You are on page 1of 8

Tourism Management 33 (2012) 1293e1300

Contents lists available at SciVerse ScienceDirect

Tourism Management
journal homepage: www.elsevier.com/locate/tourman

Conicting preferences among tourists and residents


Nanni Concu a, b, Gianfranco Atzeni b, c, *
a

CAEPR, Australian National University, Australia


CRENOS, Italy
c
DEIR, University of Sassari, Via Torre Tonda, 34, 07100 Sassari, Italy
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 30 November 2010
Accepted 5 December 2011

In this paper we present the results of a choice modelling (CM) experiment designed to estimate preferences of residents and tourists in Alghero, Sardinia (Italy). In 2004 Sardinias regional government
introduced a set of reforms on coastal development and environmental protection that had important
consequences for the tourism industry. The CM experiment took place in 2006, and aimed to study both
residents and tourists preferences regarding the 2004 reform and other tourist development alternatives. We also assess the hypothesis that the perceived social and environmental effects of tourism differ
among classes of respondents. The analysis indicates that there are conicting preferences within the
host community as well as between the host community and tourists. This creates a mismatch between
residents supply and tourists demand of recreational services that needs to be addressed to promote the
best tourist development strategy. It also shows that the 2004 reform is not such an effective strategy as
it matches neither residents nor tourists preferences.
2011 Elsevier Ltd. All rights reserved.

JEL classications:
C25
O13
Q01
Q56
Q58
Keywords:
Choice modelling
Preferences
Tourism
Development

1. Introduction
The relationship between tourism and host environments is
quite complex. Tourism development depends on features and
quality of natural, cultural, and heritage resources, among other
things. Tourism also demands services and goods that could alter
these environments and hence it has the potential to degrade the
resources on which its development is based. Comparing the
benets received and the costs incurred by host communities and
tourists is then necessary to determine the optimal level of tourist
development. The task is notoriously difcult because of hard-toquantify positive and negative externalities, non-market values,
and opportunity costs. For instance, tourists incur costs by both
purchasing the holiday package and suffering from overcrowding of
tourist facilities (see, for example Apostolakis & Jaffry, 2005a).
Similarly, the host community could gain from tourism revenues
and from revitalising of local traditions, and incur costs such as
disruption of social relations and environmental degradation (see
Lindberg & Johnson, 1997).

* Corresponding author. DEIR, University of Sassari, Via Torre Tonda, 34, 07100
Sassari, Italy. Tel.: 39 0792017332; fax: 39 0792017312.
E-mail addresses: nanni.concu@anu.edu.au (N. Concu), atzeni@uniss.it
(G. Atzeni).
0261-5177/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tourman.2011.12.009

In the last two decades several studies have analysed residents


and tourists attitudes and quantied the welfare effects of tourism.
This literature is vast and spans over different policy goals (e.g.
environmental protection, development, social impact assessment,
recreation demand modelling) and methods (e.g. attitudinal
surveys, revealed preference methods, stated preference methods,
etc.). In this paper we use a stated preference techniqueein
particular, the choice modelling (CM) method1 e to provide
a monetary estimation of residents and tourists perceptions of the
impacts of recreational uses of resources in host communities.
Through this technique, the welfare impacts of tourism externalities are estimated by capturing the trade-offs residents and tourists
are willing to make between attributes of hypothetical recreational
scenarios. This approach has been widely used and tested.
Lindberg, Dellaert, and Rassing (1999) employ the CM method to
estimate both the environmental and social impacts of tourism on
Bornholm Island (Denmark). Hanley, Wright, and Koop (2002)
provide CM estimates of the effects of overcrowding of rockclimbing sites in Scotland. In Apostolakis and Jaffry (2005b), CM
estimates indicate that visitors to Greek heritage sites do not seem
to be overly sensitive to congestion although reducing

1
According to a recent classication (Carson & Louviere, 2011) an alternative way
to term this type of exercise is discrete choice experiment.

1294

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

overcrowding determines welfare gains. Vici, Castellani, and Figini


(2007) survey residents of one of the major Italian seaside resorts to
estimate the welfare consequences of several tourist development
scenarios using CM. Brau (2008) provides one of the few CM
application to tourism demand in Sardinia (IT). He shows that
overcrowding and environmental degradation have substantial
detrimental effects on the welfare of tourists. Oh, Draper, and
Dixons (2009) CM analysis conrms that overcrowding and noise
have detrimental effects on beach visitors in South Caroline (US).
Brau, Scorcu, and Vici (2009) provide similar results on the effects
of congestion on tourists in Rimini (IT), and show that high risk of
overcrowding is associated with a loss of welfare. These studies
indicate the negative effects of tourism externalities both on residents and visitors. Increasingly, stated preference techniques are
also applied to provide an economic valuation of cultural and
heritage resources (see Choi, Ritchie, Papandrea, & Bennett, 2010).
In the present analysis we estimate residents and tourists
welfare changes associated with alternative tourist development
options using the CM method. One of these options describes Sardinias 2004 reforms on coastal development and environmental
protection that had important consequences for the tourism
industry. In the literature there are only few papers that use the CM
technique to analyse both sides (residents and tourists) of the same
tourist market. Lindberg, Andersson, and Dellaert (2001) compare
welfare changes of residents and tourists of a hypothetical ski
resort development in Sweden. Employing two different payment
vehicles (tax increase/reduction for residents and price increase for
lift tickets for tourists), they nd that the net effect on the
community is negative, and, although tourists benet from the
slope expansion, their gains are not enough to outset the net losses
of residents.
Hearne and Santos (2005) compare preferences towards alternative scenarios of ecotourism of foreign tourists and educated
local residents in Maya Biosphere Reserve in Peten, Guatemala. The
CM analysis is made using qualitative attributes and the entrance
fee as payment vehicle, with different levels for residents and
foreign tourists. Their results show that both groups want adequate
park management, wildlife viewing with guides, and are willing to
accept an entrance fee. The preferences of tourists and residents
diverged in respect to the presence of colonists within the reserve
core zone and paved access roads.
Hearne and Tuscherer (2007) carry out a choice experiment
analysis of potential tourist and local resident preferences for
alternative ecotourism development scenarios for the Standing
Rock Sioux Indian Reservation. They calculate willingness to pay for
tourist and local people using the per person price of a tour package
as payment vehicle. They observe that residents did not have
a signicant preference for lower prices, because they do not expect
to pay the price themselves, and they may consider high prices as
a source of additional income for the area.
Our study aims to give a contribution to this literature with
some distinctive features. We evaluate the overall effect of a real
(rather than hypothetical) policy reform on coastal management,
that affects the welfare of stakeholders in an important tourist
destination in Sardinia. We design the payment vehicle to provide
the same overall cost of policy scenarios for both residents and
tourists. Further, the residents sample is segmented to highlight
the effect of sources of income (tourism vs non-tourism) on
preferences. This helps to clarify the role of a potential source of
divergent preferences in host communities (see Petrzelka,
Krannich, Brehm, & Trentelman, 2005). Contrasting residents
and tourists perceptions helps to identify the development
trajectories that minimise conicting demands, and to ne tune
the policy framework towards promoting the best development
options.

The paper is organised as follows. In the next section, we


describe the guiding principle of the 2004 reform and an overview
of the study site. Then we describe the methodology, the questionnaire design, and the survey administration (Section 3). In
Section 4 we illustrate the results and discuss their implications in
Section 5. Finally, we provide conclusions in Section 6.
2. The study area
Alghero is a town of about 40,000 residents located on the
north-western coast of Sardinia (Italy). In the 1950s, Alghero was
one of the rst Sardinian towns to invest in tourism, and its residents are now mostly employed in the wholesale and retail trade
(29%), constructions (13%) and hospitality (13%) industries (ISTAT,
2001). Tourists concentrate mainly in the town causing congestion. Large tracts of Algheros coastline are still pristine or only
partly developed. Marine protected areas and parks account for
around 8000 ha of land and coastal waters, and around 100 km of
coastline (Citt di Alghero, 2008). Tourism, agriculture and environmental conservation compete over these areas. In the last 10
years Alghero has experienced an increase of tourist ows, with
considerable economic benets as well as overcrowding, disruptive
congestion, and some environmental degradation (Giordano &
Marini, 2008; Porcu, 2010).
Coastal development policies have been the source of serious
political debate in recent years in Sardinia. In 2004 the regional
government introduced two coordinated measures: the Sustainable
Tourism Development Regional Plan (Piano Regionale di Sviluppo
Turistico Sostenibile, PRSTS), and the Regional Landscape Plan
(Piano Paesaggistico Regionale, PPR). The PRSTS represents the rst
strategic plan for the preservation and use of natural resources for
tourism purposes in Sardinia (Biagi & Pulina, 2007). The PPR is the
main instrument for implementing conservation and protection
measures. Two principles guided the 2004 framework: sustainable
development, and homogeneity of planning processes. It also
identied the landscape2 as the basic planning unit. The major
prescription of the PPR is the total ban of new buildings and infrastructures within 2 km from the seashore.3 On one hand, this
regulation aimed at protecting the coastal landscape and environment; on the other hand, it posed extensive limits on the use of this
resource for tourism development. The regulation was particularly
controversial. Legislators claimed it necessary to both protect the
environment and ensure the sustainability of the Sardinian tourism
industry. In other words, they assumed Sardinian residents would
benet from both increased environmental protection and
sustainable tourism. Others claimed the regulation was too restrictive and severely constrained economic opportunities for residents.
Another important regulation of the 2004 reform includes a set
of incentives for the renovation and restoration of old suburbs in
tourist towns. This measure aimed at reducing the spread of tourist
infrastructures, and making better use of the existing urban assets
(Regione Sardegna, 2010). An unintended consequence of such
a measure is the concentration of tourists in urban areas, with
increased congestion affecting both residents quality of life and the
recreational experience.
Alghero is an ideal place to investigate on communities and
tourist preferences, and the potential conicts between them.
Alghero is one of the most important destination in Sardinia, and
hence any change in its tourist ows has consequences for the

2
The LR 8/2004 denes landscape as the set of natural, architectural, and
archaeological elements of the Sardinian countryside.
3
The ban is valid for all towns that failed to adopt a local development plan. At
the time of the reform, most of Sardinian towns did not have one.

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

economy of the region. It is a mature locality with a long history


of tourism development. Also, residents in Alghero have already
experienced the negative effects of tourism, such as overcrowding
and declining environmental quality. Furthermore, the town has
a lot of development potential with over 20% of its territory under
some form of environmental protection.
Given these features, Alghero provides an example of the
problems and successes of communities that have heavily invested
in tourism development. Any other community considering tourist
development could draw from the experience of Alghero. Should
Alghero invest in more environmental protection, as dictated in the
PPR? Do tourists and residents actually demand more environmental protection and less congestion?

3. The choice modelling application

1295

means; vzi is a random term usually normally distributed (but other


distribution are possible) with mean 0 and variance 1, so that the wz
is the standard deviation of the marginal distribution of bij. Random
parameters (RP) models are based on this specication.
An additional level of heterogeneity could be introduced in RP
models by letting the distributions of individual-specic parameters be heteroscedastic. Parameter heteroscedasticity can also be
made dependent on individual characteristics:
0

w2iz w2z expuz xi

(5)

Recent development of the RP models allows for the capture of


additional alternative specic variation that can be rendered individual specic by deep parameterisation of the variance in the error
component (EC) models (Greene & Hensher, 2007). The EC model
adds up to J alternative and individual-specic random terms to
equation (1) above:

X


Uij Vij Zj ; Si 3 ij
djm qm uim

3.1. Method

(6)

CM is a questionnaire-based data-generation technique that


produces rich data sets in which choice-specic attributes explain
individual choices. CM is rooted in Random Utility (RU) theory. The
utility Uij of a given alternative j for an individual i is a function of
the attributes Zj of alternative j, as well as a function of individual
characteristics Si:

where djm equal to 1 if effect m appears in utility function j,


0 otherwise, and uim are random terms with zero mean. The variance of the error components is then qm, that can be made heteroscedastic by deep parameterisation:



Uij Vij Zj ; Si 3 ij

q2mi q2m 0 exptm xi

(1)

where Vij(Zj, Si) is the observable or systematic component, and 3 ij is


a stochastic element. A second utility relation links the probability
of an outcome to the utility of each alternative. That is, individuals
are assumed to choose the alternative yielding the highest utility:






Prij jZj ; Zk ; Si Pr Vij 3 ij >Vik 3 ik ;

c jsk

(2)

c jsk

(3)

and




 

Prij jZj ; Zk ; Si Pr Vij  Vik > 3 ik  3 ij ;

The parameters of the systematic component of equation (1) can


be estimated conditional on the distribution properties of error
terms 3 ij. The most common assumption is that 3 ij are independently and identically distributed (IID) extreme value (EV) type 1
(Train, 2003). This assumption is at the core of the Multinomial
Logit model, of which the conditional logit (CL) is the appropriate
version to analyse choices dependent on Zj characteristics of
alternatives J. In this specication the parameter vector b is not
individual specic.
Individual heterogeneity could be accounted for by the interactions of individual socio-economic characteristics with choicespecic attributes (Bhat, 2000; Concu, 2009; Train, 2003). In this
case, estimated parameters and coefcient means are specied as:

bij bjz d0z xi w0z vzi

(4)

The vector xi is the source of individual heterogeneity, bjz is


a constant term, and dz is a coefcient that produces individual

(7)

We tested all these models accounting for individual


heterogeneity.
One of the major advantages of CM applications is the possibility
to provide estimates of the value of marginal changes in the level of
each attribute (Morrison & Bennett, 2004). These are known as
implicit prices (IPs), and are point estimates of the value of a unit
change in an attribute. IPs identify the relative importance
respondents place on attributes. They also provide useful information for management decisions as they indicate, for instance, the
value of an extra job, or the value of a marginal increase in environmental protection along the coast. IPs are calculated as:

IPz

bz

!,
dzn Xn

bCOST

!
(8)

dCOSTn Xn

where dzn is the n-th coefcient of the interaction term of attribute


z and socio-economic characteristic n, Xn is the mean value of
variable n, and bCOST represents the coefcient for the attribute
measured in monetary units.
3.2. Sampling
We drew a sample stratied according to age classes and gender
according to the 2001 Census for Alghero (ISTAT, 2010). Table 1
shows the distribution of the sample and the population. The
sample is 1.4% of the population and mirrors the population characteristics for gender and age distribution. The characteristics of the
tourist sample are summarised in Table 2. At the time of writing no

Table 1
Age and gender distribution of Alghero residents (population and sample).
Age class

Population
Male

16e40
41e65
>65
Total

Sample

Population

Male

7220
6624
2914

20.6
18.9
8.3

104
96
42

20.8
19.2
8.4

16,758

47.9

242

48.3

Population

Sample

Female

Total

Total

7176
7120
3895

20.5
20.4
11.1

102
102
55

20.4
20.4
11.0

14.396
13.744
6.809

206
198
97

18,191

52.1

259

51.7

34.949

501

Female

Sample

1296

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

Table 2
Age distribution and gender in the tourist sample.
Age class
16e40
41e65
>65

Male %
59.8
35.3
4.9
100

Female %
49.5
43.8
6.7
100

Total %
54.2
39.9
5.9
100

data is available on the characteristics of the tourist population in


Alghero.
3.3. Survey design and administration
The choice modelling application was organised in three stages.
First, we used a preliminary questionnaire to interview a random
sample of residents. The questionnaire contained open and Likerttype questions on the effects of tourism on natural resources, public
services, residents quality of life, and on other socio-economic
indicators. The preliminary questionnaire provided important
indications of residents perceptions of tourism, and allowed us to
identify the most important effects of tourism development in
residents opinion. Out of this information, we dened four attributes to describe the main perceived effects of tourism and to
create various scenarios of tourism development. Further, the
preliminary questionnaire contained some questions on policies for
promotion and management of tourist ows, and on how to fund
these policies. Finally, sampled individuals stated their willingness
to pay for the proposed policies. This was an open-ended question
designed to provide a range of monetary values to be employed in
the nal questionnaire.
In the second stage, we used this information to design the
choice modelling questionnaire. This questionnaire has three parts.
The rst contains a set of questions on respondents attitudes, and
some reminders of income constraints on individual and public
choices. The second part contains the choice sets generated by
combining ve levels of the four attributes. These combinations
describe the claimed effects of PPR (the status quo at the time) and
hypothetical alternatives. The four attributes are:
 Distance of new buildings from the seashore: This attribute
describes the level of protection assigned to the coastal environment, and it has ve levels: a) 150 m from the shoreline.
This is the limit set by the legislation for tourism infrastructures prior to the 2004 reform; b) 500 m from the shoreline;
this is the limit set by the legislation for private houses prior to
the 2004 reform; c) 1000 m from the coast; this is an intermediary level between the pre-reform limits and the 2004
prescriptions; d) 2000 m (status quo level); this is the limit
established in the PPR; e) 3000 m from the coast; this level
captures preferences for more stringent environmental
protection.
 Number of new jobs in the tourist sector (per year in
Alghero): this indicator measures the economic impact of
tourism. The attribute has ve levels: a) no new jobs. As noted
before, the 2 km limits in the PPR was controversial, with
several operators claiming it would seriously hinder the
Sardinian tourism industry. Sardinian tourism development
depends on new investment that would ensure the industry
stay competitive in the international market. Limits to tourism
infrastructure constrain new investments and may cause loss
of competitiveness; b) 20 new jobs per year; this level corresponds to 1% increase of employment in the tourism sector, and
it is lower than the 10 year trend in Sardinia; c) 40 new jobs
per year (status quo level); this is the expected effect of the

PPR. This level corresponds to a 2% increase in the number of


jobs, and it equals the 10 year trend of job growth in the
tourism industry in Sardinia. The underlying assumption is that
the PPR would neither worsen nor improve the competitiveness of the Sardinia tourism industry; d) 60 new jobs per year
equal to a 3% growth in the number of jobs; e) 80 new jobs per
year; this level correspond to a 4% increase in the number of
jobs in the tourism sector. In summary, the ve levels describe
the effects of tourism development on the economy, and they
correspond to a serious decrease, a decrease, no change, an
improvement and a substantial improvement in the number of
jobs.
 Increase of time required to perform daily activities such
as nding a parking space, shopping, visiting the local post
ofce or bank. The PPR provides incentives to renovating
urban areas, especially old suburbs of tourist towns. This
might attract tourists, reduce their dispersion and concentrate them in existing infrastructures. The attribute captures
the impact of the increased concentration of tourists in
towns, as more tourists in less space would increase
congestion. As seen above, the literature has shown congestion caused by visitors has negative effects both on residents
and other visitors. This attribute has ve levels: a) 5 min: this
is the expected increase in congestion if tourist ows
remained unchanged; b) 10 min; c) 15 min (status quo level);
d) 30 min; e) over 30 min.
 Payment vehicle. This is the only attribute that changes
between residents and tourists questionnaires. For residents,
the scenarios proposed a local tax increase (absolute value
per year). This monetary attribute is the respondents
contribution to the proposed policy. It takes the form of
a willingness to pay (WTP) measure. In the present application, the WTP was chosen because tax increases are more
plausible than tax reductions. This attribute has ve levels: a)
0 euro per year (status quo); b) 10 euros per year; c) 20
euros per year; d) 30 euros per year; e) 40 euro per year. For
tourists, the attribute took the form of a daily tourist tax that
is, an increase of the cost of their holiday in Alghero per day.
As the mean length of stay in the study area is around 5 days
(Vannini & Puggioni, 2007), the levels for this attribute were
set to: a) 0 euro per day (status quo); b) 2 euros per day; c) 4
euros per day; d) 6 euros per day; e) 8 euro per day. This
way the overall cost of a policy scenario is the same for
residents and tourists.
The status quo levels describe the expected effects of the PPR,
based on the policy makers claim and past trends of the tourism
industry. However, there was still a great deal of uncertainty
regarding the true impact of the PPR. Hence, the time frame
proposed to residents was ve years, that is, the proposed changes
take place every year for ve years. We expected this short time
frame to make it easier for respondents to choose the preferred
alternative. Choice attributes are summarised in Table 3.
The design of the experiment is based on a Hyper Graeco-Latin
(HGL) square. This design creates choice sets e i.e. combinations of
attributes and levels e that have the same balanced and orthogonal
properties of fractional factorial designs (Cox & Reid, 2000). It is
also a main-effect-only design which allows estimating only the
effect of each single attribute on the probability of choices. The HGL
square design gives 25 choices.
Finally, the questionnaire contains a set of questions on the
socio-economic characteristics of the sampled individuals. One of
these questions was designed to capture the degree to which an
individuals income is dependent on tourism. The question was
structured as a ve-point scale ranging from 100% to 0% of income

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

1297

4. Results

Table 3
Choice attributes.
Attributes

Description

Label

Levels

Distance of new
buildings from
the seashore

Level of protection
assigned to the
coastal environment

DIST






Number of new
jobs in the
tourist sector
(per year in
Alghero)

Economic
impact of tourism

EMPL

 0
 20
 40
(Status quo)
 60
 80

Increase of time
required by
daily activities
(in minutes)

Impact on residents
quality of life of
increased tourist ows

TIME

 5
 10
 15
(Status Quo)
 30
 over 30

Payment vehicle

Residents:
 local tax increase
(absolute value
in euros per year)

COST












Tourists
 increase of cost of
holidaying in Alghero
(absolute value in
euros per day)

150 m
500 m
1000 m
2000 m
(Status quo)
 3000 m

0 (Status quo)
10
20
30
40
0 (Status quo)
2
4
6
8

derived from tourism. Two focus groups comprised of postgraduate and under-graduate students were organised to assess
the effectiveness of the choice format and the questionnaire design.
This focus group suggested some modications that were then
incorporated in the nal draft of the questionnaire. The survey was
administered in person. Each interviewee received a set of cards
each containing two choice sets: the status quo option and an
alternative scenario. We divided the sample in two blocks and
randomised the order of choice selection. Trained staff administered the survey in different areas of Alghero in June 2006.

501 Residents and 175 tourists were interviewed. We analysed


the data using several model specications to account for individual heterogeneity. These models are listed in Table 4. We nd
that these models produce statistically similar results, the only
exception being a RP model that doesnt include socio-economic
characteristics. In Fig. 1 we have collected the implicit price and
95% condence interval of the choice attribute DIST estimated using
the listed models. IPs are signicant for the conditional logit (CL)
model, and for the RP1 model (Random parameter with no stratication). In the other models, residents are stratied according to
their source of income. In these cases, IPs are signicant for those
residents who earn no income from tourism (models RP2, RP4, RP6,
and RP8). For those residents who earn at least some income from
tourism, IPs are calculated using model RP3, RP5, RP7 and RP9, as
explained in Table 4. For these residents only, IPs are not signicant.
Model RP1 gives estimates that are statistically different from the
other models. Fig. 1 also shows that the CL model estimates are not
statistically different from the more general models, albeit with
larger variances. In what follows, we present results of the RP
models with mean heterogeneity for sampled residents and tourists (Table 5).
Likelihood ratio tests (Louviere, Hensher, & Swait, 2000) helped
to determine the nal set of socio-economic variables that enter the
models. We use the pseudo R2 statistics and the percentage of
correct predictions to assess model performance. A good model has
a pseudo R2 between 0.2 and 0.4 (Louviere et al., 2000). As shown in
Table 5, the pseudo R2 statistics has a value of 0.24 and 0.34 for
resident and tourist models respectively. The percentage of correct
predictions is calculated comparing actual and predicted choices. In
our application the model for residents correctly predicts 73% of
actual choices, and the RP model for tourists predicts 78% of actual
choices.
The alternative specic constants (ASCs) are variables that take
value of 0 for the current scenario (the status quo) and 1 for the
alternative policy option. They capture the average impact of any
source of heterogeneity left unexplained by the model. For both
residents and tourists, the ASC parameters have a negative sign. It is
an indication that sampled individuals prefer the status quo over

Table 4
List of model specications.
Model

Label

Description

Means & variances

Conditional logit

CL

E[bij] bj

Var[3 ij] s2j

Random parameter

RP1

E[bij] bij

Random parameters
with mean heterogeneity

RP2, RP3

Var[3 ij] s2j


Var[bij] w2z
Var[3 ij] s2j
Var[bij] w2z

Random parameters with


mean heterogeneity and
error component

RP4, RP5

Random parameters with


mean heterogeneity, error
component and heteroskedasticity
in error component

RP6, RP7

Random parameters with mean


heterogeneity, and parameter
variance heteroskedasticity

RP8, RP9

Standard conditional logit with interaction


terms to account for individual heterogeneity
Random parameter model e no socio-economic
characteristics
Random parameter model e means of random
parameters are dependent on socio-economic
characteristics. Respondents are also stratied
according source of income: tourism (RP3)
vs non-tourism (RP2)
Random parameter model e means of random
parameters are dependent on socio-economic
characteristics. Respondents are also stratied
according source of income: tourism (RP5) vs
non-tourism (RP4). Errors are choice-specic.
Random parameter model e means of random
parameters are dependent on socio-economic
characteristics. Respondents are also stratied
according source of income: tourism (RP7) vs
non-tourism (RP6). Errors are choice-specic.
Error components are heteroskedastic.
Random parameter model e means and variance
of random parameters are dependent on socio-economic
characteristics. Respondents are also stratied according
source of income: tourism (RP9) vs non-tourism (RP8).

E[bij] bj f(xi)

E[bij] bj f(xi)

Var[3 ij] s2j


Var[bij] w2z
Var[uim] q2m

E[bij] bj f(xi)

Var[3 ij] s2j


Var[bij] w2z
Var[uim] q2m0 exp(tmxi)

E[bij] bj f(xi)

Var[3 ij] s2j


Var[bij] w2z 0 exp(uzxi)

1298

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

Fig. 1. Implicit price and 95% CI for DIST e residents (in V per year).

the alternatives, everything else being the same. This is a welldocumented empirical nding, also known as status quo bias:
individuals have the tendency to stick with the familiar alternative
because the disadvantages of leaving it loom larger than the
advantages of a change (Kahneman, Knetsch, & Thaler, 1991).
Parameters of choice attributes have the expected signs, and are
consistent across the two samples. The coefcient for the attribute
DIST is positive, but statistically signicant only for tourists. Because
increasing the minimum distance from the coast for new buildings
means more environmental protection, the probability of tourists
choosing a development option increases as environmental
protection increases. The coefcient for the attribute EMPL is also

Table 5
Random parameter models with mean heterogeneity.
Variables

Residents
Coeff.

Tourists
Standard
error

Non-random parameters (b)


Alternative
0.6477*** 0.101
specic constant
Random parameters of choice attributes (bij)
Distance of new
5.17E05 0.788E04
building from the
seashore (DIST)
Number of new jobs
0.0796*** 0.009
in the tourist sector
(EMPL)
Increase of
0.0452*** 0.009
congestion (TIME)
Local tax increase
0.0276*** 0.006
(in euro) (COST)
Increase of cost
of the holiday
(in euro) (COST)
Heterogeneity in mean (dij)
DIST:gender (female 1) 0.0005***
DIST:income from
0.0004***
tourism (yes 1)
DIST:# of dependants
EMPL:age
0.0006***
EMPL:income
TIME:income from
0.0324***
tourism (yes 1)
TIME:age
COST:gender (female 1) 0.0096***
COST:income from
0.0155***
tourism (yes 1)
COST:# of dependants
0.0048***
COST:age

Coeff.

Standard
error

0.9290***

0.246

3.67E04*** 0.000

0.0910***

0.019

0.0331

0.037

0.063

0.064

0.000
0.000
0.0002 *** 0.779976E04
0.000
6.01E7*** 0.283362E06
0.014
0.0024***

0.001

0.0047***

0.002

0.005
0.002
0.006

Derived standard deviations of parameter distributions (wiz)


NsDIST
0.0015*** 0.000
0.0003
NsEMPL
0.0146
0.010
0.0791***
NsTIME
0.0639*** 0.024
0.1299***
NsCOST
0.0450*** 0.013
0.14
Pseudo Rsq
Log likelihood
Observations
% Correct predictions

0.237
3306.55
6264
72.9

0.338
997.75
2188
77.7

***Signicant at 1%, **signicant at 5%; *signicant at 10%.

0.000
0.023
0.050
0.104

signicant and positive indicating that both residents and tourists


prefer more jobs. The coefcients of the other two attributes have
negative signs. Higher cost (COST) and increasing congestion (TIME)
were both valued negatively, and they both are signicant only for
the resident sample. That is, while residents are concerned about
a tax increase and increased congestion, tourists are not concerned
about an increase in costs of their holidays and tourist services.
Socio-economic characteristics affect coefcient estimates. The
model shows that female residents prefer more environmental
protection than males. Gender also affects the COST attribute in the
resident model, and female respondents are less likely to gain from
policies as their cost increases. No effect of the gender dummy is
found for the tourist sample. Another source of individual heterogeneity is respondents income. It is usually assumed that higher
income is associated with stronger preferences for environmental
conservation e the so called environmental Kuznets curve hypothesis
(see Stern, Common, & Barbier, 1996). We nd no such effect on the
models.4 Further, tourists income has a statistically signicant
negative effect only on the EMPL attribute. It shows that tourists
with higher income are less worried about supporting employment
in the host localities.
We also stratied the resident sample according to their source
of income. This stratication is based on the self-reported degree of
dependence of individual income on tourism measured on a 5 class
Likert scale. However, as there was not enough data for each class,
sampled respondents were stratied in two groups. We used
a dummy variable taking value of 0 when the respondent earns no
income from tourism or tourism related activities (72% of sample),
and 1 otherwise (28% of sample). Through this stratication we test
the hypothesis that preferences depend on the perceived impact
a policy has on respondents main source of income. In particular,
we expect that individuals in the second sub-sample (dummy 1)
have weaker preferences for increased constraints to tourism
development than other respondents. According to the model
results, we do not reject this hypothesis. The coefcient for the
DIST: Income from tourism interaction term is negative and statistically signicant. Residents involved in tourism would gain less
from an increase in constraints to new buildings. The same dummy
variable has also a negative and statistically signicant impact on
the TIME attribute. Respondents earning at least part of their
income from tourism value negatively an increase in congestion
levels. They may associate it with a decrease in the quality of their
tourist offer. A positive impact on the coefcient of the COST attribute is recorded for the sub-sample that earns their income from
tourism. In other words, this class of respondents are less concerned about a tax increase. Tourists stratication according to
other socio-economic characteristics (e.g. level of education) does
not provide signicant results.
Table 6 shows the implicit prices (IPs) and 95% condence
intervals of the three non-monetary attributes. Condence intervals are calculated using WALD procedure in Limdep (see Greene,
2009). The WALD procedure computes values and standard errors
for specied linear and non-linear function of estimated parameters. We used the Delta method with the WALD command that, as
shown by Hole (2007), is as accurate as the other procedures to
compute condence intervals for IPs. IPs are monetary measures of
welfare changes associated with a unit variation of an attribute. For
instance, the IP of the attribute DIST for the sub-sample of residents
that earn no income from tourism is V0.013 per year for every
metre buildings are pushed away from the shoreline. Implicit prices

4
The microfoundation of EKC are controversial (Dinda, 2004; Roca, 2003). The
results of this study indicates indeed that the EKC hypothesis should be treated
with caution.

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

1299

Table 6
Implicit prices.
Attribute

Distance of new building from the seashore


Number of new jobs in the tourist sector
Increase of congestion

Residents (income from tourism no)

Residents (income from tourism yes)

Tourists

IP (V)

95% CI

IP (V)

95% CI

IP (V)

95% CI

0.013***
1.95***
1.11***

0.007e0.019
1.374e2.527
1.576 to 0.642

0.003
3.15***
3.07***

0.009 to 0.015
1.539e4.753
4.838 to 1.303

0.005
1.28
0.46

0.004 to 0.015
0.997 to 3.552
1.631 to 0.702

***Signicant at 1%, **signicant at 5%; *signicant at 10%.

for the other attributes have analogous interpretations. Increasing


the number of jobs by one unit would create a welfare increase for
residents with no income from tourism equal to V1.9. For the same
sub-sample, decreasing congestion also create a welfare increase
valued at V1.1. IPs exhibit interesting differences between the two
resident sub-samples, and between them and tourists. While
respondents with no income from tourism are concerned about
environmental protection, both tourists and residents involved in
the tourism sector are less interested in increasing coastal conservation. Tourists seem to be happy about the quality of environmental goods they have access to in Alghero. Individuals involved
in the tourism industry have also little to gain from further
constraints to tourism development. Residents are also clearly
interested in increasing the number of jobs in the tourism industry,
while tourists have no similar concern. The implicit prices of the
TIME attribute suggest a mismatch between residents involved in
tourism and tourists. The latter show no indication they are worried
about congestion (the IP is not statistically different from zero). On
the contrary, respondents earning an income from tourism are
heavily affected by tourism related congestion. Their loss of welfare
is over V3 per a marginal increase in congestion level. This IP may
be capturing perceived negative externalities caused by tourists on
tourist service providers own customers and business. Tourists
other than own customers decrease the quality of holidaying in
Alghero, and hinder efforts to improve the services, or increase
their costs; this interpretation, however, seems at odds with tourists preferences represented by a not-signicant IP for the TIME
attribute.
5. Discussion
Results indicate that there are several conicting preferences
over tourism development. First, increasing the level of environmental conservation in Alghero provides welfare gains only to
those residents that do not earn an income from tourism. Both
tourists and tourist service providers are not concerned about
environmental protection. Second, congestion seems to negatively
affect Alghero residents e including those that earn their income
from tourism; however, tourists do not indicate they are impacted
by congestion. Third, increasing employment opportunities is very
important for residents, while tourists have no concern for the
impact of tourist development on local jobs. Arguably, any tourism
development project or policy should be able to reconcile these
different effects so as to minimise conicts and increase both
residents and tourists welfare.
Although we are unable to determine conclusively the welfare
impact of the reform, as we have no information on residents and
tourists preference prior to its implementation, we can analyse
some possible scenarios. We use equation (8) to calculate the
average welfare change, for the individuals in the sample, generated by a return to the environmental standards and development
path existing before the 2004 reform. This option implies new
tourism infrastructures could be build as close as 150 m from the
shoreline. The construction and tourism industries would then
drive an increase in the number of jobs. Employment is assumed to
increase by 3% per year rather than a mere 2% as expected under the

2004 reform.5 Congestion is assumed to stay the same. Under this


scenario, the sample average welfare change is positive. Residents
would gain around V14.75 per person per year by returning to the pre2004 policy. We then consider two hypothesis: a) prior to its
implementation, residents expected some welfare loss from the
2004 reform; and b) prior to the 2004 reform, residents expected
some welfare gains from its implementation. Given the rst
hypothesis, and considering the residents preference to returning
to the pre-reform conditions, one can speculate that the 2004
reform was a top-down policy that didnt reect residents preferences, and failed to change them. Under the second hypothesis, it
could be said that the reform did not match the initial expectations,
possibly because by the time of our study it had not generate
meaningful results, and was subsequently rejected. Preferences
could also have changed as a result of the intense political debate
over the reform, that possibly polarised the community. The PPR
might have placed too much emphasis on environmental conservation by dening a too-stringent standard for new tourism infrastructure while failing to address the most important issues, i.e.
congestion and employment. Further, as it does not match the
current tourists and residents preferences, the 2004 reform might
have targeted a market segment different from the actual tourist
demand and supply, and hence it might have increased business
risk for tourism enterprises.
Arguably the 2004 reform has failed to mitigate conicts over
resource use and development in Alghero. This does not imply that
in other localities the reform could not provide a welfare improving
framework for tourism development. It does imply however that
the reform does not t every destination.
What policy could instead address residents concerns while
promoting environmental conservation?
The command-and-control policy (PPR) failed to match both
residents and tourists preferences and generated uncertain effects
on tourism development, employment and congestion. Our study
shows that tourists are not sensitive to changes in the cost of their
holidays in the price range used in the analysis (from 0 to 8 euros per
day). Hence the local authorities could rise revenues at the cost of
only a small reduction of the number of tourists and their length of
stay. Collecting these revenues through a tourist tax and investing
them in job creation, environmental protection, and improving the
quality of tourist services and attractions could be more effective
solution to mitigate conicting preferences in the community.

6. Conclusion
Identifying the best alternative for tourism development
requires knowledge of its effects on the environment and the
economy, as well as the welfare impacts on local residents. This
information is also necessary to assess present and past policy

5
In the run up to the 2008 regional election, the incumbent coalition, that
implemented the 2004 reform, and the opposition parties engage in a hot debate
on the effects of the policy on employment and growth. The incumbent was later
defeated. The rst act of the new regional government was to repel the 2004
reform.

1300

N. Concu, G. Atzeni / Tourism Management 33 (2012) 1293e1300

proposals for tourism development. This study is the rst attempt


to understand these issues in Sardinia. We use the choice modelling
technique to analyse Alghero residents and tourists preferences
regarding several alternative tourism development strategies. Each
alternative describes a different combination of environmental
protection, number of jobs, and congestion generated by tourists.
These alternatives are contrasted with the policy scenario dened
by the 2004 reform of the legal framework for planning and
building in coastal areas in Sardinia.
Results indicate that increasing environmental protection has
positive effects only on the welfare of residents that do not earn
their income from tourism. Decreasing congestion and increasing
job opportunities, on the other hand, have substantial impacts on
the welfare of all residents. Tourists do not seem to be concerned
about the effects of congestion on their recreational experiences.
There appears to be a mismatch between residents desired supply
of recreational services and tourists demand.
These results provide some policy indications. Firstly, the 2004
reform appears unable to address these conicting preferences. The
PPR and its standards for tourism infrastructure are not perceived
as able to meet both residents and tourists preferences. An alternative policy based on a tourist tax is more likely to generate
revenues to support job creation, reduce congestion, and limit or
decrease tourisms environmental impact. In other words, a tax
may help Alghero to mitigate conicting preferences in the
community, and ne tune the supply of tourism services.
Caution should be exercise when extrapolating these results to
the Sardinian population. The 2004 reform applies uniformly to the
entire island, but localities and towns are in different stages of
tourism development.
Assuming decreasing marginal utility of natural resources, one
could speculate that in localities with unexploited resources the
natural environment would be not the major cause for concern. In
such localities environmental protection may be a secondary issue
when valuing policies for tourism development. An empirical test
of this conclusion is clearly desirable. The study should then be
replicated in localities with different degrees of tourism development. Such extension would also enable assessment of the applicability of a uniform legislation across the region. Further
investigation on the relationship between preferences and
distances from environmental amenities is also required. In our
study this relationship is assumed to be linear, but testing other
possible functional forms can be a direction for further research.
References
Apostolakis, A., & Jaffry, S. (2005a). Stated preferences for two Cretan heritage
attractions. Annals of Tourism Research, 32(4), 985e1005.
Apostolakis, A., & Jaffry, S. (2005b). A choice modelling application for Greek
heritage attractions. Journal of Travel Research, 43, 309e318.
Bhat, C. R. (2000). Incorporating observed and unobserved heterogeneity in urban
work travel mode choice modelling. Transportation Science, 34, 228e237.
Biagi, B., & Pulina, M. (2007). Stessa spiaggia, stesso mare: turismo, ambiente e qualit
della vita. In L. Punzo, & S. Usai (Eds.), Lestate al mare. Milano: McGraw-Hill.
Brau, R. (2008). Demand-driven sustainable tourism? A choice modelling analysis.
Tourism Economics, 14(4), 691e708.
Brau, R., Scorcu, A., & Vici, L. (2009). Assessing visitor satisfaction with tourism
rejuvenation policies: the case of Rimini, Italy. Journal of Environmental Planning
and Management, 52(1), 25e42.

Carson, R. T., & Louviere, J. J. (2011). A common nomenclature for stated preference
elicitation approaches. Environmental and Resource Economics, 49(4), 539e559.
Choi, A. S., Ritchie, B. W., Papandrea, F., & Bennett, J. (2010). Economic valuation of
cultural heritage sites: a choice modelling approach. Tourism Management,
31(2), 213e220.
Citt di Alghero. (2008). Il rapporto di sostenibilit della citt di Alghero. Alghero.
Available
at
http://www.comune.alghero.ss.it/sito/atti_documenti/piani_
programmi.htm Last retrieved 15.03.11.
Concu, G. B. (2009). Measuring environmental externality spillovers through choice
modelling. Environment and Planning, 41, 199e212.
Cox, D. R., & Reid, N. (2000). The theory of the design of experiments. London:
Chapman & Hall/CRC.
Dinda, S. (2004). Environmental Kuznets curve hypothesis: a survey. Ecological
Economics, 49(4), 432e455.
Giordano, F., & Marini, A. (2008). A landscape approach for detecting and assessing
changes in an area prone to desertication in Sardinia (Italy). International
Journal of Navigation and Observation, 1e5.
Greene, W. H. (2009). LIMDEP version 9.0 econometric modelling guide. Econometric
Software Inc.
Greene, W. H., & Hensher, D. A. (2007). Heteroscedastic control for random coefcients and error components in mixed logit. Transportation Research Part E:
Logistics and Transportation Review, 43, 610e623.
Hanley, N., Wright, R. E., & Koop, G. (2002). Modelling recreation demand using
choice experiments: rock climbing in Scotland. Environmental and Resource
Economics, 22, 449e466.
Hearne, R. R., & Santos, C. A. (2005). Tourists and locals preferences toward
ecotourism development in the Maya biosphere reserve, Guatemala. Environment, Development and Sustainability, 7, 303e318.
Hearne, R. R., & Tuscherer, S. (2007). Stated preferences for ecotourism alternatives
on the Standing Rock Sioux Indian reservation. Agribusiness and Applied
Economics Report, 601.
Hole, A. R. (2007). A comparison of approaches to estimating condence intervals
for willingness to pay measures. Health Economics, 16, 827e840.
ISTAT. (2001). 8 Censimento dellindustria e dei servizi. Istituto Nazionale di Statistica. Available at http://dwcis.istat.it/cis/index.htm Last retrieved 26.05.11.
ISTAT. (2010). 14th General census of population and housing. Istituto Nazionale di
Statistica.
Available
at
http://dawinci.istat.it/daWinci/jsp/dawinci.jsp?
qpl02000102032311 Last retrieved 10.01.10.
Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1991). The endowment effect, loss
aversion, and status quo bias. The Journal of Economic Perspectives, 5(1),
193e206.
Lindberg, K., Andersson, T. D., & Dellaert, B. G. C. (2001). Tourism development:
assessing social gains and losses. Annals of Tourism Research, 28(4), 1010e1030.
Lindberg, K., Dellaert, B. G. C., & Rassing, C. R. (1999). Resident tradeoff: a choice
modelling approach. Annals of Tourism Research, 26(3), 554e569.
Lindberg, K., & Johnson, R. L. (1997). Modelling resident attitudes toward tourism.
Annals of Tourism Research, 24(2), 402e424.
Louviere, J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods. Analysis
and applications. Cambridge: Cambridge University Press.
Morrison, M., & Bennett, J. (2004). Valuing New South Wales rivers for use in
benet transfer. Australian Journal of Agricultural and Resource Economics, 48,
591e611.
Oh, C., Draper, J., & Dixon, A. W. (2009). Assessing tourists multi-attribute preferences for public beach access. Coastal Management, 37, 119e135.
Petrzelka, P., Krannich, R. S., Brehm, J., & Trentelman, C. K. (2005). Rural tourism and
gendered nuances. Annals of Tourism Research, 32(4), 1121e1137.
Porcu, O. (2010). Lanalisi dei mercati. Lattuale congurazione della domanda: provenienze, consistenze e scelte dei turisti nel Nordovest. 2nd Report 2010. Sassari:
Sistema Turistico Locale Sardegna Nord Ovest S.c.ar.l..
Regione Sardegna. (2010). Piano Paesaggistico Regionale. Available at http://www.
regionesardegna.it/pianopaesaggistico/download.html Last retrieved 13.01.10.
Roca, J. (2003). Do individual preferences explain the environmental Kuznets curve.
Ecological Economics, 45, 3e10.
Stern, D. I., Common, M. S., & Barbier, E. B. (1996). Economic growth and environmental degradation: the environmental Kuznets curve and sustainable development. World Development, 24, 1151e1160.
Train, K. E. (2003). Discrete choice methods with simulation. Cambridge University
Press.
Vannini, M., & Puggioni, P. (2007). Sassari e il Turismo Piano strategico della citt di
Sassari. Sassari: Comune di Sassari, Available at http://www.comune.sassari.it/
sito_piano_strategico/piano_index.htm Last retrieved 04.09.010.
Vici, L., Castellani, M., & Figini, P. (2007). Le esternalit tra residenti e turisti. In
L. Punzo, & S. Usai (Eds.), Lestate al mare. Milano: McGraw-Hill.

You might also like