METHODS
Comparative analysis of non-market valuation techniques for
the Eduardo Avaroa Reserve, Bolivia
Lindsey Ellingson⁎, Andrew Seidl
Department of Agricultural and Resource Economics at Colorado State University, Fort Collins, Colorado 80523, United States
A R T I C L E I N F O A B S T R A C T
Article history:
Received 22 July 2005
Received in revised form 11 July 2006
Accepted 25 July 2006
Available online 18 September 2006
Ecological tourism has held up as a potentially important development alternative in
relatively undeveloped regions and countries. However, tourism is a highly vulnerable
activity where tourists display voluntary and discretionary behavior. While tourism can
serve as an income-generating activity for the region, uncontrolled tourism that is
developed without consideration of the environmental impact can adversely affect the
natural areas. Determining what to charge tourists to view a natural area is difficult since
demand elasticities are not readily available from observable transactions. In order to better
gauge the value tourists place on the reserve and their sensitivity to changes in the costs of
visiting Eduardo Avaroa, non-market valuation provides several valuable tools. Here, we
test the convergent validity between two popular non-market valuation alternatives: the
contingent valuation method and the contingent behavior method. Tourists to the Eduardo
Avaroa Reserve in Bolivia were asked their willingness to pay (in visitor-days or in dollars)
due to an improvement in tourism services at the reserve. Our models show that the
traditionally important variables predict willingness to pay for improved services for the
contingent behavior, contingent valuation and the pooled model. However, we find
statistically and empirically meaningful distinctions in the estimates derived from the
two methods. Comparisons of the implications of adopting the entrance fee
recommendations from the two non-market valuation techniques are made based on the
revenues predicted to accrue to Eduardo Avaroa Reserve if park management techniques
were improved. This research highlights the challenges in using non-market valuation
techniques for policy formation, particularly in a developing country setting.
© 2006 Elsevier B.V. All rights reserved.
Keywords:
Contingent valuation
Contingent behavior
Bolivia
1. Introduction
Tourism is among the world's largest and fastest growing
industries. Ecological tourism has been held up as a potentially
important development alternative in relatively undeveloped
regions and countries (Wall, 1997). In the case of developing
countries, a rapidly growing tourism industry has proved to be
an increasingly important source of foreign exchange inflows
(Wunder, 2000). In fact, the demand for ecotourism flows mainly
from the developed countries to the developing countries, as the
willingness to pay for non-consumptive use and nonuse values
of natural resources among residents of developing countries is
often relatively lower than in developed nations, even after
controlling for income differences (Herath, 2002).
However, tourism is a highly vulnerable activity where
tourists display voluntary and discretionary behavior. Although
E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
⁎ Corresponding author. Fax: +1 970 491 2067.
E-mail address: Lindsey.Ellingson@Colostate.edu (L. Ellingson).
0921-8009/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2006.07.014
available at www.sciencedirect.com
www.elsevier.com/locate/ecolecon
entry fees are among the few policy tools available to capture
directly tourism expenditures at the reserve gate, determining
what to charge tourists to experience a natural area is challenging
since there are no readily available market signals (Pigram
and Wahab, 1997). Nonetheless, Schultz et al. (1998) find that the
use of differential entrance fees for national parks and protected
areas are useful to generate revenues to recover costs, to ensure
quality goods and services, to reduce congestion in overcrowded
parks, and promoting visitation to less crowded parks.
In order to better gauge the value tourists place on the
reserve and their sensitivity to changes in the costs of visiting
Eduardo Avaroa, non-market valuation provides several
valuable tools. Here, we explore the policy implications of
two popular non-market valuation alternatives; the contingent
valuation method and the contingent behavior method.
We compare the implications of adopting the entrance fee
recommendations of these techniques on the revenues
predicted to accrue to Eduardo Avaroa Reserve. The objective
is to test whether there is a statistical difference, or even
noticeable distinction in park entrance fee policy recommendations,
between the two non-market valuation elicitation
methods to facilitate the use of tourism as an engine of
economic development. Our research highlights the challenges
in using non-market valuation techniques for policy
formation, particularly in a developing country setting.
2. Study site: Eduardo Avaroa Reserve, Bolivia
Nestled in the southwest corner of Bolivia, the Eduardo Avaroa
Reserve or Reserva Eduardo Avaroa (REA) is a Priority I ecoregion
due to its unique ecological and archeological features. The
reserve has an arid landscape, which consists of volcanoes,
geysers, salt marshes and mountain lakes. Eduardo Avaroa
Reserve spans an area of 1.8 million acres and has more than 50
archeological sites making it rich in cultural and archeological
heritage. REA houses at least 190 species of plants and trees and
80 different species of birds including three of the world's six
flamingo species. It is one of the most visited reserves in Bolivia
and also one of the most economically depressed regions in the
country with threats to the region consisting of mining, poor
farming practices and unregulated tourism. While tourism can
serve as an income-generating activity for the region, uncontrolled
tourism that is developed without consideration of the
environmental and cultural impacts can adversely affect the
natural areas. In order to avoid the potential negative impacts, a
proposed fee increase is examined so to improve the park's
reserve management (The Nature Conservancy).
3. Estimating the demand for ecotourism
services
This study employs two methods to estimate the willingness to
pay (WTP) for entrance to the Eduardo Avaroa Reserve and
compares results from the contingent valuation method (CVM)
and the contingent behavior (CB) estimates of willingness to
pay. The convergent validity between the two valuation
methods is examined. Convergent validity investigates consistency
of contingent valuation estimates with estimates provided
by another non-market valuation method, in this case
contingent behavior (Champ et al., 2003). One issue that arises
with convergent validity is strategic bias. Strategic bias exists
when the respondent has the incentive to misrepresent their
true preferences in order to achieve a more desired outcome.
However, Boardman et al. (2005) state that when potential users
are asked to choose between an existing and an alternative
private good, the respondents have no incentive to misstate
their true preferences over their choices. Since entrance into
Eduardo Avaroa Reserve is rival and excludable it can be viewed
as a private good. Further, the survey respondents were asked if
they would visit the park again if it included additional services
(i.e. visitor's center and improved park facilities) only available
to those who paid the increased fee level. Therefore, strategic
bias is not an issue in valuing the willingness to pay for entrance
to Eduardo Avaroa Reserve.
3.1. Contingent valuation method
The contingent valuation method is facilitated by positing
behavioral changes due to different fee levels contingent on the
provision of a new visitor center, improved park services and
facilities and local naturalist tour guides. It is a direct interview
(i.e. survey) approach that can be used to provide acceptable
measures of the economic value of recreation opportunities as
well as the preservation of natural resources (Loomis and
Walsh, 1997). The object of CVM is to measure consumer surplus
for the environmental attributes of a tourism product at a
specific destination (i.e. National Park or Reserve). To remain
consistent with consumer choice theory, the elicitation of WTP
needs to propose hypothetical or contingent changes and
improvements to the parks or other tourism product being
valued (Shultz et al., 1998).
Herath (2002) argues that in cases where revealed preference
methods (e.g., travel cost method) are not appropriate,
stated preference methods, such as CVM, are the most useful
analytical approach. There are two advantages of using the
contingent valuation method. First, CVM is able to assess not
only an individual's WTP of the present conditions of a park,
but it also values their WTP with hypothetical changes to the
park. Secondly, CVM is able to value trips with multidestinations
by asking hypothetical questions for each
specified destination (Lee and Han, 2002). Since the Eduardo
Avaroa Reserve is not a tourist's sole destination in Bolivia,
CVM is a good valuation measure in this case.
3.2. Contingent behavior method
While the contingent valuation method elicits a value
statement for a non-market good, contingent behavior is
used to estimate changes in behavior for a non-market good.
The contingent behavior method asks respondents about their
intended visitation behavior given a proposed change to the
site (Chase et al., 1998). Specifically, the contingent behavior
method is used to measure the current number of days the
respondent plans to spend at REA with the current fee level
and the number of days they would plan to spend at REA with
hypothetical changes to the park under a new fee level that is
randomly chosen among respondents. The hypothetical
changes proposed under the contingent behavior scenario
518 E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
are identical to those proposed under the contingent valuation
scenario.
Chase et al. (1998) use the contingent behavior method to
generate experimental data and assess the effects of differential
pricing of entrance fees on the demand of park
visitation in Costa Rica. Specifically, park visitation demand
functions and price and income elasticities are estimated for
three parks in Costa Rica. Further, applications of differential
pricing to the parks are discussed. It is concluded that WTP
estimates for park entrance fees act as a “reference point”
from which judgments are based (Chase et al., 1998). Grijalva
et al. (2002) tested the validity of data obtained via the
contingent behavior method, while valuing the demand for
rock climbing at Hueco Tanks Texas State Park (USA). Their
results indicate that climbers do not appear to overstate their
behavioral changes when presented with a hypothetical
situation (Grijalva et al., 2002).
In the case of REA, there is no differential pricing strategy
under consideration in determining optimal entrance fee
levels. However, the contingent behavior method is also
applicable to non-differential pricing of entrance fees. Here,
the contingent behavior method is used to measure the
current number of days the respondent plans to spend at
REA with the current fee level and the number of days they
would plan to spend at REA with hypothetical changes to the
park under a new fee level chosen from a range of feasible
changes that is randomly exposed to respondents.
3.3. Contingent valuation and contingent behavior combined
The contingent valuation method is not a perfect substitute
for having revealed preference information, and will not
provide all the answers for valuing ecotourism, but it is
among the very few methods available to measure economic
value where there is no market information. However, Herath
(2002) contends that diversity in CVM applications is required
to suit different circumstances, particularly in an ecotourism
context. Adamowicz et al. (1994) argue that the strength of the
contingent valuation method as a natural resource valuation
tool is improved if it is implemented in combination with
another non-market valuation technique. Here, we explore
this contention by including both contingent valuation and
contingent behavior estimates.
Relative to contingent behavior, the contingent valuation
method might be a more common method for valuing nonmarket
goods such as National Parks or Reserves; however, it
has one limitation that the contingent behavior method can
overcome. In using standard survey techniques for CVM, it is
not possible to collect data necessary to estimate an unrestricted
system of demand equations that can be used in
designing effective pricing policies. In order to accomplish
this, the contingent behavior method needs to be applied in
conjunction with contingent valuation (Chase et al., 1998).
The contingent behavior method is viewed as controversial
due to its intrinsically hypothetical nature; however, since
there is only minimal focus of nonuse values using CB, the
limitations of the contingent valuation method might not hold
for data obtained through its combination with the contingent
behavior method (Grijalva et al., 2002). Grijalva et al. (2002) find
that methods of augmenting revealed preference data with
stated preference data show promise as a tool for estimating
demand for a choice-based sample.
To date, no literature combining the contingent valuation
method and the contingent behavior used to value a National
Park or Reserve in a developing country has been identified.
This is among the contributions of this study to the body of
natural resource economic valuation literature. Our study of
entry fee policy of the REA compares and contrasts CV data
with CB data and then pools them in helping us to understand
the use of these methods within a developing country setting.
4. Methodology
Deriving an accurate non-consumptive use value is dependent
on the survey method and the direct face-to-face interview is the
most commonly used approach at recreation sites. Similar to our
approach in the REA, Lee and Han (2002) used the face-to-face
interview method in valuing multi-destination trips and,
logically, use entrance fees as their payment vehicle. Lee and
Han (2002) conclude that an increase in entrance fees is justified
in order to maintainthe quality of the environmentbeing valued.
4.1. Data
Visitation to the Eduardo Avaroa Reserve is valued through
cross-sectional data obtained from a combination of in-person
interviews and self-administered surveys collected at the
reserve during April 2003. The survey was administered in
English, French, and Spanish to accommodate REA park
visitors from abroad. A randomized design was employed to
distribute the referendum format elicitation question in the
contingent valuation and contingent behavior treatments.
4.2. Question format
Creating a market for the non-market good being valued
requires an appropriate question format. Among the several
types of question formats, the dichotomous choice format
follows the recommendation of the US Department of Commerce's
National Oceanic and Atmospheric Administration
(NOAA) report on CVM (Leon, 1997). In a dichotomous choice
format, the respondent is asked whether they would accept or
reject a suggested price, which varies randomly among
respondents, under a hypothetical situation. Loomis and
Walsh (1997) state that it is easier for respondents to make
their decisions in using this format because discrete choices are
similar to choices made in market transactions. Lee and Han
(2002) use the dichotomous choice question format to value five
different National Parks in Korea. Leon's (1997) study in the
Canary Islands found that the single-bounded dichotomous
choice method produced higher values than the doublebounded
and open–ended methods, implying the singlebounded
method might overestimate consumer surplus, while
the double-bounded method obtained more consistent results.
Dichotomous choice produces more valid responses than open–
ended and iterative bidding question formats (Loomis and
Walsh, 1997; Leon, 1997; Fredman and Emmelin, 2001).
Our survey was distributed to current visitors of the
Eduardo Avaroa Reserve by intercepting them at the entrance
519E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
and exit of the park. The contingent scenario provided to the
respondents was the same for both the contingent behavior
and the contingent valuation surveys. The contingent scenario
was defined as the following:
In Eduardo Avaroa, the Bolivian Park Service charges visitors
US $4/person to pay for protecting the reserve (the fee is only
paid once, regardless of how many days you stay). The Park
Service may decide to charge a higher fee — with money
being used to improve visitor services and facilities. For
example, the income would be used to build and maintain a
visitor center and to train local naturalist guides.
This fee would increase the cost of your visit, as operators
would add it to your tour price. We would like to know how
this would affect your trip. Please assume that the fee
changes only at Eduardo Avaroa — not at other parks.
The difference between the two valuation methods lies
with the willingness to pay question. The contingent valuation
question was a dichotomous choice question and was
stated as the following:
If the trip price had been US $X higher than what you paid,
would you still have come to the reserve?
The respondents could either state yes, they would still
come to the reserve (coded as 1) or no they would not have
come to the reserve (coded as 0).
For the contingent behavior question, the respondents
were given three options to respond to the question, which
was stated as the following:
If the price had been US $X higher than what you paid,
what would you have done?
The three options the respondents could choose from were
indicated as follows:
1. I would have kept the same itinerary of visiting parks.
2. I would have made a different itinerary of visiting parks.
3. I would not have visited any parks.
If the respondents stated that they would have kept the same
itinerary of visiting parks it was coded as a 1. If the respondents
stated that they wouldhave made a differentitinerary, they were
further asked the days they would spend at the Eduardo Avaroa
Reserve along with other parks in Bolivia and in other countries.
If the respondent stated that they would spend 0 days at the
Eduardo Avaroa Reserve after the fee increase, their response
was coded as a 0. If the respondent stated that they would spend
at least 1 day at the reserve after the fee increase, their response
was coded as a 1. Lastly, if the respondents chose the option of
not visiting any park, their response was coded as a 0.
4.3. Model
Using non-market valuation and a dichotomous choice
question format, the binomial logit regression is an appropriate
statistical analysis tool for this study. The binomial logit is
an estimation technique for equations with a dichotomous
qualitative dependent variable. The dependent variable is the
log of the odds that the choice in question will be made
(Studenmund, 1997). Under both scenarios, the dependent
variable is 1 if the respondent would continue to visit the
Eduardo Avaroa Reserve and 0 if the respondent would not
continue to visit the Eduardo Avaroa Reserve, given a site
improvement.
4.4. Hypotheses
The objective of this study is to test whether there is a
statistical difference between two non-market valuation
elicitation methods and a pooled model incorporating both
techniques. In addition, the willingness to pay for visiting the
park is compared between the contingent valuation and the
contingent behavior methods. The following logit regression
was used for these analyses:
Vote¼b0−b1 Bid Amtþb2 Europeþb3 Incomeþb4 TourPackage
The variables used for the analysis were fee amount (Bid
Amt), European country of origin (Europe), annual household
income (Income) and tour package inclusion (Tour Package).
The entrance fee bid amounts took on the following values: $5,
$10, $20, $30, $50, and $75. The bid amounts were evenly
distributed among respondents and between the two valuation
techniques. The European country of origin variable is a dummy
variable based on the country where the respondent lives and
was coded as: 1=European Country and 0=Non-European
Country. The countries classified as European countries were:
United Kingdom, France, Germany, Netherlands, Switzerland,
and Ireland. The countries classified as non-European countries
were: Israel, Australia, United States, and Canada. There was
also a category of “Other Countries” that fell under the nonEuropean
category. The income variable was the respondents'
median annual household income. Further, any questions
pertaining to specification of a monetary level (i.e. annual
income) allowed for answers in foreign currencies. Currency
amounts were normalized to United States dollars. The tour
package inclusion variable is a dummy variable on whether the
respondent traveled as part of tour package (coded as 1) or
whether the respondent traveled independently (coded as 0)
during his or her time in Bolivia (Table 1).
The gender variable was coded as 1 for female respondents
and 0 for male respondents. The age variable was divided into
the following categories: 1=Under 18 years, 2=18–29 years,
3=30–39 years, 4=40–49 years, 5=50–59 years and 6=60 years or
over. The education level of the respondents was defined as:
1=Primary school, 2=High school (diploma), 3=Undergraduate
college or university (e.g. Bachelor's), and 4=Graduate school
(e.g. Master's or Ph.D.). The sample size between the two
scenarios was 187 and 194 for the contingent behavior and the
contingent valuation scenarios, respectively.
In order to compare the difference between contingent
valuation and contingent behavior, descriptive statistics and
demand elasticities were calculated. Econometric tests were
also conducted to evaluate the statistical difference between the
two models as well as the statistical difference between
willingness to pay under the contingent valuation scenario
520 E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
versus the contingent behavior scenario. In order to evaluate the
statistical difference between the two models, the following null
hypothesis was tested:
H0 : Contingent Valuation ¼ Contingent Behavior
The vector of coefficients for each valuation method is
denoted as β. In order to test for the coefficient equality, a
likelihood ratio (LR) test needs to be performed. The LR test
involves taking the difference of the restricted log likelihood
function (log likelihood of the pooled model) from the
unrestricted log likelihood function (summation of log likelihood
for each ethnicity model) and multiplying the difference
by negative two. The result is the calculated chi-square and if
it is greater than the critical chi-square statistic then the
coefficients across the two methods are statistically different
from each other (Gujarati, 2003). The three models used for
this analysis were the contingent behavior model only,
contingent valuation model only and a pooled model of all
the data. In addition, a pairwise t-test was conducted to test
for significant difference between the contingent behavior and
contingent valuation models. The t-test is used to test for
significance across coefficients between the two models
(Stephenson). The equation for the t-test is as follows:
SEðbCVM−bCBÞ ¼ SPM½ð1=nCVMÞ þ ð1=nCBފ
It is, essentially, the standard error between the two
coefficients, where n is the number of observations for each
sample and β is the coefficient value for the willingness to pay
for each sample. The pooled standard error, denoted as SP, is
calculated as follows:
SP ¼ Mf½ðnCVM−1Þ4S2
CVM þ ðnCB−1Þ4S2
CBŠ=ðnCVM þ nCB−2Þg
Where, S2
is the variance of the willingness to pay
coefficient for the contingent valuation logit regression and
the contingent behavior logit regression, respectively.
It is hypothesized that the willingness to pay for the
Eduardo Avaroa Reserve will vary across valuation methods.
In order to test the difference in willingness to pay across
elicitation methods, the following hypothesis is tested:
H0 : WTPContingent Valuation ¼ WTPContingent Behavior
To test for statistical difference between willingness to pay
across non-market valuation methods, the confidence intervals
for the mean willingness to pay values for each scenario
were derived.
Further, the differences in the price elasticities and the
differences in the expected value of revenue among the two
elicitation methods are compared. The hypothesis is that the
price elasticities will not be equal under each valuation
scenario. The following null hypothesis is examined:
H0 : Price ElasticityContingent Valuation
¼ Price ElasticityContingent Behavior
Table 1 – Explanation of variables
Variable Explanation
Age Age of the respondent. Coded as: 1=Under
18 years, 2=18–29 years, 3=30–39 years,
4=40–49 years, 5=50–59 years, and
6=60 years or over
Bid amount Bid amount, with the following values: $5,
$10, $20, $30, $50, and $75
Days in Bolivia The number of days the respondent will
spend in Bolivia on his/her current trip
Days in Eduardo
Avaroa Reserve (REA)
The number of days the respondent will
spend in the Eduardo Avaroa Reserve on
his/her current trip
Days in Latin America The number of days the respondent will
spend in Latin America on his/her current
trip
Days in other Bolivian
parks
The number of days the respondent will
spend in other nature reserves or parks in
Bolivia on his/her current trip
Days in other
countries' parks
The number of days the respondent will
spend in other nature reserves or parks in
other countries on his/her current trip
Education Education level of the respondent. Coded
as: 1=Primary school, 2=High school
(diploma), 3=Undergraduate college/
university (e.g. Bachelors), and 4=Graduate
(e.g. Masters or Ph.D.)
European Dummy variable: 1 if the respondent lives
in a European country, 0, if the respondent
lives in a non-European country
Gender Dummy variable: 1 if the respondent is
female, 0 if the respondent is male
Income Median household income of the
respondent (converted into US dollars)
Total days on trip The number of days the respondent will
spend in total on his/her current trip
Total amount spent in
Bolivia
The amount of money the respondent will
spend in total for his/her time in Bolivia,
including airfare, accommodations, food,
souvenirs and other expenses
Tour package Dummy variable: 1 if the respondent is
traveling as part of a tour package during
his/her time in Bolivia, 0 if the respondent
is traveling independently during his/her
time in Bolivia
Table 2 – Descriptive statistics
Mean values CB CV Pooled
Gender 0.47 0.46 0.47
Age 2.44 2.38 2.41
Education 2.92 2.86 2.89
Income $36,684.49 $37,886.60 $37,296.59
Travel Statistics
Days in Latin America 126.3 111.2 118.6
Days in Bolivia 26.5 26.2 26.4
Days in REA 3.0 3.0 3.0
Days in other Bolivian parks 3.2 3.0 3.1
Days in other countries' parks 11.0 12.3 11.7
Total days on trip 174.1 177.2 175.7
Tour package 0.56 0.58 0.57
Avg. amt. spent in REA $83.52 $72.19 $77.70
Avg. amt. spent in other
Bolivian parks
$89.23 $72.07 $80.26
Total amount spent in Bolivia $730.43 $637.09 $682.66
Number of observations 187 194 381
521E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
In addition, it is hypothesized that the total expected value
of revenue impact for each fee level will be different under
each valuation scenario. Therefore, the null hypothesis to be
evaluated is:
H0 : Expected RevenueContingent Valuation
¼ Expected RevenueContingent Behavior
The method used to calculate the price elasticities and total
expected value of revenue impacts will be described in detail
in the next section.
4.5. Elasticity and expenditure calculations
A logit regression model has a dependent variable that is the log
of the odds ratio instead of the actual variable; therefore, it is
difficult to calculate price elasticity. However, through some
mathematical manipulation, elasticities canbe derived. By using
the forecasting command in EViews, an econometric software
package, one can expand the sample. The reason for expanding
the sample is to forecast the willingness to pay values, given our
estimates, for each bid amount level. From this one can derive
the elasticity estimates. The sample size was expanded by six
observations to account for each bid amount level. Next each bid
amount was entered into the forecast sample. With the
expanded sample size, the logit regression was run with the
dependent variable as is in our previous model and the only
independent variable was the bid amount. Once the regression
was run, the forecasted dependent variable for each given bid
amount was obtained from the dependent variable data series.
Since all of the respondents were currently visiting the park
at the current fee level of $4.00, it was assumed that all the
respondents would attend the park at this fee level. It can then
be assumed that the mean dependent variable would be equal
to one at the $4.00. With this assumption and the forecasted
dependent variables for each bid amount, the elasticity can be
calculated. The elasticity was calculated by dividing the
percentage change in the dependent variable by the percentage
change in the bid amount. The elasticity was calculated for
each given bid amount: $5, $10, $20, $30, $50 and $75.
With the elasticity calculations, the average amount spent in
the Eduardo Avaroa Reserve obtained from the survey and the
annual visitation obtained directly from REA, the expected value
of revenue can be calculated. Since the elasticities were
calculated for each proposed new bid amount, then the annual
expected expenditure levels for REA based on each given fee level
can also be calculated. The average amount spent in REA per
visitor was derived by multiplying the number of days spent at
the park by the respondents' average daily expenditure in Bolivia.
The total visitation for REA in 2000 was 26,150 visitors, which will
be used in the expected expenditure impact calculation.
Before the expected value of revenue can be derived, the
expected probability of the number of visitors to REA needs to
be determined for each bid amount. The expected probability
of number of visitors was calculated by multiplying the
number of visitors in 2000 by one minus the elasticity
associated with each bid amount. This calculation displays
how responsive park visitation is relative to a change in the
entrance fee level. The expected value of revenue can then be
determined by multiplying the expected probability number of
visitors by the daily average amount spent in REA for each fee
level.
Table 3 – Respondents' countries of origin
Country
count
CB
count
CB % CV
count
CV % Pool
count
Pool
%
United Kingdom 65 34.6% 72 36.7% 137 35.7%
France 19 10.1% 21 10.7% 40 10.4%
Germany 14 7.4% 17 8.7% 31 8.1%
Netherlands 12 6.4% 14 7.1% 26 6.8%
Switzerland 12 6.4% 6 3.1% 18 4.7%
Ireland 4 2.1% 4 2.0% 8 2.1%
Total European
countries
126 67.0% 134 68.4% 260 67.7%
Israel 17 9.0% 26 13.3% 43 11.2%
United States 14 7.4% 6 3.1% 20 5.2%
Australia 6 3.2% 7 3.6% 13 3.4%
Canada 8 4.3% 1 0.5% 9 2.3%
Other 17 9.0% 22 11.2% 39 10.2%
Total nonEuropean
countries
62 33.0% 62 31.6% 124 32.3%
Total 188 100.0% 196 100.0% 384 100.0%
Table 4 – Contingent behavior, contingent valuation and pooled logit regressions
Dependent variable: whether (coded as 1) or not (coded as 0) the respondent would visit REA with a fee increase
Contingent behavior Contingent valuation Pooled model
Variable Coeff. Std. error z-stat Prob. Coeff. Std. error z-stat Prob. Coeff. Std. error z-stat Prob.
Constant 0.516 0.588 0.877 0.381 0.949 0.466 2.037 0.042 0.770 0.342 2.254 0.024
Bid amt. −0.034 0.008 −4.329 0.000 −0.069 0.010 −6.626 0.000 −0.045 0.006 −8.023 0.000
European 0.681 0.399 1.707 0.088 1.534 0.430 3.568 0.000 0.918 0.268 3.429 0.001
Income 0.000 0.000 2.491 0.013 0.000 0.000 0.812 0.417 0.000 0.000 2.637 0.008
Tour package 0.926 0.395 2.345 0.019 0.385 0.384 1.002 0.316 0.592 0.259 2.288 0.022
Mean dep. var. 0.7647 0.5825 0.6719
S.E. of regression 0.3846 0.3835 0.4026
Log likelihood −83.1764 −86.5862 −186.7190
Restricted log
likelihood
−102.0262 −131.8193 −241.1019
LR statistic (3 df) 37.6995 90.4663 108.7658
Probability. (LR stat.) 0.0000 0.0000 0.0000
Mc Fadden R^2 0.1848 0.3431 0.2256
Number of observations 187 194 381
522 E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
5. Results
Pairwise t-tests were performed comparing the mean values
of the respondents' demographics across both valuation
samples. The demographics under both scenarios were not
significantly different from each other with the exception of
the number of days the respondents spend in Latin America,
which was significant at the 90% level. The mean values of the
descriptive statistics for the contingent behavior sample, the
contingent valuation sample and the samples pooled together
can be seen in Table 2.
In summary, the survey sample was split evenly between
male and female respondents. The average respondent ranged
in ages from 18 to 29 and completed a Bachelor's degree as
their highest level of education. They spent an average of
26 days in Bolivia, while 3 days were spent at the Eduardo
Avaroa Reserve and 3 days were spent at other national parks
within Bolivia. Approximately 60% of the respondents spent
their time in Bolivia through a tour package. The respondents
from the contingent behavior sample spent roughly $100 more
on their trip than did the respondents from the contingent
valuation sample. On average, the respondents from the
contingent behavior sample spent $10 more/day in REA than
did the respondents from the contingent valuation sample. In
addition, the respondents traveled from a variety of countries
for their trips to Bolivia (Table 3).
Approximately 67% of the respondents traveled from
European countries. The top three countries the respondents
traveled from in descending order was United Kingdom, Israel,
and France, consisting of over half the sample size.
The majority of these variables were used to evaluate the
data, however, the only variables that were significant were the
increased entrance fee level (Bid Amt), whether the respondent
originated from a European country or not (European), the
respondents annual median household income (Income), and
whether the respondent did his or her traveling through a tour
package or not (Tour Package) (Table 4).
As can be seen in Table 4, the direction of the estimated
coefficients of both elicitation methods are consistent with
economic theory, resulting in theoretical convergent validity.
Further, the coefficients can be compared to determine if they
are significantly different from one another for the CB and CV
regressions through a likelihood ratio test. Also, a pairwise ttest
on the mean dependent variable for both models can be
performed in order to determine if the two regressions are
statistically different from one another. Since the Bid Amt
variable is significant in the contingent behavior and contingent
valuation models, the willingness to pay extracted from
each model can be evaluated to see if they are significantly
different from one another.
In order to test for the coefficient equality, a likelihood ratio
(LR) test was performed. The resulting calculated chi-square is
33.91. With 5° of freedom, the critical chi-square value is 15.09 at
the 99% confidence level. Therefore, we can reject the null
hypotheses and conclude that the contingent valuation and
contingent behavior regressions are significantly different at
even the 99% level.
In order to further analyze the statistical difference among
regressions, a pairwise t-test on the mean of the dependent
variables was used. The calculated t-statistic was 4.63 and the
corresponding critical t-statistic was 2.33 at the 99% confidence
level. This provides further evidence that the contingent
behavior model is significantly different from the
contingent valuation model.
Since the bid amount variable is statistically significant for
both models, the confidence intervals on the mean willingness
to pay values need to be calculated in order to test for
significant difference between the willingness to pay across
the valuation methods. The mean willingness to pay was
calculated using the following formula:
Mean WTP þ ðlnð1 þ expðaÞÞÞ=b
The product of the coefficient and mean values for all
independent variables (constant, income level, European
country, tour package) excluding the bid coefficient is denoted
by α and β is the absolute value of the bid coefficient (Park et
al., 1991). The mean willingness to pay values and the
respective confidence intervals for each valuation technique
is in Table 5.
The mean willingness to pay under the contingent
behavior scenario is more than double the mean willingness
to pay under the contingent valuation scenario. The confidence
intervals for each scenario do not overlap with each
other; therefore, the willingness to pay under the contingent
behavior scenario is significantly different from the willingness
to pay under the contingent valuation scenario.
Table 5 – Mean WTP and confidence intervals for the two
elicitation methods
Mean willingness to pay
Mean 90% confidence interval
Contingent behavior $76.50 $62.58–$104.37
Contingent valuation $36.73 $32.38–$42.68
Table 6 – Price elasticities for each proposed fee level
across valuation techniques
Elasticities
Bid amt CB CV Pool
$5 −0.421 −0.538 −0.514
$10 −0.081 −0.115 −0.103
$20 −0.040 −0.068 −0.054
$30 −0.032 −0.062 −0.045
$50 −0.029 −0.060 −0.043
$75 −0.029 −0.051 −0.041
Table 7 – Total expected expenditure impact for each
proposed fee level across valuation techniques
Total expected expenditure impact
Bid amt CB CV Difference
$5 $2,174,856 $1,877,617 $297,239
$10 $2,182,283 $1,885,597 $296,687
$20 $2,183,178 $1,886,479 $296,698
$30 $2,183,353 $1,886,595 $296,758
$50 $2,183,421 $1,886,640 $296,781
$75 $2,183,405 $1,886,805 $296,600
523E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
Further, by calculating the price elasticities for each
scenario, it reveals that we are in the inelastic portion of the
demand curve. If entry fees to the park increased by 25% to
$5.00, the contingent behavior analysis predicts a 0.43%
decrease in the likelihood the respondent would visit the
park, whereas our contingent valuation analysis is slightly
more elastic; The contingent valuation analysis predicts a
0.54% decrease in the likelihood of a visit given a 25% increase
in the entrance fee. The combined analysis predicts a 0.51%
decrease in the likelihood of a visit, given a 25% increase in the
entrance fee. The elasticities for each corresponding bid
amount under each scenario can be seen in Table 6.
Therefore, it can be seen that the results of the contingent
behavior model are significantly different than the results
generated from the contingent valuation model. Further, it is
important to note that the respondents to the contingent
valuation scenario are more price sensitive to changes in the
entry fee than the respondents to the contingent behavior
scenario. As a result, the optimal entry fee pricing strategy will
differ across the valuation technique employed.
The total expected expenditure impact can be evaluated
from these elasticities combined with the 2000 visitation level
and the average REA expenditures. The total impact for each
bid amount under the two elicitation methods can be seen in
detail in Table 7.
The expected annual impact under the contingent behavior
scenario is approximately $300,000 greater than the expected
impact under the contingent valuation scenario. This result
illustrates that there is a financially important difference in
interpreting the results from among non-market valuation
techniques and therefore any potential compensatory policy
alternatives. Hence, it is important to incorporate both of these
models when performing non-market valuation as they produce
differing results.
6. Conclusion
Ecotourism has grown in importance during the past decade
and it has become a major asset to the economies of
developing countries (Chase et al., 1998). Our results show
that visitors' willingness to pay contingent on park management
improvements is greater than what they are currently
being charged. Chase et al. (1998) suggest that as fees are
increased, the reference point from which judgments of WTP
are based will shift and cause the current level of WTP to
increase. This occurs in the case of valuing the three parks in
Costa Rica and if more countries follow the format of
increasing entrance fees tourists may become accustomed to
paying more substantial fees (Chase et al., 1998).
In order to put a value on goods where there is no market
available, such as a national park, one needs to use a form of
non-market valuation to obtain a measure of willingness to
pay for the national park. Two forms of non-market valuation
are the contingent valuation method and the contingent
behavior method. The logit model is used to determine the
likelihood of the respondent visiting the park, contingent on a
hypothetical scenario and a given bid amount.
In order to examine the difference between the two
samples (contingent valuation and contingent behavior), the
likelihood ratio test was used to compare the significance
between the coefficients of the regressions. It was determined
that the regressions between the two samples were theoretically
similar but statistically different. It was concluded that
there was a significant difference between willingness to pay
for the contingent behavior scenario versus contingent behavior
scenario. The willingness to pay under the contingent
behavior scenario is greater than the willingness to pay under
the contingent valuation scenario. Further, the elasticity with
respect to the bid amount is less inelastic in the contingent
behavior scenario than the contingent valuation scenario,
which leads to a $300,000 annual difference in total expected
expenditure impacts.
Theoretically, both valuation techniques provide similar
results. However, with further econometric testing it was
shown that the two non-market valuation techniques provide
different statistical results. Further research needs to be
conducted in order to determine whether one valuation technique
is superior. In the case of valuation of Eduardo Avaroa
Reserve, the statistical differences in willingness to pay
responses lead to financially important differences in predicted
fee levels, visitation levels and revenues. The results
produce a large divergence in expected economic impacts
across the two non-market valuation methods. Therefore, in
developing countries, it is advisable to employ more than one
non-market valuation technique toward the support of any
protected area.
Acknowledgement
The authors would like to acknowledge Kreg Lindberg, Oregon
State University–Cascades Campus, Andy Drumm and the staff
at The Nature Conservancy Worldwide Office, Arlington,
Virginia 22203 and John Loomis, Colorado State University and
two anonymous reviewers for their comments and suggestions.
R E F E R E N C E S
Adamowicz, Wiktor L., Louviere, J., Williams, M., 1994. Combining
revealed and stated preference methods for valuing environmental
amenities. Journal of Environmental Economics and
Management 26 (3), 271–292 May.
Boardman, Anthony E., Greenberg, David H., Vining, Aidan R.,
Weimer, David L., 2005. Cost-Benefit Analysis: Concepts and
Practice, 3rd ed. Prentice Hall, Upper Saddle River, NJ.
Champ, Patricia A., Boyle, Kevin J., Brown, Thomas C., 2003. A
Primer on Nonmarket Valuation. Kluwer Academic Publishers,
Norwell, MA, pp. 269–329.
Chase, Lisa C., Lee, David R., Schulze, William D., Anderson,
Deborah J., 1998. Ecotourism demand and differential pricing of
national park access in Costa Rica. Land Economics 74 (4),
466–482 November.
Fredman, Peter, Emmelin, Lars, 2001. Wilderness purism, willingness
to pay and management preferences: a study of Swedish
mountain tourists. Tourism Economics 7 (1), 5–20.
Grijalva, Therese C., Berrens, Robert P., Bohara, Alok K., Shaw, W.
Douglass, 2002. Testing the validity of contingent behavior trip
responses. American Journal of Agricultural Economics 84 (2),
401–414 May.
Gujarati, Damodar N., 2003. Basic Econometrics, 4th ed. McGrawHill
Company. p. 595–8, 604–7, 633–5.
524 E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5
Herath, Gamini, 2002. Research methodologies for planning
ecotourism and nature conservation. Tourism Economics 8 (1),
77–101.
Lee, Choong-Ki, Han, Sang-Yoel, 2002. Estimating the use and
preservation values of national parks' tourism resources using
a contingent valuation method. Tourism Management 23,
531–540.
Leon, Carmelo J., 1997. Valuing international tourism benefits
from natural areas. Tourism Economics 3 (2), 119–136.
Loomis, John B., Walsh, Richard G., 1997. Recreation Economic
Decisions: Comparing Benefits and Costs, 2nd ed. Venture
Publishing, Inc., State College, PA.
Park, Timothy, Loomis, John B., Creel, Michael, 1991. Confidence
intervals for evaluating benefits estimates from dichotomous
choice contingent valuation studies. Land Economics 67 (1),
64–73 February.
Pigram, John J., Wahab, Salah, 1997. Sustainable tourism in a
changing world. In: Salah Wahab, Pigram, John J. (Eds.),
Tourism Development and Growth: The Challenge of Sustainability.
Routledge, New York, pp. 17–32.
Shultz, Steven, Pinazzo, Jorge, Cifuentes, Miguel, 1998. Opportunities
and limitations of contingent valuation surveys to
determine national park entrance fees: evidence from Costa
Rica. Environment and Development Economics 3, 131–149.
Stephenson, W. Robert, Undated. “Two Independent Samples.”
Department of Statistics. Iowa State University.
Studenmund, A.H., 1997. Using Econometrics: A Practical Guide,
3rd ed. Addison-Wesley Educational Publishers, pp. 509–515.
The Nature Conservancy, Undated. “Eduardo Avaroa Andean
Fauna National Reserve, Bolivia.” Fact Sheet.
Wall, Geoffrey, 1997. Sustainable tourism–unsustainable development.
In: Wahah, Salah, Pigram, John J. (Eds.), Tourism
Development and Growth: The Challenge of Sustainability.
Routledge, New York, pp. 33–49.
Wunder, Sven, 2000. Ecotourism and economic incentives — an
empirical approach. Ecological Economics 32, 465–479.
525E C O L O G I C A L E C O N O M I C S 6 0 ( 2 0 0 7 ) 5 1 7 – 5 2 5