manager)

Please enter a ‘0' if the question does not apply to you.

1. What is the farthest, in miles, that you have ever traveled for a birding daytrip“.
MILES ONE WAY

2. How many birding daytrips‘ did you take this past spring, that is during March,
April, and May?

 

TIMES
3. How many birding daytrips* did you take during the past 12 months?
TIMES
4. How many of these birding daytrips were in Michigan?
TIMES
5. How many bird species are on your Michigan life list?
BIRD SPECIES
6. How many bird species are on your ABA area or North America life list?
ABA BIRD SPECIES
NORTH AMERICA SPECIES
7. How many bird species are on your world life list?
BIRD SPECIES
8. How many bird books do you own
BOOKS

 

9. How many subscriptions to bird/birding magazines do you have?
. SUBSCRIPTIONS

10. How many bird species can you ident'fl by sight in North America without a field
guide?
SPECIES
* A daytn'p is a trip where you leave and return on the same day, that is, you are mt
staying ovemight.

102

1998 Birding Questionnaire
This questionnaire is designed to help me learn about how birders view different
features of their birding activities.
I will be asking about your birding activities, what you enjoy most about birding, and
which characteristics of birding are important to you. The information collected will
be used to show what kinds of trade-offs different birders make when faced with
decisions about which birding site to visit You were selected to participate in this
study because of your interest in and knowledge of birds.
Section I: First I would like to ask you some questions about your birding
activities.
Q. 1 In just a couple of sentences, how were you first introduced to birding?

 

Q. 2 What do you find most enjoyable about birding?

 

Q. 3 Other than observing birds at a feeder, for how many years have you
been actively birding?
years

Q. 4 Did you get a chance to fill out the questions I had mailed to you?

1 Yes 2 No \

Cabot questioma're Go throum mailQ questionnaie
The following questions ask you about birding vacations, trips where you stay
at least one night.
Q. 15 How often do you take vacations where birding is your primary activity?
(Prompt: Howmany mekendtripsdidyoutake andhowmanylongertn‘ps? Tryandrememberwhere
you went.)

1 never

2 less than every 5 years

3 about every 3- 5 years

4 about every 1-2 years

6 2 or3timesa year

7 4 or 5 times a year

8 more than 6 times a year

Q. 16 What is the farthest distance, one way, (or destination) you have
traveled in the past 12 months where the primary purpose of the trip was
bIrding? mm in orprompt)

1 more than 500 miles ?

2 more than 300 miles ?

3 more than 200 miles ?

4 more than 100 miles 75 more then 50 miles ?

6 20-50 miles?

7 less than 20 miles?
Next, I will ask you some questions about your birding skills and bird lists.

103

Q. 17 Thinking about your overall birding skills in your home area, would you
rate yourself

1 as a Beginning birder,

2 an lnterrnediate birder,

3 an Advanced birder, or

4 an Expert birder

Q. 18 How about for North America, how would you rate your overall birding
skills at that level?

1 Beginning,

2 Intermediate,

3 Advanced, or

4 Expert birder

Q. 19 Which of the following best descibes how much you rely on sound when
birding: would you say you

1 Don’t use sound at all

2 Follow sound for visual identification

3 Identify some birds by song or call

4 Identify most birds by song or call

Q. 20 Do you keep a birdlist?
1 Yes gotonext 2 No gotoZZ

Q. 21 Now I will read you different types of lists birders may keep. For each
one, please tell me whether you keep this type of list. (Stop reading list if R.
indicates that those are the only lists R. has)

1 all birds ever seen

2 ABA area (if R. not familiar with ABA area: essentially North Amen’ca)

3 United States

4 backyard

5 home county

6 other county

7 Michigan

8 other state(s)

9 other country

10 daylist

11 triplist

Do you keep any other type of list?
1 Yes goto next 2 No gotozz
( lfyes,) what type of list is it?

 

So much for skills and lists. Now, I would like to find out about resources you
use.

104

Q. 22 I will read you a list of birding or conservation organizations, please tell
me for each one whether you are a member. (Stop reading when R. indicates
those are the only memberships.)
American Birding Association
Capital Area Audubon Society
Cornell Lab. of Ornithology
Kalamazoo Audubon Society
Michigan Audubon Society
National Audubon Society
Sierra Club
The Nature Conservancy
Whitefish Pt. Bird Observatory
10 World Wildlife Federation
11 Do you belong to any other birding or conservation organization?
( If yes) which one? (Record answer)

00NGU$ODNJ

 

Q. 23 I will new list some possible birding partners, please tell me which of
those you have birded with most of the time in the past 12 months. The
categories are not exclusive and you may choose more than one category.

1 By yourself

2 With a knowledgeable birder

3 With a friend or several friends

4 Vlfith one or more family members

5 With one or more fellow bird club members (but not as part of an organized
outing)

6 With a birding club or nature/natural history organization as part of an
organized outing

7 Vlfith none of the above (who is your birding

partner )

I have just a couple of general questions about your birding.

Q. 24 Did you participate in a birding activity, such as the Christmas Bird
Count or the Breeding Bird Survey, to promote knowledge and understanding
about birds in the past 12 months?

1 Yes 2 No

Q. 25 Have you ever participated in a competitive birding event such as a Big
Day or Big Year?
1 Yes 2 No

Q. 26 Have you ever served as a guide for organized birding field trips?
1 Yes 2 No

105

Section II: Birding Sites

In this next section I am presenting you 8 pairs of birding sites and would like you to
tell me for each pair which birding site you prefer. In some cases you may not like
either of the sites. If that is the case, I would like you to tell me the site with the
characteristics that you find more acceptable or less unacceptable.

Here Is the Information sheet for you to look over (provide Information sheet).

For the purposes of this study, it was necessary to simplify the information on the

birding sites from what some birders may want to know about a site before deciding

on whether to visit the site or not.

Let me give you some information about the birding sites. I will be repeating some of

what you have read.

All sites have the following characteristics:

1. There is a trail system in place which provides easy viewing access in habitats
where birds an be expected to be found.

2. The sites are closed to motorized vehicles and mountain bikes and hunting
and fishing are not allowed at the sites.

3. Bird checklists are not available.

4. The number of visitors to the site is about the same for all sites. You will see
some people on the trails, most of them will be birding.

5. The sites are about the same distance from your home. (If asked how far. not
any farther than you are willing to travel for a daytrip).

Q. 27 Are you familiar with birding sites that have the features I just
mentioned?

1 Yes

2 No

I will spend a few minutes to explain the site descriptions to you. Please look at the

following table (Point to top page and relevant items for respondent to follow along.)

In the upper part of the table you see information on the number of species recorded

in the spring over the past 10 years, and in the lower part of the table the number of

species you can expect at the site on a spring day.

The sites differ in the following ways:

0 there are different numbers of warbler species,

0 and there are different numbers of rare/unusual species, rare/unusual species
are species not usually found in your area,

. you will also find different numbers of other species, that is species that are
neither warblers nor rare/unusual species,

a when you add up these three categories you get the total number of species,
which also varies across the sites,

0 there are different abundance categories, designated by a capital letter. The
abundance category only varies for warbler species which is why it is
represented in its own column. The other codes are listed in the far left column .

106

At some sites the warblers appear in low numbers, that is indicated by the letter
R, and at other sites they appear in high numbers, indicated by C. The
abundance category for rare/unusual species is always R. Other species are
evenly divided, into thirds, among the abundance categories, R, U, and C. The
abundance codes are explained at the bottom of the sheet. R stands for rarely
seen, U for present but not certain to be seen, and C for usually seen in proper
habitat.

lwould like you to note that there is a difference between a rare/unusual

species and R as an abundance category. R as an abundance category just

means that there are not many individuals of a species in this category found at
the site in the spring. The species may be C, common, or U, uncommon, at other
times of the year or at other sites.

0 the habitat varies across the sites. Here I would like you to be aware that the
less diverse habitat may still harbor a large number of bird species because it
includes edge habitat where the forest habitat gradually changes to a
different type of habitat.

o and different entrance fees are charged. The entrance fees are used to maintain
the site for birds and birding. The difference in entrance fees does not reflect
differences in services other than the characteristics listed in the table.

Q. 28 Do you have any questions about the site descriptions?
1 Yes (folow i4) on questions)
2 No

I would like to come back to the Information sheet for a minute. Please, look at the
table in the middle of the sheet. (Point to the relevant items)

For the purposes of this study each abundance category is directly related to an
expert’s chance of seeing the species in that category at the site on a given day in
the spring and it is also directly related to the number of individual birds that are
present at the site on any given day in the spring. Your chance of seeing a bird may
differ from a birding expert’s chance because of differences in birding skills, for
instance. To illustrate the relationships, for species in the abundance category C,
which means: birds are usually seen in proper habitat, a birding expert’s chance of
seeing a species in that category at the site on a day in the spring is between 80
and 100%, and there will be between 31 and 101 individuals of the species at the
site that day. Similarly you find the correspondence for species in the abundance
categories U (uncommon) and R (rarely seen).

Q. 29 Thinking about your chance of seeing a species at a birding site, when
you go birding would you be most interested in knowing the abundance
category for the species, a birding expert’s chance of seeing the species at
the site, or the number of individual birds per species per day?

1 abundance category

2 expert's chance of seeing species

3 number of individuals

107

I would like you to consider a daytrip in the spring. Suppose, that this is the first
birding trip for the season. It is a nice day, no rain is in the forecast. Assume that
you have to decide between two areas that you have never visited before.

Now, please review the birding site descriptions and tell me which site you prefer

and why.
(Altereach si‘e choice aslc Wouldyou be interestedin Visiti'ngyownelened site?)
Site A Site B REASON
Yes No

30 Scenario 1 0 Cl 0 CI
31 Scenario 2 0 Cl 0 D
32 Scenario 3 Cl Cl El El
33 Scenario 4 D D D D
34 Scenario 5 Cl 0 Cl C]
35 Scenario 6 Cl E] El El
36 Scenario 7 Cl CI CI Cl
37 Scenario 8 E] Cl E] El

Q. 30 Now, I will read you a list of the birding site characteristics that
appeared in the scenarios I just showed you. For each one, please tell me
on a scale from 1-7, where one Is not at all important, and 7 is very
important, how important each one was for your choice of birding site.

not at very
all imp. imp.
how important were warblers 1 2 3 4 5 6 7
and how important were 1 2 3 4 5 6 7
rare/unusual species
how about the total number of 1 2 3 4 5 6 7
species
and the type of habitat given 1 2 3 4 5 6 7
in the table?
the entrance fees, how 1 2 3 4 5 6 7
important were they
and finally, how important 1 2 3 4 5 6 7

were the abundance codes

Next, I would like you to take a few minutes to fill out the following list of
questions on how important a variety of considerations are for yo_ur birding
activities.(Provide importance sheet)

Thank you very much.

In order to determine how representative our findings are with a larger group of
birders than I am able to interview, we need you to fill out the following questions
about yourself.

(Hovide socio-demogaphic sheet) Thank you verymuch.

108

Section N: Debrlefing Questions
To conclude the interview, I would like to ask you a handful of questions about the
birding site descriptions.

Overall, before reading the definitions of the terms used in the section on the
choice of birding sites, how familiar were you with the terms?

VERY FAMILIAR FAMILIAR NOT FAMILIAR NOT AT ALL
FAMILIAR -

Overall, how difficult or how easy was it for you to choose one site from each pair?
VERY EASY EASY DIF F ICULT VERY DIFFICULT

Please, describe what was difficult about the choices?

 

When you selected a site from each pair of birding sites, did you look at the
information on the species recorded in the spring over a 10-year period or the
species you can expect at the sites on a spring day?

10-YEAR COUNT EXPECTED NO. OF SPECIES

How did the abundance category or categories enter into your choice of a
birding site?

 

One last question:
What is the most you haVe ever paid for a daily entrance fee to go birding?
5

Thank you very much for your participation. Do you have any comments or questions
about this interview?

109

lnioanafioflheet
Eu are going on a birding daytrip in the spring. This is your first birding trip of the

 

eason. It is a nice day, no rain is in the forecast. Assume that you have to decide
tween two areas that you have never visited before.

All sites have the following characteristics:

a A trail system is in place which provides easy viewing access in habitats where
birds can be expected to be found.

0 The sites are closed to motorized vehicles and mountain bikes and hunting
and fishing are not allowed at the sites.
Bird checklists are not available.
The number of visitors to the site is about the same for all sites.
The sites are about the same distance from your home.

Terms in these columns have the same meaning in this study:

 

 

 

 

 

Abundance Explanation in Birding expert’s Number of birds of

words chance of seeing the same species
species at the site

C (common) usually seen in 80-100% 31-101
proper habitat

U (uncommon) present, but not 21-79% 4-30
certain to be seen

R (rare) rarely seen less than 20% 1-3

 

 

 

 

Rare/unusual species‘: These species are not typically seen in your area.
Rarel unusual species fall into the ‘R’ abundance category.

* Note: There is a difference between a rare/unusual species and R(are) as an
abundance category. R as an abundance category just means that there are not
many individuals of this species found at the site in the spring. The species may
be common or uncommon at other times of the year or at other sites.

Other species: Other species are species that are neither warblers nor rare
species. These species are evenly divided between the three abundance
classes (common, uncommon, rare).

Number of species you can expect to have at the site in one day: some
species have been seen at the site in the past 10 years but occur only
infrequently at the birding site and they may not be at the site every day in the
spnng.

Habitat: The habitat may be very diverse (forest, wetlands, etc.) or less diverse
(forest with edge). The less diverse habitat may still harbor a large number of
bird species because it includes edge habitat where the forest habitat gradually
changes to a different type of habitat.

Entrance fee: Entrance fees are charged at both sites to maintain and enhance
the site for birds and bird-watching. The different fees do not reflect differences
in services other than the characteristics listed in the table.

110

10

11

12

13

14

15

Q. 31 Different birders enjoy different things about birding. Please, indicate how

important die following considerations are for your birding by circling the

appropriate category.

For my birding

to study birds in their natural habitat is

to see rare bird species is ...................

 

 

 

to get outdoors for a chance to enjoy the

natural environment is ...................

to challenge my birdwatching abilities is
to compete with other birdwatchers .......
the social interactions with others is ......

to contribute to society's general

knowledge and understanding of birds is

to see new bird species is ...................
to add species to a list is .....................

to get away from everyday problems is.

to help others develop their birdwatching

skills is ...........................

to photograph, draw, or paint birds in
their natural habitat is ...........................

to be with family or relatives is ..............

to see many different species of birds is

to contribute to the conservation of
birds is .................................................

111

Level of Importance
not at very
all impor
impor tant
tant
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7
1 2 3 4 5 6 7

 

Section III: Sociodemographic Information
Q. 32 What is your gender?

1 Female

2 Male

Q. 33 What is your age group?

1 less than 20

2 20-29

3 30—39

4 40-49

5 50-59

6 60-69

7 over 69

Q. 34 What is your marital status?
1 Single
2 Married

Q. 35 What Is your current or highest level of education?
1 Some high school
2 High school graduate or equivalent
3 Some college
4 Associate's degree
5 Bachelor's degree
6 Masters degree
7 Doctorate

Q. 36 Are you retired?
1Yes 2 No

Q. 37 Do you have an occupation that allows you to spend some (or all) of
your work-time birding?
1 Yes 2 No

Q. 38 Do you have any children less than 5 years of age?
1 Yes 2 No

Q. 39 What was your approximate 1997 pre-tax household income?
1 Under $20,000
2 $20,000-39,999
3 340,000-59999
4 360,000-79999
5 380,000-99999
6 $100,000-120,000
6 Over $120,000

112

Interviewer Observations

Overall attitude of the respondent:
1 enthusiastic participation

2 cooperative

3 slightly resistant

4 very resistant

Overall understanding of the tasks required:
1 very good understanding

2 good understanding

3 some problems in understanding

4 substantial problems in understanding

Difficulty/Ease in choosing from pairs of scenarios:

1 generally easy

2 some minor difficulties (mark # scenario(s) if possible: )
3 some major difficulties (mark # scenario(s) if possible: )
4 generally difficult

Level of acceptance of entrance fee as payment vehicle:
1 no problem (e.g. no comments or questions about entrance fee)
2 minor problem (R. accepts entrance fee after minor comments or clarification)

3 major problem (R. has major reservations about entrance fee but completes
tasks as specified)

4 rejection of entrance fee (R. rejects tasks, refusing to choose between sites
because of entrance fee)

Level of effort made by R. to give meaningful answers (esp. to scenario
choices):

1 substantial effort (R. weighs characteristics in scenario and takes time
thinking about answers)

2 some effort (R. takes some time but gives up on some tasks)

3 lacking effort (R. gives up on tasks that require more than basic effort,
refuses serious consideration)

Comments:

113

 

3

25228
565.35» £5 .2!»
coco 8.85:0; .528

 

    

 

avian
mafia

 

 

 

 

o:

 

mu

 

 

.K m

£28 3.8% .o
85252 . ..onEsz

< 35

 

 

 

 

 

9.22: 50> on 3.8 :82; p

114

 

.238: .0828 c_ coon > Eco» $.28 u m .

>

 

oaoo 53> 7.22

 

avian

 

mmrwn

 

To

 

9N

 

mo—

 

mu
m
cm mm

$8828 moved» .0
85.253. .onEaz

< 35.

 

 

 

 

 

 

 

 

$828 so» on 3.» 5.55 N

 

.252. .328 E coon 2.38: n 0 Eden on o. 52.3 .2. So E328 u D ”coon >22 u m .
5 . .. . . . ..
22:22.8 . .

635.288 53 .22,
coco .mucwzo; .322

 

 

        

 

mviov

 

marwm

 

To

 

miv

 

m

 

m
m
.0 m

b08280 3.3% .o
3:23:32 .3632

< 85

 

 

 

 

 

 

9.228 :9. on 2.8 5.55 a

116

.232. 6828 :_ coon 2.83.: n

on»

once 5? .822

 

 

 

 

 

 

 

 

 

 

 

 

$4.0
8-9.”
Wm
mw.NN
3.
ms
on
.0 mm
$8823. 388% .0
8:86:22 23:52
< cam

 

 

«.22.. 50> on 2.8 coE>> e

8%;

mega».

o ”coon on 2 £2.00 .0: Sn .5on u 3 Eden $.22 .i.

m.

 

 
   

117

 

3.0. .

\

.238: .328 :_ 8035.83: u 0 Eco» on o. 52.3 .0: So 5828 .i. 3 ”coon >22 .1. m .
8; . ,. . ,

     

. .8.

 

case 53. 7.22

 

STNOP

 

No ..-mm

 

m

 

 

8N

 

5N
on

 

 

.m m

€028 moved» .0
3cmuc3< 83:52

< 35

 

 

 

 

 

9.228 :9. os 23 :03; a

118

 

a;

2:22.».
385.858 55 .295
:80 828.33.19.22

 

 

875—.

 

No 7.8

 

To

 

m.N

 

5N

 

wow
m

.m mm

E0823 8202.8 .0
3coo§n< cfiEaz

< 33

 

 

 

 

 

 

 

.2228 so» on 2.» :255 8

 

 

..... ..1

.252. .828 c. :03 2.32.. n o ”coon on o. 58:8 .0: So €828 u 3 Eco» $.22

...w ... .81

     

   

3

 

omoo £5 .822

 

mo Tmov

 

No Tam

 

To

 

mi?

 

3N

 

 

 

.0 m

bonnie 3.00% .0
85252 58:52

< 35

     

 

 

 

 

9.22: so» on 28 :82; 2.

"N...

120

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.859. Back. 5 zoom 2.33 n 0 Emma on
www
3:205
.99??ch 5.3 .2“;
:30 £95.53 .692
mm. 2 v9
Nov 2 mm
m
nub. mm
mam
5N
on
.0 mm
r328 3.00% .o
3cmuc3< .3532
< 33

 

 

$895 30> on 33 5.55 a

121

APPENDIX B

BINOMIAL LOGIT REGRESSIONS

122

_m>m. 853:8 $8 6 .585ch t ”.02.... 853:8 $3 “a E3553 .

 

 

$3 «.3 8% $3 $8 §m «am 93.0. *2. m 25
$8 $$ *3 $3 $8 $8 $3 $8 *8 < 27.
86605
£8 £3 *3 $3 $3 :3 «am *8 $3 .99 2.850
3.89 3.8% 8.25- 8.8”- 8.9”- ~38- ~03? 3.8m- 93$ .6“. 32:2... 8..
38.9 68.3 $8.9 $8.3 $8.3 $8.9 $8.9 68.8 58.3
Bod 886 Sod 38¢ ~8o.o Bod- mood- wood- :m~o.o- 3,. 85:5
68.9 68.9 88.9 68.3 38.3 68.9 68.9 :89 82:
~26 53 ~25 63 $3 veto BE .83 .m~.o aim:
98.9 are
2.3 :36 Em; “.0 8:352
38.9
2.386 868m .26 .o .02
68.3
:Rod 36QO 0.6.. ho .oz
:89
2.086 29995 he .02
~89 :85 ~89 68.9 a~te $8.9 $8.9 $8.8 .
:83 :8~.m :83 :83 :o~v.o :83 :33 ..3o.o xouc. 236235
.955 z 325.”. $25.”. Es
.383 case—cs. 58.5w 95:55 x2555: 33822 3.33 8.0on 9883.0 o.nm_._m>

 

$355.3 :_ Echo 23cm.» 53> 365:3 220530 96 notoaom

83%.; 350 22 828.2. 3.83.35 "3m 2...:

123

85. 8:03:00 $3 6 9:85:06 1 ”.96. 853:8 $3 8 88557. .

 

 

$3 $3 $3 $3 $05 $3 $3 $~u m 2?.
$3 $3 $3 $3 $3 $3 $3 $3 < 25

3862:
$8 $3 $3 $3 $3 $3 $mm $3 .92 38:00
3.05- 8.08. 2.9m- mod—b- 8.3m- omdom- Nwdom- 96mm- .«om 3959:... ac.—

Amoodv $00.8 $00.9 600.9 $8.8 $8.8 $8.8 43.8.8
wood Sod mood Sod oocod moood- wood- tamed. m3 oucmhcm

629
.88 55:...

6:9
2.3.: .995 :o 8:352

:89
2.888 88% .98 8 .02

38.9
:39: 3.0QO 9.8 .0 .oz

98.9
2.98.: 328:; 90 .oz

$8.8 Cong 38.9 85.3 69.8 88.8 2.8.9

3306 2.899 2.086 .3596 2.9.3 2.30.: 2.306 xmuE 3.90285

:95.» I 33:21 3.:
.898 58:62 .8856 m.:o::a:w 5.55.22 8.3.22 3.00% @2835 223:0)

 

$355.3 :_ 20:0 Emucmfi 53> mofiezmo 206530 2: 3:33.
con. 02.8.5 2.: 9.83.0:- Efigfiofi "vwé 2%...

124

BIBLIOGRAPHY

125

Bibliography

Abdelbasit, KM. and R. L. Plackett “Experimental Design for Binary Data.” Journal
of the American Statistical Association 78 (381): 90-98, 1983.

Adamowicz, W., J. Louviere, J. Swait. “Introduction to Attribute-Based Stated Choice
Methods.” January 1998. (Final Report).

Addelman, S. “Symmetrical and Asymmetrical Fractional Factorial Plans.”
Technometn‘cs 4 (1): 47-58, 1962 b.

Alberini, A “Optimal Designs for Discrete Choice Contingent Valuation Surveys:
Single-Bound, Double-Bound, and Bivariate Models.” Journal of
Environmental Economics and Management 28 (3): 287-306, 1995.

Ameriwn Birding Association. “A Birders Resource Guide: 1998 ABA Membership
Directory, Member Handbook, and Yellow Pages.” Supplement to Birding
(5). 1998. ‘

. “1997 Membership Directory and Yellow Pages: A Birder's Resource
Guide.” Supplement to Birding 29 (5). 1997.

. “1996 Membership Directory and Yellow Pages.” Supplement to Birding 28
(5). 1 996.

_‘ .1996 ABA Convention. Park City, Utah, June 17-23’ Supplement to
VVrnging It 7(12): 1995.

. "ABA Survey of Members - 1994." Colorado Springs, CO: Amerimn Birding
Association.

Applegate, J.E., R.A Otto, and JA Buttitta. “A Cluster Analysis of Appreciative
Wildlife Users.” Wildlife Society Bulletin. 10 (1): 65-69, 1982.

Barbier, E.B., J.C. Burgess, and C. Folke. Paradise Lost? London: Earthscan
Publications Ltd., 1994.

126

Bibby, C.J., N.J. Collar, M.J. Crosby, M.F. Heath, C. Imboden, T.H. Johnson, AJ.
Long, AJ. Stattersfiel, and S.J.Thirgood. Putting Biodiversity on the Map:
Priority Areas for Global Conservation. Cambridge: lntemational Council for
Bird Preservation, 1992.

Bockstael, N.E., KE. McConnell, and LE. Strand. “Recreationf' In: Braden, J.E. and
CD. Kolstad (eds.). Measuring the Demand for Environmental Quality. North
Holland: Elsevier Science Publishers, 1991.

Boettner, J., J. Buecking, and G. Hickok " A Checklist of Genessee County Birds.
Including For-Mar Nature Preserve and Arboretum.” Genessee Audubon,
Inc., 1 983.

Boxall, PC. and BL. McFarlane. “Human Dimensions of Christmas Bird Counts:
Implications for Nonconsumptive Wildlife Recreation Programs.” Wildlife
Society Bulletin. 21: 390-396, 1993.

Buchanan, T., E.R. Morey, and 0M. Waldmann. “Happy (Hypothetical) Trails to
You: The Impact of Trail Characteristics and Access Fees on a Mountain
Biker’s Trail Selection and Consumer’s Surplus.” March 17, 1998. (mimeo).

Burley, F.W. “Monitoring Biological Diversity for Setting Priorities in Conservation.”
In: VWson, E.O., eds. Biodiversity. National Academy Press: Washington,
DC, 1988.

Burton, P.J., AC. Balisky, LP. Coward, S.G. Cumming, and DD. Kneeshaw. “The
Value of Managing for Biodiversity.” The Forestry Chronicle 68 (2): 225-237,
1 992.

Butcher, G. "Demographics of ABA Members." Wrnging It 7(4): 2-3, April, 1995.

Butler, JR. and GD. Fenton. Bird Watchers of Point Pelee National Park, Canada:
Their Characteristics and Activities, with Special Consideration to Their
Social and Resource Impacts. Paper Presented at First National Symposium
on Social Science in Resource Management, Oregon State University,
Corvallis, Oregon, May 12-16, 1986.

Cook, RD. and OJ. Nachtsheim. “A Comparison of Algorithms for Constructing
Exact D-Optimal Designs.” Technometrics 22 (3): 315-324, 1980.

127

Cooper, D. “Genes For Sustainable Development.” In: Shiva, V.P.A, H. Schi'icking,
A Gray, L. Lohmam, and D. Cooper (eds). Biodiversity: Social and
Ecological Perspectives. New Jersey. Zed Books, 1991.

Dillman, D.A Mail and Telephone Surveys. The Total Design Method. New York:
Vlfiley-lnterscience Publication. 1977.

Evers, D. and J. Granlund. “The Checklist to the Birds of Whitefish Point Bird
Observatory'.1 991 .

Fetterrnan, D.M. “Ethnography.” In: Bickman, L. and DJ. Rog. Handbook of Applied
Social Research Methods. Sage Publications, 1998.

Fowler Jr. , F. J. Improving Survey Questions. Design and Evaluation. Applied Social
Research Methods Series, Vol. 38. Sage Publications: Thousand Oaks,
1 995.

Freeman III, AM. “Factorial Survey Methods and Willingness to Pay for Housing
Characteristics: A Comment.” Journal of Environmental Economics and
Management 20: 92-96, 1991.

Friends of Point Pelee and Pelee Island Winery. Point Pelee National Park 8
Vicinity & Pelee Island. Checklist of Birds. Leamington, Ontario, Canada:
Friends of Point Pelee and Pelee Island Winery, March 1994.

Graham, J.R. Point Pelee National Park and Vicinity. Seasonal Status of Birds.
Leamington, Ontario, Canada: Friends of Point Pelee, May 1997.

Gray, A “The Impact Of Biodiversity Conservation On Indigenous Peoples.” In:
Shiva, V.P.A, H. Schi‘icking, A Gray, L. Lohmann, and D. Cooper (eds.).
Biodiversity: Social and Ecological Perspectives. New Jersey: Zed Books,
1991.

Greene, W. H. LIMDEP. Version 7.0. User’s Manual. Plainview, New York:
Econometric Software, Inc., 1995.

Econometric Analysis. New Yorlc Macmillan Publishing Co., 1993.

128

Hanemann, W.M. "Economics and the Preservation of Biodiversity." In: Vlfilson,
E.O., eds. Biodiversity. National Academy Press: Washington, DC, 1988.

Hensher, D.A "Stated Preference Analysis of Travel Choices: The State of
Practice.” Transportation. 21 2107-1 33.

Jakobsson, KM. and AK Dragun. Contingent Valuation and Endangered Spades.
Methodological Issues and Applications. Cheltenham, UK Edward Elgar,
1 996.

Jacquemot, A and F. L. F ilion. “The Economic Significance of Birds in Canada.” In:
Diamond, AW. and FL. Filion (eds). The Value of Birds. Cambridge, UK:
lntemational Council for Bird Preservation, Technical Publication No. 6,
1986.

Jenkins Jr., RE. “Information Management for the Conservation of Biodiversity." In:
VViIson, E.O., eds. Biodiversity. National Academy Press: Washington, DC,
1 988.

Johnson, 8. Oral Communication

Johnston, R.J., T.F. Weaver, L.A Smith, and SK Swallow. “Contingent Valuation
Focus Groups: Insights from Ethnographic Interview Techniques.”
Agricultural and Resource Economics Review. 56-69, April 1995.

Jones, J.O. Where the Birds Are. New York: William Morrow and Company, Inc,
1 990.

Kanninen, B. J. “Optimal Experimental Design for Binary Choice Experiments.”
1 998. (mimeo).

Kellert, S.R. “Birdwatching in American Society." Leisure Sciences. 7(3): 343-360,
1 985.

Kerlinger, P. “Birders as Ecotourists.” Birder’s World: 74-76, April 1995a.

Kerlinger, P. “The Economic Impact of Birding Ecotourists on Ten National Wildlife
Refuges.” Winging lt: 10-11, September 1995b.

129

Kim, S., 8., D. Scott, J.L. Crompton. “An Exploration of the Relationships Among
Social, Psychological Involvement, Behavioral Involvement, Commitment,
and Future Intentions in the Context of Birdwatching.” Journal of Leisure
Research. 29(3): 320-341, 1997.

Kitching, J. Birdwatchers Guide to Wildlife Sanctuaries. Arco Publishing Co.: New
York, 1976.

Kuhfeld, W.F., R.D. Tobias, and M. Garratt. “Efficient Experimental Design with
Marketing Applications”. Journal of Marketing Research. 31 (1994): 545-557.

Lazari, AG. and DA Anderson. “Designs of Discrete Choice Set Experiments for
Estimating Both Attribute and Availability Cross Effects.” Journal of Marketing
Research 31: 375-383, 1994.

Lerg, J. “Birds of Rose Lake. “ East Lansing, Michigan: Michigan Department of
Natural Resources. \Nildlife Division. Rose Lake Wildlife Research Center.
Undated. (mimeo). (undated).

Louviere, J.J. and G. Woodworth. “Design and Analysis of Simulated Consumer
Choice or Allocation Experiments: An Approach Based on Aggregate Data.”
Journal of Marketing Research 20: 350-367, 1983.

Lovejoy, T.E. “Species Leave the Ark One by One”. In Norton, B.G. (ed.). The
Preservation of Species. The Value of Biological Diversity. Princeton
University Press. 1986.

Magurran, E. Ecological Diversity and Its Measurement. Princeton University Press,
1 988.

Margules, C. and MB. Usher. “Criteria Used in Assessing Wildlife Conservation
Potential: A Review.” Biological Conservation 21 : 79-109, 1981.

McFadane, B.L. Oral Communication 06-29-1999.
“Socialization Influences of Specialization Among Birdwatchers." Human

Dimensions Of Mldlife. 1 (1): 35-50, 1996.

130

“Specialization and Motivations of Birdwatchers.’ Wildlife Society Bulletin.
22: 361370, 1 994.

McFarlane, B.L., and PC. Boxall. “Participation in Wildlife Conservation by
Birdwatchers.’ Human Dimensions of Ill/ildlife. 1 (3): 1-14, 1996.

McNeely, J. Economics and Biological Diversity: Developing and Using Economic
Incentives to Conserve Biological Resources. Gland, Switzerland: IUCN,
1 988.

McNeely, J.A, KR. Miller, W.R. Reid, R.A Mitterrneier, and TB. Werner.
Conserving the World’s Biological Diversity. Gland, Switzerland: IUCN, 1990.

McWhirter, D. “Birds of the Lansing Area. An Annotated Checklist.” 1997. (mimeo).

Mitchell, RC. and RT. Carson. Using Surveys to Value Public Goods. The
Contingent Valuation Method. Washington, DC: Resources for the Future,
1 989.

Montgomery, 0.0. Design and Analysis of Experiments. New York: John Wiley and
Sons, 1 991 .

Norton, 8. "Commodity, Amenity, and Morality." In: Wilson, E.O., eds. Biodiversity.
Washington, DC: National Academy Press, 1988.

Pearce, D.W. Economic Values and the Natural World. Cambridge, MA: MIT Press,
1993.

Pearce, D.W. and D. Moran. The Economic Value of Biodiversity. Earthscan
Publications, LTD, 1 994.

Plackett, R. L. and JP. Burman. “The Design of Optimum Mulitfactorial Experiments.”
Biometn'ka 33: 305-325, 1946.

Polasky, S. , C. Langpap, S. Pavich, C. Vossler, N. Bergeron, and M. Jaspin.
“Bibliography on the Conservation of Biological Diversity: Biological!
Ecological, Economic, and Policy Issues."
http:/Iosu.orst.eduldeptlag_resrc_econlbiodivlbiblio.html, April 8, 1999.

131

Rae, D. A The Value to Visitors of Improving Visibility at Mesa Verde and Great
Smoky National Parks. In: Rowe, RD. and LG. Chestnut. Managing Air
Quality and Scenic Resources at National Parks and Wildemess Areas.
Westview Press: Boulder, Colorado, 1985.

Randall, A 'What Mainstream Economists Have to Say About the Value of
Biodiversity." In: VWIson, E.O., eds. Biodiversity. National Academy Press:
Washington, DC, 1988.

Reid, W.V., J.A McNeely, D.B. Tunstall, D.A Bryant, and M. Winograd. Biodiversity
Indicators for Policy-Makers. Washington, DC: World Resources Institute,
1993.

Reid, W.V. and KR. Miller. Keeping Options Alive. The Scientific Basis for
Conserving Biodiversity. World Resources Institute, 1989.

Rubey, L. and F. Lupi. “Predicting the Effects of Market Reform in Zimbabwe: A
Stated Preference Approach.” American Journal of Agricultural Economics 79
(1): 89-99, 1997.

Ryan, J.C. Life Support: Conserving Biological Diversity. Worldwatch Paper 108.
Worldwatch Institute, 1 992.

Salant, P. and DA Dillman. How to conduct your own survey. New York : Wiley,
1994.

Sauer, J. R., J. E. Hines, G. Gough, l. Thomas, and B. G. Peterjohn. 1997. The
North American Breeding Bird Survey Results and Analysis. Version 96.4.
Patuxent Wildlife Research Center, Laurel, MD.
http:/lwww.mbr.nbs.govlbbslbbs.html

Sault Naturalists Checklist. “Birds of Sault Ste. Marie.” (mimeo). (Undated).

Schucking, H. and P. Anderson. “Voices Unheard and Unheeded.” In: Shiva, V.P.A,
H. SchCicIdng, A Gray, L. Lohmann, and D. Cooper (eds). Biodiversity:
Social and Ecological Perspectives. New Jersey: Zed Books, 1991.

132

Scott, J.M., B. Csuti, J.D. Jacobi, and J.E. Estes. “Species Richness. A Geographic
Approach to Protecting Future Biological Diversity”. BioScience 37(11): 782-
788, 1 987.

Smith, R., W. Booth, and C. Witkoske. “The Birds of Berrien County Michigan. 30
Year Summary 1962-1991.“ Berrien Audubon Society, Undated. (mimeo).
(undated).

Solow, A, S. Polasky, and J. Broadus. "On the Measurement of Biological
Diversity." Journal of Environmental Economics and Management 24 (1 )260-
68, 1993.

Solow, AR. and S. Polasky. "Measuring Biological Diversity." Environmental and
Ecological Statistics 1:95-107, 1994.

Spradley, JP. The Ethnographic Interview. Fort Worth: Harcourt Brace Jovanovich
College Publishers, 1979.

Steffens, K and JP. Hoehn. “Valuing Biodiversity: Issues and Illustrative Example.”
Michigan State University, Department of Agricultural Economics, Staffpaper
97-7, 1 997.

Sudman, 8., NM. Bradbum, and N. Schwarz. Thinking About Answers. The
Application of Cognitive Processes to Survey Methodology. San Francisco:
Jossey-Bass Publishers, 1996.

Swaney, J.A, and RI. Olson. "The Economics of Biodiversity: Lives and Lifestyles."
Journal of Economic Issues 26(1):1-25, March, 1992.

Swingland, I.R “Tropical Forests and Biodiversity Conservation: A New Ecological
Imperative." In: Barbier, E.B. (ed). Economics and Ecology: New Frontiers
and Sustainable Development. London: Chapman and Hall, 1993.

The Nature Conservancy. “Birds of The Les Cheneaux Islands Area.” (Undated).
US. Department of the Interior, Fish and Wildlife Service. “Birds. Seney, National

Wildlife Refuge.“ Seney, Michigan, May 1991.

133

. “Birds of Shiawassee National \lVildlife Refuge/MI.“ Saginaw, Michigan,
May 1979.

von Kammen, W. B. and M. Stouthamer-Loeber. “Practical Aspects of Interview Data
Collection and Data Management.“ In: Bickman, L. and D. J. Rog (eds).
Handbook of Applied Social Research Methods. Sage Publications. 1998.

Vuong, Q. H. “Likelihood Ratio Tests for Model Selection and Non-Nested
Hypotheses.“ Econometn'ca 57 (2): 307-333, 1989.

Wauer, RH. The Visitor's Guide to the Birds of the Rocky Mountain National Parks.
John Muir Publications: Santa Fe, New Mexico, 1993.

Wauer, R'Protile of an ABA Birder.“ Birding 146-154, June 1991.

Weiss, RS. Learning from Strangers. The Art and Method of Qualitative lntenri'ew
Studies. New York: The Free Press. 1994.

Weitzman, M. “On diversity.“ Quarterly Journal of Economics 107(2): 363-405, 1992.

Wiedner, D. and P. Kerlinger. "Economics of Birding: A National Survey of Active
Birders." American Birds 44 (2):209-213, 1990.

Williams, PR, RI. Vane-Wright and OJ. Humphries. "Measuring Biodiversity for
Choosing Conservation Areas." In: LaSalIe, J. and ID. Gauld (eds).
Hymenoptera and Biodiversity. CAB lntemational, 1993.

Wilson, 5.0. “The Current State of Biological Diversity." In: \MIson, E.O. (ed).
Biodiversity. Washington, DC: National Academy Press, 1988.

Yen, S.T., RC. Boxall, and W. L. Adamowicz “Donations for Environmental
Conservation: An Econometric Analysis" 1996. (mimeo).

Zwerina, K, J. Huber, and WP. Kuhfeld. “A General Method for Constructing
Efficient Choice Designs.“ September 1996. (mimeo).

134