The total economic value of threatened, endangered and rare species, and updated meta-analysis

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / e c o l e c o n

ANALYSIS

The total economic value of threatened, endangered and rare
species: An updated meta-analysis
Leslie Richardson, John Loomis⁎
Dept. of Agricultural and Resource Economics, Colorado State University, Fort Collins, CO 80523-1172, United States

AR TIC LE D ATA

ABSTR ACT

Article history:

This paper updates a 1996 meta-analysis of studies using the Contingent Valuation Method

Received 17 June 2008

(CVM) to value threatened, endangered and rare species published in this journal by Loomis

Received in revised form

and White. A variable for studies conducted in or after 1995 was added to the model to test if

29 October 2008

new studies are systematically different from old studies and identify which explanatory

Accepted 30 October 2008

variables influencing willingness-to-pay (WTP) for these species have changed over time.

Available online 4 December 2008

Generally newer studies yielded higher WTP. Variables such as the change in the size of the
species population, payment frequency, survey mode, response rate, type of respondent,

Keywords:

type of species, and a new variable for whether a species is a ‘charismatic megafauna’ or not,

Threatened and endangered species

whether the species has use and nonuse value versus nonuse value only and year of the

Meta-analysis

study, were found to significantly influence WTP. This model is used in a benefit transfer

Benefit transfer

example and a comparison of original study estimates and model estimates is made to

Willingness-to-pay

compare its accuracy. The average within sample benefit transfer error was 34–45%.
© 2008 Elsevier B.V. All rights reserved.

1.

Introduction

Biodiversity is increasingly threatened given current trends in
human population growth and development with the number
as well as the rate of plant and animal extinctions on the rise.
According to the World Conservation Union's 2007 Red List, one
in four mammals, one in eight birds, one third of all amphibians,
and 70% of the worlds assessed plants are now endangered.
There is an awareness of the problems that arise with the loss of
biodiversity and this is reflected in the Endangered Species Act
(ESA) in the U.S.A. and similar legislation in numerous countries
around the world.
Currently, economic analyses may not be incorporated in
species listing decisions under the Endangered Species Act but
can be incorporated in designating critical habitats and
developing recovery plans after a species is listed. However,
there has been considerable concern about how these

economic analyses are conducted for critical habitat. For
instance, Defenders of Wildlife and their Conservation Economics Program has argued that the Fish and Wildlife
Service's current practice of monetizing costs while qualitatively describing benefits under the ESA is flawed. They call for
more consistent measures of the benefits provided by species
(Defenders of Wildlife, 2004). In addition, in 2004, the National
Wildlife Federation released a report documenting how the
Bush administration used flawed economic data to cut in half
critical habitat designations under the ESA, overestimating
the costs while ignoring many of the benefits of proposed
designations (National Wildlife Federation, 2004). These concerns provide justification for the need of a consistent measure of benefits provided by threatened, endangered and rare
species.
The Total Economic Value of the majority of these species
consists of both recreational use and nonuse (existence and

⁎ Corresponding author. Fax: +1 970 491 2067.
E-mail address: jloomis@lamar.colostate.edu (J. Loomis).
0921-8009/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2008.10.016

1536

Table 1 – WTP per household ($2006) for threatened and endangered species
Willingness to pay
(2006$)
Reference
Bell et al. (2003)

Survey Species
date
2000

Salmon

Gain or loss Size of Lump Annual CVM
change sum
method
Gain

100%

$138.64

DC

Survey region

Sample Response
size
rate
357

49.1%

100%

$141.27

Willapa Bay, WA
households

386

61.7%

Coos Bay, OR
households

424

58.4%

Tillamook Bay, OR
households

347

53.2%

Yaquina Bay, OR
households

357

59.7%

NM residents
TX and US households
Visitors
WI households

726
316
254
365

64.0%
36.0%
67.0%
73.0%

Trust fund
Foundation
Foundation
Foundation

Ely and St. Cloud, MN
households
NM

352

56.1%

One-time tax

723

42.0%

158

30.6%

Increase state
taxes
Lifetime membership

121

86.0%

Lifetime membership

189

46.6%

Lifetime membership

335

69.6%

Lifetime membership

345

69.6%

157

27.3%

Trust fund

688

77.1%
54.4%

Trust fund
Trust fund

$90.64
Avoid loss

100%

$57.99
$47.70

Avoid loss

100%

$91.99
$28.39

Avoid loss

100%

2001

Silvery minnow
Whooping crane
Whooping crane
Bald eagle
Striped shiner
Gray wolf

Avoid
Avoid
Avoid
Avoid
Avoid
Avoid

loss
loss
loss
loss
loss
loss

100%
100%
100%
100%
100%
100%

Avoid loss

100%

$134.00
$87.84
$37.77
$43.69
$68.55
$21.21
$8.32

Berrens et al. (1996)
Bowker and Stoll
(1988)
Boyle and Bishop
(1987)
Chambers and
Whitehead (2003)
Cummings et al. (1994)

1995
1983

1994

Squawfish

Duffield (1991)

1990

Gray wolf

Reintroduction

$93.92

DC

Duffield (1992)

1991

Gray wolf

Reintroduction

$162.10

DC

Duffield et al. (1993)

1992

Gray wolf

Reintroduction

$37.43

DC

USDOI (1994)

1993

Gray wolf

Reintroduction

$28.37

DC

USDOI (1994)

1993

Gray wolf

Reintroduction

$21.59

DC

Duffield and Patterson
(1992)

1991

Arctic grayling

33%

$26.47

PC

Yellowstone National
Park visitors
Yellowstone National
Park visitors
ID, MT, WY
household
ID, MT, WY
household
ID, MT, WY
household
US visitors

$19.84

1996

Arctic grayling
Mexican spotted owl

33%

Giraud et al. (1999)

PC
DC

US visitors
US households

1984

Improve 1 of 3
rivers
Avoid loss

$22.64
$11.65

$68.84

DC
DC
DC
DC
DC
DC
OE

Annual tax—high
income
Annual tax—low
income
Annual tax—high
income
Annual tax—low
income
Annual tax—high
income
Annual tax—low
income
Annual tax—high
income
Annual tax—low
income
Annual tax—high
income

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

Grays Harbor, WA
households

$91.55
Gain

Payment
vehicle

Giraud et al. (2002)
Hageman (1985)

2000
1984

Hageman (1985)

1984

Hagen et al. (1992)
King et al. (1988)
Kotchen and Reiling
(2000)
Layton et al. (2001)

Loomis and Ekstrand
(1997)
Loomis and Larson
(1994)

Avoid
Avoid
Avoid
Avoid
Avoid
Avoid

loss
loss
loss
loss
loss
loss

100%
100%
100%
100%
100%
100%

1985
1997

Bighorn sheep
Peregrine falcon

Avoid loss
Gain

100%
87.5%

1998

Gain

1994

Eastern WA and Columbia River
Freshwater Fish
Eastern WA and Columbia River Migratory
Fish
Western WA and Puget Sound Freshwater
Fish
Western WA and Puget Sound Migratory
Fish
Western WA and Puget Sound Saltwater
Fish
Salmon and steelhead

1996

DC
PC
PC
PC
PC
DC

AK and US households
CA households

$16.99

50%

$210.84

Gain

50%

$146.57

Gain

50%

$229.31

Gain

50%

$307.76

Gain

50%

$311.31

Gain

600%

$79.53

DC

Salmon and steelhead
Salmon and steelhead
Mexican spotted owl

Gain
Gain
Avoid loss

600%
600%

$98.41
$91.67
$51.52

Gain
Gain
Gain
Gain
Gain
Gain
Gain
% chance of
survival

50%
100%
50%
100%
100%
100%
100%
99%

Olsen et al. (1991)

1989

Gray whale
Gray whale
Gray whale
Gray whale
Salmon and steelhead

Reaves et al. (1994)

1992

Red-cockaded woodpecker

Rubin et al. (1991)

Samples and Hollyer
(1989)
Stanley (2005)

1991

$70.90
$36.41
$34.50
$45.94
$39.80
$130.19

1987

1986
2001

No. Spotted owl

Monk seal
Humpback whale
Riverside fairy shrimp

% chance of
survival

Avoid loss
Avoid loss
Avoid loss

$32.27

63.6%
21.0%

Increase federal tax
Increase federal tax

21.0%

Increase federal tax

US households

1653
180
174
180
174
409

46.0%

OE
DC

AZ households
ME residents

550
206

59.0%
63.1%

Taxes and wood
prices
Foundation
One-time tax

CE

WA households

801

68.0%

Monthly payment

CA households

(converted to annual)

284

77.0%

DC
DC
MB

Clallam County, WA
households
WA households
US households
US households

467
423
218

68.0%
55.0%
56.0%

$23.65
$26.53
$36.56
$43.46
$42.97
$95.86
$121.40
$14.69

OE
OE
OE
OE
OE
OE
OE
OE

CA households
CA households
CA visitors
CA visitors
Pac. NW households
Pac NW HH option
Pac. NW anglers
SC and US households

890
890
1003
1003
695
482
225

54.0%
54.0%
71.3%
71.3%
72.0%
72.0%
72.0%
53.0%

99%
99%
50%

$20.46
$13.14
$38.61

DC
PC
OE

WA households

223
234
249

52.0%
53.0%
23.0%

75%
100%
100%
100%
100%

$39.99
$60.84

OE
OE
DC

HI households

165

40.0%

Orange County, CA
households

242

32.1%

$165.80
$239.53
$28.38

PC

Increase federal tax

Protection fund
Protection fund
Electric bill

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

Loomis (1996)

1990

Steller sea lion
Bottlenose dolphin
Northern elephant seal
Gray-blue whale
Sea otter
No. spotted owl

Recovery fund

Unspecified

Preservation fund
Money and time
Annual tax

1537

(continued
ononnext
(continued
next page)
page)

1538

Willingness to pay
(2006$)
Reference
Stevens et al. (1991)

Survey Species
date
1989

Wild Turkey

Gain or loss Size of Lump Annual CVM
change sum
method
Avoid loss

100%

$11.38

DC

Avoid loss

100%

$15.36

OE

Atlantic salmon
Atlantic salmon
Bald eagle

Avoid loss
Avoid loss
Avoid loss

100%
100%
100%

$10.00
$11.12
$45.21

DC
OE
DC

100%
300%

$31.85

Swanson (1993)

1989

Bald eagle
Bald eagle

Avoid loss
Increase in
populations

Whitehead (1991,
1992)

1991

Sea turtle

Avoid loss

300%
100%

$349.69
$244.94
$19.01

Survey region

Sample Response
size
rate

Payment
vehicle

New England
households
New England
households
MA households

339

37.0%

Trust Fund

169

30.0%

Trust fund

New England
households

339

37.0%

Trust fund

OE
DC

WA visitors

747

57.0%

Membership fund

OE
DC

WA visitors
NC households

207

35.0%

Preservation fund

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

Table 1 (continued)

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

bequest) values, which can be measured by eliciting willingness-to-pay (WTP) for the preservation of a particular
species. However, it is important to note that some species
have nonuse values only. Currently, one of the most accepted
methods used to estimate the Total Economic Value provided
by species is the Contingent Valuation Method (CVM). This
method employs the use of surveys outlining a hypothetical
market or referendum (Mitchell and Carson, 1989). It has been
found that people are willing to pay a small portion of their
income towards the protection of endangered or rare species
for a variety of reasons. While numerous CVM studies valuing
threatened and endangered (T&E) species have been conducted, especially over the last 30 years, performing an
original study to value every single species would be costly
and time-consuming.
An important alternative is benefit transfer, a form of
secondary research which has been used extensively in the
past couple of decades. Rosenberger and Loomis (2003) define
this term as the use of existing data or information in settings
other than for what it was originally collected. The two forms
of benefit transfers are value transfer and function transfer. In
the former, a single estimate or average of multiple estimates,
is transferred from the original site where primary research
was conducted (called the study site) to a site with similar
species that are being evaluated (called the policy site). In a
function transfer, a statistical function is estimated based on
the original studies and then this function is applied to the
study site to calculate a value tailored to the study site.
Function transfers are generally viewed as more accurate than
value transfers because they can be tailored to account for
differences in the site characteristics (Rosenberger and
Loomis, 2003).
The two types of function transfers include demand functions and meta-regression analysis functions. The focus of this
paper is on the meta-analysis, which helps to statistically
explain the variation in the values obtained in different studies.
For instance, looking at various CVM studies valuing threatened
or endangered species, this method will help control for the
effect of different study variables, such as question format,
payment frequency, or type of respondent, to calculate WTP for
that type of species. This information and its use in benefit
transfer may assist in improving quantification of the economic
benefits of critical habitat.
The purpose of this paper is to: (1) update information on
the economic valuation of threatened, endangered, and rare
species first published by Loomis and White in this journal
more than twelve years ago. Using their same model, the first
goal is to add new or overlooked CVM studies valuing
threatened and endangered species conducted in the U.S.;
(2) add a variable to the model to test if WTP from the new
studies (conducted in or after 1995) are systematically
different from old studies (conducted prior to 1995). This
will help to identify whether people's valuation of threatened and endangered species has changed over time; (3) test
new specifications, such as how a species ‘charisma’ affects
the value placed on it, to identify an effective model which
can be used in benefit transfer; (4) outline an example of how
meta-analysis regression functions can be used in benefit
transfer to estimate the value of various threatened and
endangered species.

2.

Methodology

2.1.

Data sources

1539

Various economic and scientific research databases were
searched, including EconLit, JSTOR, and Web of Science.
Eleven new or overlooked CVM studies valuing threatened,
rare, or endangered species conducted inside the United
States were found, consisting of 29 estimates of value. A
summary of these added studies, as well as the 20 studies
identified in the Loomis and White (1996) meta-analysis can
be found in Table 1. All WTP values were converted to U.S.
dollars in a 2006 base year using the Consumer Price Index for
comparability. Under the CVM Method column in Table 1, DC
represents surveys which used a dichotomous choice question format, OE represents surveys which used an open-ended
format, PC represents those which used a payment card
question format, CE stands for studies using a conjoint, or
choice experiment, technique, and MB represents a multiplebounded format. Looking at Table 1, the first study by Bell et al.
(2003) uses an annual tax as the payment vehicle in their
survey. The low income group represents respondents with
incomes below $30,000 and respondents with incomes not
below $30,000 are categorized as high income.

2.2.

Meta-analysis model

In order to systematically explain the variation in the WTP values
for threatened and endangered species obtained in different
studies, a meta-analysis regression approach can be undertaken.
The first meta-analysis on CVM studies valuing threatened and
endangered species was published by Loomis and White (1996)
and the first goal of this paper is to compare the meta-analysis
regression results from their study with the results of a
regression including all 31 studies found to date. In order to
accurately compare findings, before adding any new variables to
the model, the same model used in their study is used here. Their
model takes on the following equation, which includes the
variables that economic theory would suggest as important:
WTP = b0 + b1 CHANGESIZE + b2 PAYFREQUENCY
+ b3 CVFORM + b4 VISITOR + = 
 b5 FISH + b6 MARINE
+ b7 BIRD + = 
 b8 OTHER 
 b9 RESPONSERATE
+ = 
 b10 STUDYYEAR:
Willingness to pay for a particular species is a function of:
the percentage change in the species population proposed in
the survey (CHANGESIZE); payment frequency, coded 1 for a
one-time payment or purchase of a lifetime membership and
0 for an annual payment amount (PAYFREQUENCY); contingent valuation format, coded 1 for studies using a
dichotomous choice question format in their survey and 0
for those using an open-ended or payment card format
(CVFORM); whether the survey respondents were visitors,
coded 1, or households, coded 0 (VISITOR); dummy variables
broken down by groups of similar species being valued,
including fish, mammals, marine mammals, birds, and other,
coded 1 if that represents the species being valued, 0
otherwise. MAMMAL is the omitted category from the

1540

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

model in Loomis and White (1996) and is omitted from this
model as well. For the purposes of this meta-analysis
regression two included studies valued species that did not
quite fit into these similar species group variables. So a
dummy variable labeled OTHER was included to account for
the Riverside fairy shrimp and the sea turtle, coded 1. The
final model variables include RESPONSERATE, which is
simply the survey response rate, and STUDYYEAR, which is
the year the study was performed.
The signs in front of the variables indicate their hypothesized effect on WTP, based on the results from previous
literature. The change in the size of the species population
being valued should have a positive effect on WTP because
value should increase with the size of the population. The
positive sign on CVFORM is due to the consistent findings in
the literature that when valuing public goods, dichotomous
choice, referendum format questions result in higher
estimates than open-ended questions, all else constant.
Brown et al. (1996) summarize 11 studies which elicit
hypothetical WTP values for public goods using both a
dichotomous choice and open-ended format, and find that
mean WTP values are consistently higher when the survey
question was posed using the dichotomous choice format.
More recent studies dedicated to this topic find similar
results (Balistreri et al., 2001). In addition, surveys using a
sample frame of visitors to a particular area would be
expected to result in higher values for threatened and
endangered species than households, due to the fact that
visitors have use as well as non-use values for threatened or
endangered species. The hypothesized sign on the species
variable coefficients are based on results from previous
studies. Response rate is expected to have a negative impact
on WTP. Boyle et al. (1994) find that response rate is an
overall indicator of the quality of the CVM survey and this
may lower WTP estimates. Finally, the year the study was
conducted was hypothesized to have an ambiguous effect in
Loomis and White (1996) and this is a topic which will be
explored further in this study.
Staying consistent with Loomis and White, a full linear and
double log model with all original variables thought to
influence willingness-to-pay was estimated. In the double
log model, the dependent variable, WTP, as well as the
independent variables, CHANGESIZE and RESPONSERATE
(non-dummy variables), are logged.

2.3.
Use of the Chow test to compare meta-analysis
regressions for U.S. studies
In order to test for differences between the original Loomis
and White (1996) meta-analysis regression and the updated
model with all U.S. studies, both old and new, the Chow test
is used. In particular, if the WTP relationship in new studies
is systematically different from that in old studies. The use
of the Chow test will show if one or more of the model's
variables has statistically changed, so the null hypothesis is
that there has been no structural change in the willingness
to pay meta-analysis regression model. Since Loomis and
White (1996) included studies conducted prior to 1995, the
alternative hypothesis is that there has been some structural change in the regression model after 1995. After sorting

the data, three models need to be run to conduct the Chow
test:
• One full model with all studies — 1983–2001
• One reduced model with studies conducted from 1983–1994
• One reduced model with studies conducted from 1995–2001.
The Chow test formula takes on the following form:

F=



RSSpooled 
 RSSold 
 RSSnew =K
fFðK;
ðRSSold + RSSnew Þ=ðNold + Nnew 
 2KÞ

Nold + Nnew 
2KÞ

where RSS is the sum of squared residuals, N is the number of
observations, K is the number of coefficients, pooled is the full
model with all studies included, old is studies conducted prior
to 1995 and new is studies conducted in or after 1995.

2.4.
New best fit model to explain WTP for threatened and
endangered species
Given the addition of new U.S. CVM studies valuing threatened
and endangered species, a new specification or best fit model to
explain the willingness-to-pay for threatened and endangered
species in the United States will be estimated for the purposes of
benefit transfer. Using the full sample of studies conducted in
the U.S., seven new variables will be added to the meta-analysis
regression to help find a best fit model that explains willingnessto-pay for the preservation of threatened, endangered and rare
species. These new variables are NEWSTUDY, LOSS, MAIL,
TELEPHONE, IN-PERSON, CHARISMATIC and NONUSE. It is not
clear what the sign on the NEWSTUDY variable is expected to be.
With some key environmental issues brought into the limelight
by various media sources in the late 1990's and environmental
issues making their way into the political mainstream by the
turn of the decade, the argument could be made that these
values would have increased over time. However, recently,
economic concerns, the war in Iraq, climate change, etc., may
have eclipsed T&E species as an area of concern. As a result, it is
unclear but important to measure how these values have
changed over time.
A second dummy variable, LOSS, was added to identify
whether the change in the size of the species population being
valued represented a gain or the avoidance of a loss. For
instance, many of the included studies in the sample valued
the avoidance of a certain percentage loss in the species
population or the avoidance of a total loss in the species
population rather than a percentage gain. This is expected to
result in a higher WTP value because it puts the species closer
to extinction, and thus conservation becomes a priority. Bulte
and Van Kooten (1998) point out the importance of looking at
the marginal valuation of a species and distinguishing
between the benefits of preventing a species from going
extinct versus the benefits of certain gains in the species
population above the minimum viable population. Bandara
and Tisdell (2005) find that the Total Economic Value for a
species is likely to be underestimated when respondents feel
that the population of the species is at a reasonably secure
level. The LOSS variable is coded with a 1 for studies valuing
the avoidance of a further loss in a species, and a 0 otherwise.
The LOSS variable is expected to have a positive effect on WTP.

1541

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

The next three variables added, MAIL, TELEPHONE and INPERSON, represent the various survey modes used in the
included studies. Each variable is coded with a 1 if the survey
mode represents that used in a particular study, 0 otherwise.
Mail surveys are expected to result in lower WTP than phone
or in-person interviews (Noonan, 2003). The IN-PERSON
variable was removed to avoid the ‘dummy variable trap.’
A variable CHARISMATIC was added to the model to test the
effect of a species’ ‘charisma’ or high profile status on the
public's valuation. There has been substantial evidence found
which indicates that a disproportionate amount of recovery
funding goes to relatively few species and it is important to see if
these policy measures are aligned with the public's Total
Economic Value of these species. Due to the results from
previous studies, this variable is hypothesized to have a positive
effect on WTP. Although identifying what constitutes ‘charismatic megafauna’ requires some subjectivity due to the varying
definitions of the term, they are generally thought to be large
vertebrates which are appealing to humans and focused on to
gain support for conservation campaigns. Studies valuing
species that were thought to represent ‘charismatic megafauna’
using classifications from Metrick and Weitzman (1996, 1998)
were coded with a 1, 0 otherwise.
One of the new CVM studies valuing T&E species by Layton
et al. (2001) uses a conjoint, or choice experiment, technique
rather than the Contingent Valuation Method to elicit the value of
various anadromous fish populations. This stated that preference
method differs from CVM in that it asks respondents to rate a set
of alternatives, each one having a number of attributes. In this
choice matrix, cost of the program is just one attribute, unlike in a
CVM study where cost of the program is the key element. Use of
these conjoint techniques to value natural resources have been
found to result in higher WTP estimates than when CVM is used
(Stevens et al., 2000). To account for these high values, a new
dummy variable for studies using this conjoint technique was
added, called CONJOINT. All studies are coded 0 except for the
Layton et al. (2001) observations, which are coded with a 1.
Finally, while some threatened and endangered species
have use values, such as viewing, hunting, and eating, as well
as nonuse values, such as existence and bequest values,
others have nonuse value only. A dummy variable was added
to test this effect, coded 1 for species having nonuse value only
and 0 for those having both use and nonuse value. This
variable is expected to have a negative effect on WTP.
The question has also been raised as to what effect the level
of endangerment facing a particular species has on WTP values.
While some of the literature (Metrick and Weitzman, 1996) has
shown that the likability of a species plays a more significant
role in WTP than the level of species endangerment, recent
studies (Tisdell et al., 2006) have found the opposite. While it
would have been beneficial to include a variable in the metaanalysis regression model accounting for the level of endangerment faced by each particular species, there was insufficient
information in the full sample of studies to test this effect.
However, given that a little over half of the sample of studies did
specify in the survey instrument the level of endangerment
facing the species being valued, a model was run on this subset
of the sample with an included dummy variable for threatened
versus endangered species to test this effect on WTP. This
variable came in insignificant at standard significance levels

and the results can be obtained from the authors. Given the
mixed findings on this topic, future original CVM studies valuing
threatened and endangered species should include the level of
threat facing the species so this effect can be further tested.
These new variables are added to the meta-analysis
regression model to test whether these other factors could
be affecting the public's valuation of threatened and endangered species. A new best fit model to explain WTP for the
preservation of threatened and endangered species, including
all studies conducted in the U.S., will be estimated to enhance
benefit transfer. With the addition of these variables, the new
model now takes on the following form:
WTP ð2006DÞ = b0 + b1 CHANGESIZE + b2 PAYFREQUENCY
+ b3 CVFORM + b4 VISITOR + = 
 b5 FISH
+ b6 MARINE + b7 BIRD + = 
 b8 OTHER

 b9 RESPONSERATE + = 
 b10 STUDYYEAR
+ b11 CONJOINT + = 
 b12 NEWSTUDY + b13 LOSS

 b14 MAIL + b15 TELEPHONE + b16 CHARISMATIC

 b17 NONUSE:

3.

Results

3.1.

Average values per household by species

Using the total sample of 31 studies with 67 willingness-to-pay
observations, the average value of various threatened and
endangered species can be found in Table 2, broken down by
studies which reported an annual versus lump sum payment.
Table 2 – Summary of economic value of threatened,
endangered and rare species ($2006)
Low High
value value
Studies reporting annual WTP
Bald eagle
Bighorn sheep
Dolphin
Gray whale
Owl
Salmon/Steelhead
Sea lion
Sea otter
Sea turtle
Seal
Silvery Minnow
Squawfish
Striped Shiner
Turkey
Washington state
anadromous fish populations
Whooping crane
Woodpecker
Studies reporting lump sum WTP
Arctic grayling
Bald eagle
Falcon
Humpback whale
Monk seal
Wolf

Average of all
studies

$21

$45

$24
$39
$10

$46
$130
$139

$11
$147

$15
$311

$39
$17
$36
$35
$65
$81
$71
$40
$19
$35
$38
$12
$8
$13
$241

$44
$13

$69
$20

$56
$16

$20
$245

$26
$350

$22

$162

$23
$297
$32
$240
$166
$61

1542

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

3.2.
Comparison of meta analysis regression of new and
old studies
Table 3 shows the results of the meta-analysis regression
models before the new variables are added to compare them
with the results from Loomis and White (1996). It is important to
point out that the new CONJOINT variable is included to account
for the considerable difference this set of observations was
found to have on WTP values. The adjusted R2 is 59% in the
linear model and 56% in the double log model. The results in
Table 3 show a good deal of consistency between the new expanded meta equation and Loomis and White (1996). With one
exception, the variables significant in Loomis and White are
significant in the new model. Likewise, insignificant variables in
Loomis and White are insignificant in the new model as well.
The only difference being that unlike in their model, CVFORM is
now significant in the linear and double log models.

3.3.
Use of the Chow test to compare meta-analysis
regressions for U.S. studies
Despite the similar signs and significance of the variables in
the two models, the magnitude of the coefficients may still be

Table 3 – Meta-analysis regressions of WTP for T&E
species — studies with conjoint dummy variable
Updated — all
U.S. studies

Variable
CONSTANT
(t-statistic)
CHANGESIZE
PAYFREQUENCY
CVFORM
VISITOR
FISH
MARINE
BIRD
OTHER
RESPONSERATE
STUDYYEAR
CONJOINT
Adj R2=
N=
F=

Loomis and White
(1996)

Linear

Double
log

Linear

Double
log

−422.65
(−0.14)
0.18⁎⁎⁎
(3.02)
55.24⁎⁎
(2.60)
34.89⁎⁎
(2.23)
74.12⁎⁎⁎
(3.61)
43.11
(1.51)
77.18⁎⁎⁎
(2.63)
65.79⁎⁎
(2.42)
40.72
(0.88)
−0.33
(−0.70)
0.19
(0.13)
217.70⁎⁎⁎
(7.69)
0.591
67
9.66⁎⁎⁎

−58.114
(− 1.54)
0.73⁎⁎⁎
(4.24)
0.327
(1.25)
0.43⁎⁎
(2.16)
1.18⁎⁎⁎
(4.69)
0.279
(0.80)
0.94⁎⁎
(2.61)
0.540
(1.63)
− 0.038
(−0.07)
− 0.341
(−1.33)
0.030
(1.55)
2.46⁎⁎⁎
(6.57)
0.556
67
8.51⁎⁎⁎

100.04
(0.57)
0.59⁎⁎⁎
(5.06)
45.51⁎⁎⁎
(2.89)
14.33
(1.12)
24.03⁎
(1.71)
24.26
(1.31)
49.87⁎⁎
(2.58)
33.41⁎
(1.85)

4.32
(1.06)
0.769⁎⁎
(2.57)
0.82⁎⁎
(2.53)
0.05
(0.18)
0.82⁎⁎
(2.73)
0.03
(0.07)
0.75⁎
(1.83)
0.57
(1.52)

0.00
(0.008)
−1.89
(−0.98)

− 0.12
(−0.38)
− 0.05
(−1.29)

0.682
38
9.82⁎⁎⁎

0.623
38
5.14⁎⁎⁎

⁎Significant at the 10% level.
⁎⁎Significant at the 5% level.
⁎⁎⁎Significant at the 1% level.

different between the two models. Thus, the next step is to
formally test whether the model has changed over time using
the Chow test. A check for collinearity among the explanatory
variables shows a few possible problematic high correlations,
given the small sample size of the model. In addition,
examining the variance-inflating factor (VIF) for each independent variable regressed on all the other explanatory
variables to check for multicollinearity also raises some
concerns of multicollinearity in the small sample of new
studies. To address this issue, and to conserve degrees of
freedom, the Loomis and White (1996) reduced model will be
used to conduct the Chow test. This includes variables that
came in significant in their meta-analysis regression, which
includes CHANGESIZE, PAYFREQUENCY, VISITOR, MARINE,
and BIRD.
In the linear model, the F statistic for the Chow test is 17.96
and in the double log model the F statistic is 16.30. The critical
F value at the 1% level is 3.12, meaning the null hypothesis
that there was no structural change in the willingness to pay
regression model between the two periods (prior to and after
1995) can be rejected at the 1% level. The results from these
regressions are available in Richardson (2008). Some of this
difference between new and old studies could be due to the
new species mix given that many of the new studies included
in the sample value fish, especially salmon. So the Chow test
was applied to the linear and double log model with the
inclusion of the variable FISH. The F statistic was still a rather
high 12.852 in the linear model and 11.896 in the double log
model so again, the null hypothesis that there was no
structural change in the model can be rejected at the 1%
level. These F statistic results are consistent with the fact that
the adjusted R2 of the pooled model is quite a bit lower than in
the individual models.
One final application of the Chow test was performed.
When the study using the conjoint technique (Layton et al.,
2001) is removed or accounted for, it appears that some of the
difference between new and old studies goes away. Running
the Chow test without this study gives some insight as to
whether this one unique study is driving the structural
change. This results in an F statistic for the Chow test of 3.72
in the linear model and 4.55 in the double log model. Since the
critical F value at the 1% level is 3.12, with the exclusion of the
conjoint technique study, the null hypothesis that there was
no structural change in both the linear and double log models
can still be rejected at the 1% level. However, the F statistic
went down quite a bit. The unique conjoint technique study
clearly has an effect and drives a lot of the difference between
the new and old studies, but there are still other factors driving
this difference.
Use of the Chow test provided evidence that there has been
some structural change in the willingness-to-pay metaanalysis regression model since 1995. However, the Chow
test does not show whether the structural difference in the
two regressions is due to differences in intercept terms, slope
coefficients, or both. Including the NEWSTUDY variable in the
pooled new and old studies model and interacting it with the
remaining explanatory variables will allow for statistical
testing of the differential effect studies conducted in or after
1995 have on the other variable's influence on WTP. When
testing for a structural difference between the two regressions

1543

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

Table 4 – Reduced linear WTP model with interaction
dummy variables for ‘new study’
Variable

Coefficient t-statistic p-value

CONSTANT
NEWSTUDY
CHANGESIZE
CHANGESIZE ⁎ NEWSTUDY
PAYFREQUENCY
PAYFREQUENCY ⁎ NEWSTUDY
VISITOR
VISITOR ⁎ NEWSTUDY
MARINE
MARINE ⁎ NEWSTUDY
BIRD
BIRD ⁎ NEWSTUDY
CONJOINT
Adj R2=
N=
F=
S.E. of regression

− 35.924
− 70.801
0.255⁎⁎⁎
1.143
84.567⁎⁎⁎
− 95.002
46.494⁎⁎
69.358
56.016⁎⁎⁎
− 18.191
57.914⁎⁎⁎
− 30.809
277.983
0.631
67
10.423
47.949

− 2.157
− 0.104
4.539
0.172
4.463
− 1.540
2.446
0.905
2.860
− 0.294
3.380
− 0.605
0.799

0.036
0.917
0.001
0.864
0.001
0.130
0.018
0.370
0.006
0.770
0.001
0.548
0.428

0.00001

⁎Significant at the 10% level.
⁎⁎Significant at the 5% level.
⁎⁎⁎Significant at the 1% level.

with the Chow test, when the conjoint technique study was
dropped the F statistic decreased, but there was still a
structural difference. This one study drives a lot of the
difference between new and old studies but there are still
other factors behind this change. Interacting the NEWSTUDY
variable with the other variables in the model will show which
ones have changed significantly. Due to the small sample size
and the relatively high number of variables, a reduced model

is focused on to look at the interaction effects. The Loomis and
White (1996) reduced model is a rational one to use because
the same variables that they found to be insignificant are
insignificant in this full model. Given that the Chow test
showed the significant effect of the conjoint technique study,
the variable CONJOINT will also be included in the model.
Table 4 shows the regression results using the reduced model
with the NEWSTUDY interaction term:
Looking at the coefficients as well as the significance of the
variables gives some insight into what has changed. The
variable PAYFREQUENCY is somewhat different from old to
new studies because when interacted with the NEWSTUDY
variable, it comes in marginally significant at the 13% level
and by itself is significant at the 1% level. The NEWSTUDY
variable itself does not show up statistically significant at
standard significance levels in the model but this could be due
to a collinearity issue affecting the results. Checking correlation coefficients, it is found that CONJOINT is indeed
correlated with the NEWSTUDY variable at 0.41. Given that
the sample only includes 67 observations, this is a relatively
high correlation and could explain why these two variables do
not come in significant in this model.
Conducting an F test on the joint significance of the
NEWSTUDY variable as well as this NEWSTUDY variable
interacted with the other explanatory variables shows that
there is indeed a significant difference between new and old
studies. This test results in an F statistic of 2.742 with degrees
of freedom (6, 54) and a p-value of 0.0213, meaning we can
reject the null hypothesis that these variables jointly equal
zero at the 5% level and conclude that there is a significant
difference between new and old studies. The results from this
test are consistent with the results from the Chow test.

Table 5 – Initial full WTP models for benefit transfer
Variable

Linear
Coefficient

t-statistic

−15.712
0.177⁎⁎
71.893⁎⁎⁎
15.652
62.080⁎⁎⁎
61.978⁎
85.994⁎⁎⁎
93.532⁎⁎⁎
77.009
−0.339
24.299
196.807⁎⁎⁎
−8.502
−27.661
−5.054
−50.434⁎⁎
0.639
67
8.775
47.479

−0.315
2.421
2.887
0.932
2.806
1.709
3.009
3.348
1.633
−0.656
1.333
6.124
−0.477
−1.345
−0.185
−2.055

CONSTANT
CHANGESIZE
PAYFREQUENCY
CVFORM
VISITOR
FISH
MARINE
BIRD
OTHER
REPONSERATE
NEWSTUDY
CONJOINT
LOSS
MAIL
CHARISMATIC
NONUSE
Adj R2=
N=
F=
S.E. of regression
⁎Significant at the 10% level.
⁎⁎Significant at the 5% level.
⁎⁎⁎Significant at the 1% level.

Double log
p-value
0.754
0.019
0.006
0.356
0.007
0.094
0.004
0.002
0.109
0.515
0.188
0.001
0.635
0.185
0.854
0.045

0.00001

Coefficient

t-statistic

1.264
0.714⁎⁎⁎
0.387
0.131
1.039⁎⁎⁎
0.880⁎⁎
0.917⁎⁎⁎
0.936⁎⁎⁎
0.552
−0.430⁎
0.720⁎⁎⁎
2.108⁎⁎⁎
−0.222
−0.633⁎⁎⁎
0.429
−0.469⁎
0.711
67
11.826
0.506

1.010
4.244
1.493
0.711
4.516
2.344
3.056
3.198
1.118
− 1.771
3.849
5.648
− 1.164
− 3.011
1.468
− 1.785

p-value
0.317
0.001
0.142
0.480
0.001
0.023
0.004
0.002
0.269
0.083
0.001
0.001
0.250
0.004
0.148
0.080

0.00001

1544

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

3.4.
New best fit model to explain WTP for threatened and
endangered species
Utilizing the results of the previous regressions as a guide, a new
meta-analysis regression to explain willingness-to-pay for
threatened and endangered species in the U.S. is estimated for
benefit transfer purposes. Ideally, this model includes all
variables mentioned earlier that could potentially explain willingness-to-pay for these species. Unfortunately, a check for
collinearity of the independent variables for this full model
shows that STUDYYEAR and NEWSTUDY have a correlation
coefficient of 0.86, which is to be expected. To account for this,
the full model was run with each of these variables separately.
The NEWSTUDY variable came in slightly more statistically
significant, so this variable was kept in the full model, but both
variables will be included separately in some of the reduced
models that follow. The MAIL survey mode variable is correlated
with TELEPHONE at about 0.82 and with IN-PERSON at about
0.52. To address this, the models were run with the MAIL survey
mode variable only as well as with the TELEPHONE and INPERSON survey mode variables only. The models including the
MAIL variable did slightly better in terms of statistical significance, so this variable is included in the full models below. In
addition, the variables FISH and CHARISMATIC have a correlation coefficient of 0.83. This may be due to the fact that fish
species are not classified as ‘charismatic megafauna’ and many
of the studies valuing species other than fish tend to focus on
those with ‘charisma.’ Kennedy (2003) characterizes a correlation coefficient as being high enough to be a serious problem at
around 0.9 or above. The models were run with both variables
kept in, as reported in Table 5.
When all variables are included in the linear meta-analysis
regression model measuring the willingness-to-pay for threatened and endangered species, the variables CHANGESIZE,
PAYFREQUENCY, VISITOR, FISH, MARINE, BIRD, CONJOINT
and NONUSE come in significant at standard significance
levels of 1, 5, or 10%. In the double log model, the variables
CHANGESIZE, VISITOR, FISH, MARINE, BIRD, RESPONSERATE,
NEWSTUDY, CONJOINT, MAIL and NONUSE come in statistically significant at standard significance levels. In order to find
the best fit model to explain willingness-to-pay for threatened
and endangered species for benefit transfer purposes, variables that that are not statistically different from zero at
standard significance levels of 1, 5, or 10% were not included.
However, variables that come in significant near the 10% level
may be included in order to stay consistent with a ‘test down’
approach and avoid omitted variable bias. After testing
various specifications, one linear model and two double log
models fit the data best and explain WTP for threatened and
endangered species quite well (the R2's are approximately 0.7).
Full results of the linear model can be found in Table 6 and full
results of the two double log models can be found in Table 7.
As can be seen in Table 6, the model coefficients are
statistically significant and the adjusted R2 is 71%. CHANGESIZE, as expected, positively impacts WTP meaning that as the
change in the size of the species population being valued
increases, so does WTP, ceteris paribus. Payment frequency
also comes in significant, with lump-sum payments about $50
higher than annual payments. Fish, marine mammals and
birds result in a higher WTP than other species such as land

mammals and reptiles. Studies using the conjoint, or choice
experiment, technique result in WTP of about $200 higher
than studies not using this method. In addition, species with
nonuse values only result in a WTP about $40 lower than those
with both use and nonuse values.
Finally, an interaction variable was included to test for a
differential effect on the slope coefficient of CHANGESIZE
arising from whether the respondent is a visitor or a nonvisiting household. This was tested by interacting the VISITOR
dummy variable with the CHANGESIZE variable. The slope
coefficient for visitor WTP is greater than the slope coefficient
of household WTP with respect to the percentage change in
the size of the species population being valued. This is due to
the fact that visitors have both use and nonuse values for an
increase in species population whereas households have
nonuse value only. Visitors WTP values for T&E species are
therefore more strongly affected by population increases than
are households WTP.
Table 7 presents the results of the double log models. With
the logged dependent variable, WTP, and the logged continuous variables, CHANGESIZE and RESPONSERATE, the two log
specification models appear to do the best job of explaining
WTP for threatened and endangered species. The models are
robust and include many of the explanatory variables which
theory and past literature finds important in determining
WTP, the only difference being that model 3 includes the
variable NEWSTUDY, whereas model 4 includes the variable
STUDYYEAR.
Both models have a high adjusted R2, with the explanatory
variables as a group explaining about 70% of the variation in
WTP. However, it is important to point out that in both
models, the variables FISH and CHARISMATIC are correlated at
around 0.83. As mentioned earlier, Kennedy (2003) notes that a
correlation coefficient of about 0.9 or higher should be

Table 6 – Reduced linear WTP model for benefit transfer
purposes
Variable

CONSTANT
CHANGESIZE
PAYFREQUENCY
FISH
MARINE
BIRD
CONJOINT
NONUSE
VISITOR ⁎
CHANGESIZE
Adj R2=
N=
F=
S.E. of regression
Sum squared
residuals

Model 1
Coefficient

t-statistic

−4.700
0.101⁎⁎
50.778⁎⁎⁎
42.641⁎⁎
47.745⁎⁎
40.280⁎⁎
198.189⁎⁎⁎
−39.069⁎⁎
0.583⁎⁎⁎

− 0.260
2.010
2.967
2.104
2.325
2.020
8.906
− 2.411
5.429

0.712
67
21.419
42.368
104,115

⁎Significant at the 10% level.
⁎⁎Significant at the 5% level.
⁎⁎⁎Significant at the 1% level.

p-value
0.796
0.049
0.004
0.040
0.024
0.048
0.001
0.019
0.001

0.00001

Sample
means
119.784
0.194
0.418
0.164
0.284
0.075
0.149
26.358

1545

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

Table 7 – Reduced double log WTP models for benefit transfer purposes
Variable

Model 2

Model 3

Coefficient

t-statistic

p-value

Coefficient

t-statistic

p-value

0.344
0.953⁎⁎⁎
1.299⁎⁎⁎
0.678⁎⁎
0.583⁎⁎
0.555⁎⁎
− 0.459**
2.620⁎⁎⁎
−0.798⁎⁎⁎
0.765⁎⁎⁎
0.816⁎⁎⁎

0.359
6.754
6.588
2.198
2.420
2.535
−2.203
8.389
−3.967
3.091
4.835

0.721
0.0001
0.0001
0.032
0.019
0.014
0.032
0.0001
0.0002
0.003
0.0001

−153.231⁎⁎⁎
0.870⁎⁎⁎
1.256⁎⁎⁎
1.020⁎⁎⁎
0.772⁎⁎⁎
0.826⁎⁎⁎
−0.603⁎⁎⁎
2.767⁎⁎⁎
−0.903⁎⁎⁎
1.024⁎⁎⁎

− 4.764
6.256
6.378
3.327
3.100
3.569
− 2.749
8.868
− 4.307
4.072

0.000
0.0001
0.0001
0.002
0.003
0.001
0.008
0.0001
0.0001
0.0001

0.078⁎⁎⁎
0.697
67
16.161
0.519
15.068

4.765

0.0001

CONSTANT
LN CHANGESIZE
VISITOR
FISH
MARINE
BIRD
LNRESPONSERATE
CONJOINT
MAIL
CHARISMATIC
NEWSTUDY
STUDYYEAR
Adj R2=
N=
F=
S.E. of regression
Sum squared residuals

0.699
67
16.347
0.517
14.940

0.00001

Sample means
4.596
0.231
0.418
0.164
0.284
3.894
0.075
0.851
0.493
0.328
1992.254

0.00001

⁎Significant at the 10% level.
⁎⁎Significant at the 5% level.
⁎⁎⁎Significant at the 1% level.

characterized as a serious problem while Gujarati (2003) points
out that a correlation coefficient of about 0.8 or higher can be
problematic. Collinearity does not seem to be degrading in this
model due to the fact that the explanatory variables have the
expected signs and are statistically significant. But to investigate the potential effect of high collinearity, an additional
test is employed. In order to test the joint significance of these
two variables in the models, a multiple linear restriction test is
used to test if the error variance from the restricted model
removing these variables is significantly bigger than the error
variance when the variables are included in the model. The
joint null hypothesis is that FISH and CHARISMATIC are equal
to zero, or not significant to the models. Testing this
restriction in model 2 in Table 7, the result is an F statistic of
4.779 with degrees of freedom of (2, 56) and a p-value of 0.0121.
Testing this same restriction in model 3, the result is an F
statistic of 8.429 with degrees of freedom of (2, 56) and a pvalue of 0.006. Thus, for both models the joint null hypothesis
that these two variables are not significant can be rejected at
the 1% level and the conclusion is that the error variance from
the restricted model is significantly bigger than the error
variance in the unrestricted model, providing justification to
include these variables. In addition, attempting to remove one
of these variables from the model seems to result in
specification bias.
Employing a logged model is useful in that the coefficients
are interpreted as percentage changes in the dependent
variable for a one percent change in the independent
variables. This facilitates an elasticity interpretation of the
coefficients and comparison to economic theory, such as
diminishing marginal returns. For instance, CHANGESIZE has
a coefficient of about 0.953 in model 3 and 0.870 in model 3,
showing that as the proposed population of the species being
valued increases, respondents' WTP increases but at a
decreasing rate, consistent with economic theory. The double
log models also show that visitors on average have a WTP

about 250% higher than households holding all else constant;
valuation of charismatic species results in a WTP about 115%–
180% higher than non-charismatic species; as the year the
study was performed increases, WTP increases by about 8%;
and fish, marine mammals, and birds result in higher WTP
than for other species such as land mammals and reptiles. In
addition, higher response rates and studies using a mail
survey mode decrease WTP, as expected. It should be noted
that unlike in the linear model, the interacted VISITOR ⁎ CHANGESIZE variable did not come in significant at standard
significance levels and therefore is not included here.
A full summary of the linear and two double log best fit
models can be found in Table 8, with signs and significance
levels of included variables for comparison. There is a
substantial degree of consistency across models in terms of
signs and significance.

Table 8 – Summary of signs and significance of the three
WTP models
Variable

CHANGESIZE
PAYFREQUENCY
VISITOR
FISH
MARINE
BIRD
RESPONSERATE
STUDYYEAR
NEWSTUDY
CONJOINT
MAIL
CHARISMATIC
NONUSE
VIISITOR ⁎ CHANGESIZE

Linear

Double log

Model 1

Model 2

Model 3

+/0.05
+/0.01

+/0.01

+/0.01

+/0.01
+/0.05
+/0.05
+/0.05
−/0.05

+/0.01
+/0.01
+/0.01
+/0.01
−/0.01
+/0.01

+/0.05
+/0.05
+/0.05

+/0.01

−/0.05
+/0.01

+/0.01
+/0.01
−/0.01
+/0.01

+/0.01
−/0.01
+/0.01

1546

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

Both a linear and double log model are beneficial for benefit
transfer purposes in that they provide a different interpretation of the independent variable's effect on WTP for threatened and endangered species. Looking at the double log
models, they have similar adjusted R2, with the only difference
being that model 2 includes the NEWSTUDY variable and
model 3 includes the STUDYYEAR variable, both of which
would be useful for benefit transfer purposes. Model 3 may be
slightly preferable since it can account for any year as opposed
to model 2 which simply provides a ‘before 1995’ and ‘in or
after 1995’ split. Looking at all three models explaining
willingness-to-pay for threatened and endangered species,
the double log model 3 would most likely be the best model for
benefit transfer purposes. This is due to the fact that managers
and users unfamiliar with economic modeling would find the
additional species related variables and fewer methodological
variables more useful.
These same meta-analysis regression models were run
without the conjoint technique study to see the effect this
would have on results. The results are almost identical except
for a reduced adjusted R2. To be as inclusive as possible, it is
important to keep this conjoint technique study in the model
and account for it.

3.5.
Benefit transfer example and calculating the percent
error for benefit transfer
Now that we have determined what we believe to be best fit
models to explain the variation in WTP for threatened and
endangered species, these meta-analysis functions can be
programmed for benefit transfer purposes. Using the reduced
double log model 3 results in the following equation:

particular species being valued. See Loomis (2000) for a
discussion of sample expansion issues to the population.
This example shows how a meta-analysis regression
function can be programmed to provide an estimate of the
willingness-to-pay for a particular threatened or endangered
species under various circumstances. Models such as these
can be used in various fields, and users do not need advanced
training in economics to use the meta equation. Given the cost
and time of conducting an original CVM study, the use of
benefit transfer methods for purposes such as estimating the
Total Economic Value of conserving a particular species will
most likely continue to have great importance in both
scientific and political arenas.
To test the accuracy of this benefit transfer meta model in
predicting WTP estimates of the in-sample threatened and
endangered species, a comparison of original study values
versus predicted values from the meta-analysis function was
conducted. To conserve space, the detailed table is not shown
but can be found in Richardson (2008).
On average, the model #3 does a reasonable job of
predicting WTP values for various threatened and endangered
species, as the average absolute percentage error is 34% for
studies reporting an annual WTP value and 45% for studies
reporting a lump sum WTP value. In many applications this
error bound may be acceptable, as compared to not having any
value, whereby it gets treated implicitly as zero WTP. The
meta-analysis benefit transfer provides an error estimate and
if decision makers decide it is unacceptable, they can perform
an original study. As more studies valuing threatened and
endangered species emerge, providing a larger sample size,
this model may become even more accurate, providing a lowcost and simple tool to predict WTP values for various species.

ln WTP ð2006DÞ = 
 153:231 + 0:870 ln CHANGESIZE
+ 1:256 VISITOR + 1:020 FISH + 0:772 MARINE
+ 0:826 
 0:603 ln RESPONSERATE
+ 2:767 CONJOINT + 1:024 CHARISMATIC

0:903 MAIL + 0:078 STUDYYEAR:
This model was chosen for benefit transfer purposes due to
its inclusion of more species related variables and less
methodological variables. By plugging in sample means for
the methodological variables and the appropriate values for
the policy relevant variables, the estimated WTP value can be
obtained. For instance, valuing a 50% gain in charismatic sea
otter populations to non-visitors in the year 2007 results in the
following equation:
ln WTP ð2006DÞ = 
 153:231 + 0:870ð3:912Þ + 1:256ð0Þ + 1:020ð0Þ
+ 0:772ð1Þ + 0:826ð0Þ 
 0:603ð3:894Þ + 2:767ð0Þ
+ 1:024ð1Þ 
 0:903ð0:851Þ + 0:078ð2007Þ
which results in a total economic value of about $88 dollars per
household. It is important to note that to improve the accuracy
of this benefit transfer estimate, the variable coefficients were
taken out to six decimal places. Aggregating this benefit per
household to a population of one million households would
result in a Total Economic Value of $88 million. However, this
aggregated amount may depend on the location of these
households. For instance, residents at a greater distance from
where the survey took place may have lower benefits for the

4.

Discussion

From the results of this study, it is found that the Total
Economic Value of species in the U.S. is sensitive to the change
in the size of the species population, the type of species being
valued, and whether visitors or households are valuing the
species. The frequency of the payment being made, the
response rate, survey mode, when the study is performed,
the ‘charisma’ of a species, and what kind of values a species
has (nonuse only or both use and nonuse) can also play a role.
It is also apparent that studies using a slightly different
valuation method, such as a conjoint, or choice experiment,
technique can have a considerable effect on a meta-analysis
with a relatively small sample size, so any variation such as
this should be accounted for.
Given the need for a consistent measure of the benefits
perceived by humans provided by threatened and endangered
species, along with the time and cost associated with conducting original CVM studies, the use of benefit transfer will
continue to play a significant role. By using the meta-analysis
regression equation itself, users can estimate the Total Economic Value of a particular species. The ease and convenience
of this technique is attractive to users across many fields of
study. However, it is important to remember that this technique
provides a rough estimate only and has an average in-sample
error of 34% to 45%, depending on whether the study reported an

E CO L O G I CA L EC O NO M IC S 6 8 (2 0 0 9) 1 53 5–1 54 8

annual or lump sum payment. In addition, both original CVM
studies and benefit transfer techniques provide economic, not
biological, benefits of a particular species. Like Loomis and
White (1996) point out, these values are based on a humancentered understanding of the particular ecological role these
species have. This understanding is in no way complete, and as
such a more cautious strategy may need to be employed when
determining listing and recovery plans for threatened and
endangered species, for example a Safe Minimum Standard (see
Ready and Bishop, 1991).
Given recent concerns that the Endangered Species Act is
being undermined, especially by limiting species listings, there
has emerged a very important argument that these species
provide considerable benefits and have great value, pointing to
the need for greater funding and more preventative measures
in their recovery. Evidence from this study shows that people's
valuation of T&E species has indeed increased over time,
providing greater support for this argument.

Acknowledgments
Partial funding for this project was provided by the Doris Duke
Charitable Foundation through the National Council for
Science and the Environment's (NCSE) Wildlife Habitat Policy
Research Program (WHPRP). Partial funding was also provided
by Agricultural Experiment Station Regional Research Project
W2133. We would like to thank Timm Kroeger at Defenders of
Wildlife for his assistance and support on the overall research
project. In addition, the paper benefited greatly from suggestions of three anonymous reviewers. Any errors or omissions
are the responsibility of the authors.

REFERENCES
Balistreri, E., McClelland, G., Poe, G., Schulze, W., 2001. Can
hypothetical questions reveal true values? A laboratory
comparison of dichotomous choice and open-ended
contingent values with auction values. Environmental and
Resource Economics 18, 275–292.
Bandara, R., Tisdell, C., 2005. Changing abundance of elephants
and willingness to pay for their conservation. Journal of
Environmental Management 76, 47–59.
Bell, K.P., Huppert, D., Johnson, R.L., 2003. Willingness to pay for
local coho salmon enhancement in coastal communities.
Marine Resource Economics 18, 15–31.
Berrens, R.P., Ganderton, P., Silva, C., 1996. Valuing the protection
of minimum instream flows in New Mexico. Journal of
Agricultural and Resource Economics 21 (2), 294–309.
Bowker, J.M., Stoll, J.R., 1988. Use of dichotomous choice
nonmarket methods to value the whooping crane resource.
American Journal of Agricultural Economics 70, 372–381.
Boyle, K., Bishop, R., 1987. Valuing wildlife in benefit–cost analysis:
a case study involving endangered species. Water Resources
Research 23, 943–950.
Boyle, K.J., Poe, G.L., Bergstrom, J.C., 1994. What do we know about
groundwater values? Preliminary implications from a meta
analysis of contingent-valuation studies. American Journal of
Agricultural Economics 76, 1055–1061.
Brown, T.C., Champ, P.A., Bishop, R.C., McCollum, D.W., 1996.
Which response format reveals the truth about donations to
a public good? Land Economics 72 (2), 152–166.

1547

Bulte, E.H., Van Kooten, G.C., 1998. Marginal valuation of
charismatic species: implications for conservation
Environmental and Resource Economics 14, 119–130.
Chambers, C., Whitehead, J., 2003. A contingent valuation
estimate of the benefits of wolves in Minnesota
Environmental and Resource Economics 26, 249–267.
Cummings, R., Ganderton, P., McGuckin, T., 1994. Substitution
effects in CVM values. American Journal of Agricultural
Economics 76, 205–214.
Duffield, J., 1991. Existence and non-consumptive values for wildlife:
application of wolf recovery in Yellowstone National Park
W-133/Western Regional Science Association Joint Session.
Measuring Non-Market and Non-Use Values. Monterey, CA.
Duffield, J., 1992. An economic analysis of wolf recovery in
Yellowstone: park visitor attitudes and values. In: Varley, J.,
Brewster, W. (Eds.), Wolves for Yellowstone? National Park
Service, Yellowstone National Park.
Duffield, J., Patterson, D., 1992. Field testing existence values:
comparison of hypothetical and cash transaction values. In:
Rettig, B. (Ed.), Benefits and costs in natural resource planning,
5th Report. W-133 Western Regional Research Publication.
Compiler, Dept. of Agricultural and Resource Economics.
Oregon State University, Corvallis, OR.
Duffield, J., Patterson, D., Neher, C., 1993. Wolves and people in
Yellowstone: a case study in the new resource economics. Report
to Liz Claiborne and Art Ortenberg Foundation. Department of
Economics, University of Montana, Missoula, MT.
Defenders of Wildlife, 2004. Economic impact assessment of
designating critical habitat for the lynx (Lynx Canadensis).
Prepared for The Geraldine R. Dodge Foundation.
Giraud, K., Loomis, J., Johnson, R., 1999. Internal and external
scope in willingness-to-pay estimates for threatened and
endangered wildlife. Journal of Environmental Management
56, 221–229.
Giraud, K., Turcin, B., Loomis, J., Cooper, J., 2002. Economic benefit
of the protection program for the stellar sea lion. Marine Policy
26, 451–458.
Gujarati, D.N., 2003. Basic Econometrics (4th ed.). New York:
McGraw Hill/Irwin.
Hageman, R., 1985. Valuing marine mammal populations: benefit
valuations in a multi-species ecosystem. Administrative
Report LJ-85-22. Southwest Fisheries Center, National Marine
Fisheries Service, La Jolla, CA.
Hagen, D., Vincent, J., Welle, P., 1992. Benefits of preserving
old-growth forests and the spotted owl. Contemporary Policy
Issues 10, 13–25.
Kennedy, P., 2003. A Guide to Econometrics, 5th ed. The MIT Press,
Massachusetts.
King, D., Flynn, D., Shaw, W., 1988. Total and existence values of a
herd of desert bighorn sheep. Benefits and Costs in Natural
Resource Planning, Interim Report. . Western Regional Research
Publication W-133. University of California, Davis, CA.
Kotchen, M., Reiling, S., 2000. Environmental attitudes,
motivations, and contingent valuation of nonuse values: a case
study involving endangered species. Ecological Economics 32,
93–107.
Layton, D., Brown, G., Plummer, M., 2001. Valuing Multiple
Programs to Improve Fish Populations. Washington State
Department of Ecology.
Loomis, J.B., 1996. Measuring the economic benefits of removing
dams and restoring the Elwha river: results of a contingent
valuation survey. Water Resources Research 32 (2), 441–447.
Loomis, J.B., 2000. Vertically summing public good demand curves:
an empirical comparison of economic versus political
jurisdictions. Land Economics 76 (2), 312–321.
Loomis, J.B., Larson, D., 1994. Total economic values of increasing
gray whale populations: results from a contingent valuation
survey of visitors and households. Marine Resource Economics
9, 275–286.

1548

E C O L O G IC A L E C O N O M IC S 6 8 ( 2 0 09 ) 15 3 5 –1 54 8

Loomis, J.B., White, D.S., 1996. Economic benefits of rare and
endangered species: summary and meta-analysis. Ecological
Economics 18, 197–206.
Loomis, J.B., Ekstrand, E., 1997. Economic benefits of
critical habitat for the Mexican spotted owl: a scope test
using a multiple-bounded contingent valuation survey.
Journal of Agricultural and Resource Economics 22 (2),
356–366.
Metrick, A., Weitzman, M.L., 1996. Patterns of behavior in
endangered species preservation. Land Economics 72 (1), 1–16.
Metrick, A., Weitzman, M.L., 1998. Conflicts and choices in
biodiversity preservation. Journal of Economic Perspectives 12 (3),
21–34.
Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public
Goods: The Contingent Valuation Method. Resources for the
Future, Washington D.C.
National Wildlife Federation, 2004. “Bush administration
suppresses facts to weaken endangered species act: NWF
Report Finds Critical Habitat Reductions Based on Distorted Economic
Data.” bhttp://www.nwf.org/news/story.cfm?
pageId=57139C4F-DED5-3CB0-3F478525211A303CN.
Noonan, D.S., 2003. Contingent valuation and cultural resources: a
meta-analytic review of the Literature. Journal of Cultural
Economics 27, 159–176.
Olsen, D., Richards, J., Scott, D., 1991. Existence and sport values
for doubling the size of Columbia river basin salmon and
steelhead runs. Rivers 2, 44–56.
Ready, R.C., Bishop, R.C., 1991. Endangered species and the safe
minimum standard. American Journal of Agricultural
Economics 73 (2), 309.
Reaves, D.W., Kramer, R.A., Holmes, T.P., 1994. Valuing the
endangered red cockaded woodpecker and its habitat: a
comparison of contingent valuation elicitation techniques and
a test for embedding. AAEA meetings paper.
Richardson, L., 2008. The total economic value of threatened and
endangered species: an updated meta-analysis and extension
to international studies. In: MS Thesis, Department of
Agricultural and Resource Economics, Colorado State
University.

Rosenberger, R.S., Loomis, J.B., 2003. Benefit transfer. In: Champ, P.A.,
Boyle, K.J., Brown, T.C. (Eds.), A Primer on Nonmarket Valuation.
Kluwer Academic Publishers, Boston, pp. 445–482.
Rubin, J., Helfand, G., Loomis, J., 1991. A benefit–cost analysis of the
northern spotted owl. Journal of Forestry 89 (12), 25–30.
Samples, K., Hollyer, J., 1989. Contingent valuation of wildlife
resources in the presence of substitutes and complements. In:
Johnson, R., Johnson, G. (Eds.), Economic Valuation of Natural
Resources: Issues, Theory and Application. Westview Press,
Boulder, CO.
Stanley, D.L., 2005. Local perception of public goods: recent
assessments of willingness-to-pay for endangered species.
Contemporary Economic Policy 2, 165–179.
Stevens, T., Echeverria, J., Glass, R., Hager, T., More, T., 1991.
Measuring the existence value of wildlife: what do CVM
estimates really show? Land Economics 67, 390–400.
Stevens, T.H., Belkner, R., Dennis, D., Kittredge, D., Willis, C., 2000.
Comparison of contingent valuation and conjoint analysis in
ecosystem management. Ecological Economics 32, 63–74.
Swanson, C., 1993. Economics of non-game management: bald
eagles on the Skagit river bald eagle natural area, Washington.
In: Ph.D. Dissertation, Department of Agricultural Economics,
Ohio State University.
Tisdell, C., Nantha, H.S., Wilson, C., 2006. Endangerment and
likeability of wildlife species: how important are they for
payments proposed for conservation? Ecological Economics 60,
627–633.
U.S. Department of the Interior, Fish and Wildlife Service, 1994.
The reintroduction of gray wolves to Yellowstone National
Park and central Idaho. Final Environmental Impact
Statement. Helena, MT, pp. 4.21–4.27.
Whitehead, J., 1991. Economic values of threatened and
endangered wildlife: a case study of coastal nongame wildlife.
Transactions of the 57th North American Wildlife and Natural
Resources Conference. Wildlife Management Institute,
Washington, DC.
Whitehead, J., 1992. Ex ante willingness-to-pay with supply and
demand uncertainty: implications for valuing a sea turtle
protection programme. Applied Economics 24, 981–988.