Combining multivariate analysis and cost analysis in outdoor recreation planning

Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Contents lists available at ScienceDirect

Journal of Outdoor Recreation and Tourism journal homepage: www.elsevier.com/locate/jort

Combining multivariate analysis and cost analysis in outdoor recreation planning Jeoffrey Dehez n, Sandrine Lyser 1 Irstea, Environnement Territoires et Infrastructures Research Unit, 50 Avenue de Verdun, 33612 Gazinet Cestas, France

art ic l e i nf o

a b s t r a c t

Article history: Received 15 January 2014 Received in revised form 16 September 2014 Accepted 16 September 2014

In this paper we examine the usefulness of combining multivariate analysis and costs analysis in recreation planning. Although these approaches have sometimes been developed in previous recreation studies, they have never been combined in this way. We apply this approach to a regional beach planning policy called the “Beach Plan”, in Aquitaine, south-western France. A multivariate procedure is used to assess the current environmental and social conditions of the 91 beaches included in the Plan. It reveals some connections between the variables we selected at the inventory step and leads to the definition of four homogeneous clusters of sites. We also identify possible social inequities. We find that the partition obtained by the cluster analysis does not coincide with the classification defined in the Plan. This confirms the necessity of an iterative process between inventory and implementation steps. We then examine the cost consequences of the implementation of the Plan. To do so, we define “total incremental cost functions” which enable us to evaluate the cost impacts of introducing sites in the Plan. We show that the application of the strict efficiency criterion may lead to several socially undesirable effects. We therefore propose an alternative integration path, called the “no social cost” path, after combining results provided by the multivariate analysis and the cost analysis.

Keywords: Outdoor recreation planning Beach recreation Multivariate analysis Cost analysis France

M a n a g e m en t i m p l i c a t i o n s


At the inventory stage of recreation planning, multivariate analyses provide a synoptic description of






recreational activities, out of a large quantity of data. Multiple correspondence analysis is a factorial method designed to deal with categorical variables The combination of physical and socioeconomic data is of great help to decision makers in defining new social objectives in recreational planning. By identifying possible correlations between the variables, multivariate analyses provide a better understanding of the functioning of sites. In the French case for instance, the level of use and the management effort are much more discriminating than the environmental characteristics of the beaches. Cost analysis aims at giving full account of the variability of management costs at the site level. Several drivers are identified accordingly. Cost analysis is also a means to compare several strategies for implementing the plan. In the present case, ranking the sites according to a “pure efficiency” criterion may be appealing because it allows for the introduction of 66 sites (out of the 91) into the planning process at a zero overall cost. The selection rule based on the efficiency criterion tends to increase social and economic inequalities between the municipalities. Another selection rule (hereafter called the “no social costs”) may reduce such undesirable effects. & 2014 Elsevier Ltd. All rights reserved.

n

Corresponding author. Tel.: þ 33 5 57 89 08 44; fax: þ 33 5 57 89 08 01. E-mail addresses: [email protected] (J. Dehez), [email protected] (S. Lyser). 1 Tel.: þ33 5 57 89 01 62; fax: þ33 5 57 89 08 01. http://dx.doi.org/10.1016/j.jort.2014.09.002 2213-0780/& 2014 Elsevier Ltd. All rights reserved.

1. Introduction Recreation planning is a process for ensuring the sustainable use and conservation of open spaces (Hammitt & Cole, 1998; Stein,

76

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

2013). In this respect, planning generally involves the task of defining specific areas where recreation is a key function (Stein, 2013). Depending on the anticipated activities and the environmental conditions, various uses can be assigned to these recreation sites, ranging from sports areas equipped with many facilities, to more natural land designed to meet very specific needs of nature-based tourists. As Pröbstl, Wirth, Elands, and Bell (2010) pointed out, approaches to recreation planning may be diverse. Compared to the North American countries, European managers seem to use more tailor-made solutions which are adapted to local conditions. Nevertheless, any planning process should include key steps such as an inventory of current (social and environmental) conditions, the definition of goals and objectives, and the preparation of an implementation stage (Hammitt & Cole, 1998; Pröbstl et al., 2010). Although a planning process usually starts with an inventory step, objectives are sometimes determined first before investigating the various ways of achieving them (as in Roig, Comas, Rodriguez-Perea, and Martin-Prieto (2005)). In any event, recreation planning clearly remains an integrated process that requires interdisciplinary research approaches. In this paper we would like to demonstrate the insights to be gained when combining multivariate analysis and cost analysis in recreation planning. More specifically, we use factor and cluster analysis. Both are data reduction techniques designed to provide a synoptic view of a large amount of data. Factor analysis is a method that reduces the number of variables (Lebart, Piron, & Morineau, 2006) while cluster analysis reduces the number of statistical units by a grouping of observations into homogeneous “groups” (Mirkin, 2005). These methods have sometimes been used for the classification of recreational users (Fyhri, Jacobsen, & Tommervik, 2009; Roca, Villares, & Ortego, 2009) or of the natural environment (Kaplunovsky, 2005). More recently, some authors have applied them in studies of recreational sites management, such as camping zones (Leung & Marion, 1999; Monz & Twardock, 2010) or beaches (Roig et al., 2005). As such cluster analyses are very promising for informing the “inventory step” of planning processes. Furthermore, they are complementary to the planning process of recreational zoning (as in Roig et al. (2005)) and the implementation phase would benefit from appropriate analytical tools that evaluate options. In biodiversity conservation, close attention has recently been paid to cost considerations, in order to optimize the use of existing resources and achieve more ambitious goals (Armsworth, Cantu-Salazar, Parnell, & Stoneman, 2011; Naidoo et al., 2006). In recreation planning, we found little if any scientific articles dedicated to this issue. Compared to conservation planning, economic research in recreation planning or land use planning seems to be much more interested in the evaluation of welfare impacts and non-monetary benefits, rather than costs (Cheshire & Sheppard, 2002; Hanley, Shaw, & Wright, 2003; Loomis & Walsh, 1997). Nevertheless, we think that cost analysis may also be of assistance, although the benefits are expressed in non-monetary terms. In this paper we combine multivariate and costs analyses to study the effects of a French recreational planning policy, the “Beach Plan” on the Aquitaine coastline in south-western France. To improve the recreational quality of the various beaches of Aquitaine, the Plan defines a typology of sites with new management objectives. However, until now no clear objectives have been formulated to support the typology, or the implementation of the Plan. The paper is structured as follows. First, we describe the study area. We then detail the data collection and the methodology for multivariate analysis and cost assessment. In Section 3 we present the results produced by both approaches as well as the implications for the implementation of the Plan. We discuss our results and conclude our arguments in Section 4.

2. Materials and models 2.1. Presentation of the study area and the “Beach Plan” The Aquitaine coastline is characterized by relatively limited urbanization (approximately 10% of the total shoreline) and the presence of large areas of natural land. The prevailing landscape consists of sandy beaches along its 230 km of shoreline, with rocks and cliffs in the southern part (approx. 30 km). A few kilometres inland, numerous freshwater lakes offer another type of beaches. Unsurprisingly, this coastline is highly attractive to tourists: in 2010, the coastal zones accounted for 40 million overnight stays, e.g. 42% of the total for Aquitaine (Conseil Régional du Tourisme d'Aquitaine, 2011). Tourism is however concentrated during a very limited period of the year (i.e. summer). The region experiences continuous population growth, concentrated around the biggest cities.2 As a consequence, human densities vary widely along the coastline (from 23 inhabitants/km2 in the least developed areas to 570 inhabitants/ km2 in the biggest coastal cities). This permanent population actually represents another major source of recreational demand. The current use pattern of the Aquitaine coastline is defined mostly by a long history of spatial planning, which started during the late 1960s when the inter-ministerial committee to develop Aquitaine's coast (Mission Interministérielle pour l'Aménagement de la Côte Aquitaine – MIACA) managed the recreational activities along the coast. Following its first “Beach Plan”, that committee funded most recreational facilities (parking lots, access, etc.). Since then, several public policies have followed (Daubet, Dehez, & Figura, 2010), dedicated to various aspects of beach management (organization of beach cleaning, monitoring erosion, installation of new facilities, etc.). In 2010 a new Beach Plan was launched by the local public authorities (Schéma Plan Plage Littoral Aquitain – Stratégie régionale, 2010). It was designed to (i) assess actual conditions of the coastline, (ii) devise new management standards to improve outdoor recreational facilities, and (iii) set objectives for the distribution of such standards along the coastline. Of the 170 beaches identified in Aquitaine (including oceanside and lakeside beaches), 91 were included in the Plan (Fig. 1). These 91 beaches were selected for their natural features (i.e. urban beaches were excluded) and the guarantee of public access (thus excluding private camps). The definition of management standards led to a classification of four types of beaches:


“Extended recreation” (ER) beaches (31 sites) offer a variety










a facilities and services. They are located in the immediate vicinity of urban areas and management practices are determined accordingly. As far as possible, coordination is sought with services provided by urban centres (e.g. transport). Recreation is the main objective. A high use rate is therefore expected. “Recreation and nature” (RN) beaches (11 sites) may not be situated in the vicinity of urban areas but rather in a natural environment. Fewer services are therefore available, compared to “Extended recreation” beaches. High use is still expected, requiring a high standard of facilities and services. “Natural” (N) beaches (27 sites), situated in a natural environment, are supposed to meet the demand for nature-based tourism. Few facilities are offered and particular attention is paid to environmental quality. Similar to others beaches, the safety of individuals must be ensured. A lower level of recreational use is expected. “Lakeside” (L) beaches (22 sites) are situated inland. The environmental characteristics of lakes and the types of

2

Source: French National Census.

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

77

Fig. 1. Location of the 91 study sites.

recreational activities pursued there have been used as the main arguments to separate these beaches into a distinct type. Lakes are supposed to be defined as a more homogeneous class (for a given level of use, the same services are provided). The level of use is supposed to remain low.

Hence, the typology is based on a tacit gradient of “wilderness” quality and the desirable level of use. Environmental features (i.e. the distinction between oceans and lakes) also serve as a discriminating factor. According to the authorities, this typology offers a “wide variety of recreational services while protecting

78

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

large tracts of land”. As a consequence, “(…) municipalities situated in the vicinity of these protected areas will fully benefit from nature-based tourism development” (Schéma Plan Plage Littoral Aquitain – Stratégie régionale, 2010). Nevertheless, the definition of the previously mentioned “variety of service” is based on very general assumptions about the nature of recreational demand, i.e. some users are looking for “touristic services” while others are looking for more “nature”. In the Plan, each of the 91 beaches is assigned to one of these new types. Furthermore, such a distribution among the categories remains a long-term objective and the Plan does not explain any details about its implementation. 2.2. Data collection and processing 2.2.1. Multivariate analysis As Leung and Marion (1999) suggest, multivariate approaches applied to recreational sites can facilitate a better understanding of the structure of the descriptive data, and hence of the functioning of these sites, by identifying possible correlations between the monitored variables. In addition, the implementation of cluster analysis (CA) affords a synoptic view of the cases by identifying a limited number of homogeneous groups of sites out of a large quantity of data. In this paper, we adapted a conventional strategy similar to that of Leung and Marion (1999). We use factor and cluster analyses in a complementary fashion (this procedure is sometimes referred to as “tandem analysis”):  the multiple correspondence analysis (MCA) is a factorial method that provides an exploration of the similarity relationships among categorical variables, by transforming them into continuous variables;  scores of individuals derived from MCA are then used to perform an ascendant hierarchical clustering algorithm (AHC). This method contributes to a better understanding of the relationships among the observations. Most of the examples in the literature on the clustering of natural sites use principal component analysis (PCA) (Leung & Marion, 1999; Monz & Twardock, 2010; Roig et al., 2005). In the cited examples, the variables selected to describe the sites are quantitative indicators (e.g. ground vegetation coverage, number of tree stumps). As our case study deals with categorical explanatory variables, PCA is not appropriate here, whereas MCA is.3 To describe the 91 relevant sites, a set of 27 variables was selected based on discussions with managers and an extensive literature research. One of the main objectives was to characterize the various dimensions of the recreational activities and the natural environment. For every site, the variables were assessed with the assistance of experts and local managers in 2010 along the following six themes: environment of the site, natural hazards, physical alterations and environmental management, level of use, facilities and service, and access conditions. As shown in Table 1, the number of categories varied from 2 (e.g. Type) to 6 (e.g. Ground movements). A preliminary study with the correlation ratio helped us to identify those variables that were strongly linked to each other. MCA is thus applied with 13 “active” variables. We add the 14 remaining variables as supplementary information in the MCA. As mentioned above, the MCA is completed with an AHC algorithm with Ward criterion, to cluster the 91 beaches. 2.2.2. Cost analysis Our cost analysis was based on a distinct strategy. Following Naidoo et al. (2006) and Armsworth et al. (2011), it aims to reflect 3 Nevertheless, MCA shares strong similarities with Principal Correspondence Analysis (PCA). As a consequence, many proprieties and criteria which are actually used in PCA are also used in MCA (Abdi & Valentin, 2007; Abdi & Williams, 2010).

the variability of the costs at the site level, which seems to be an unusual type of analysis in recreation research.4 In this paper we focus on direct costs. Acquisition costs and opportunity cost are considered to be zero, based on the fact that the sites were originally designated for recreation, and that no significant change in this use will occur. In the absence of dedicated databases, we developed a specific empirical method. Cost data were obtained by directly interviewing managers who were generally employed by municipalities or public agencies. These managers were able to estimate the management costs of the natural areas in a way that was suitable for cost analysis (see Escobedo et al. (2008) for instance). To limit selection bias, we selected a large variety of sites and considered a four-year time period. All figures were converted into equivalents of the 2010 Euro. The investigation period covered the entire year of 2010. A questionnaire was left with the managers and data were checked with the qualified persons afterwards. Whenever possible, we compared the costs of activities with the few studies available in other parts of the country. Expenses covered operational costs as well as investment costs. All cost estimates were expressed on an annual basis. Capital costs were annualized with the usual formula.5 Three main operations were considered for the cost analysis: bathing supervision, beach clean-up and the management of natural areas (dune forests) and recreational facilities. Explanatory variables were selected to reflect the variability of management costs across the various beaches. As far as possible, these variables were selected from the set of variables used in previous multivariate analyses but specific additional variables were also introduced. On this basis, a total of 42 values for annual management costs were estimated at the site level (Tables 2–5). The annual management cost of a given site is estimated by adding the costs of the three previously mentioned operations (including capital costs if necessary). Tables 2–5 show that each component of the direct costs has specific drivers. The cost of beach supervision contains more than 80% of expenses for manpower. They are related to the level of use and the type of beach (i.e. ocean or lake). Supervision costs are lower for lakes than for ocean beaches. As all these facilities are removed during off seasons and many are rented on an annual basis, no capital costs are incurred. In comparison, beach clean-up costs do not depend exclusively on the level of use. Each département has its own organization and uses its own techniques for cleaning, which strongly impacts the costs. In the Landes département for instance, mechanical cleaning is used intensively, while in the Pyrénées Atlantique département manual operations are the most prevalent method. The Gironde area appears to be a mixed case. These institutional characteristics add one additional component to our cost analysis. At least half of the beach cleaning operating costs relate to manpower. When manual collecting is applied on beaches with low levels of use, the cost for manpower can account for up to 90% of the total annual costs. On the other hand, capital costs for beach cleaning remain low, irrespective of the technique. As mentioned earlier, the management of natural areas and recreational facilities is the third operation to be considered in our analysis. Related operating costs have proved to be somewhat more complex and relate to the level of use, the type of site (i.e. ocean or lake), the natural surroundings, and the variety of the services offered. “Natural surroundings” relate to the

4 In the economics literature, cost assessment is rarely a priority and is usually just a small part of cost-benefit analysis. Thus, cost estimates are often limited to rough values referring to large programmes (see Chen and Jim (2008) or Montes et al. (2011) for instance). Noteworthy exceptions are given in the manual of Loomis and Walsh (1997), although the references appear quite old today. 5 A ¼ Iðr=ð1 ð1 þ rÞ  L ÞÞ, where I is the initial investment outlay, r is the interest rate (3%), L the lifetime of the investment and A the annualized value.

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

79

Table 1 Variables selected for the beach analysis. Variables

Active variable

Measurement scale

Value

3 3 2 4 4

3 class level

Gironde; Landes, Pyrénées Atlantiques Natural; tourist facilities; semi-urban Ocean; lake o 22 Ha; [22–37 [Ha; [37–75 [Ha; 475 Ha o €855,000, [€855,000–€1,030,000[; [€1,030,000–€1,776,500[; 4€1,776,500 None; low; intermediate; high None, low, intermediate, high None; rockfalls low; rockfalls high; landslide low; landslide intermediate; landslide high Low; intermediate; high

3 class level

None; intermediate; high

(LAND)

5 level scale

None (0); ineffective (1); small (2); intermediate (3); large (4)

(FACIL) (PKG) (OVER) (SAFE)

3 3 3 3 4 3

Poor (0); intermediate (1); good (2) Poor (0); intermediate (1); good (2) None (0); intermediate (1); high (2) Poor (0); intermediate (1); good (2) None; low; intermediate; high Low (0); intermediate (1); high (2)

Départementa Natural surroundings Type Site area (SITE) Municipal annual budget (/beach) Marine erosion Wind erosion Ground movements

Natural value of surroundings and administrative zoning

Natural hazards

Physical alteration and environmental management

Level of use

Facilities and services

General condition of the site Signs of physical alterations Land reservation-access restriction Adequacy of facilities Adequacy of parking size Overcrowding Safety Fire risks Number of visits (summer) Density of population Touristic area Parking lot size (vehicles) Access to the beach (number) Picnic areas Shops Others services Cars Bicycles Walkers

Access conditions

(VISIT)

zones types types level scale level scale

4 level scale 4 level scale 6 values

level level level level level level

scale scale scale scale scale scale

4 zones 4 zones

(PICN) (SHOP) (SERV) (CAR) (BIKE) (WALK)

5 level scale 4 level scale

Medoc; Bordeaux & Arcachon Bay; Landes; Basque country Medoc coastline; Arcachon Bay; Landes coastline; Basque coastline r 50;] 50–200]; ]200–500]; ]500–1000]; 4 1000 0; 1; 2; Z 3

2 4 2 3 4 3

No (0); yes (1) 0; 1; 2-3; 43 No (0); yes (1) Poor (0); intermediate (1); good (2) None (0); poor (1); intermediate (2); good (3) Poor (0); intermediate (1); good (2)

level level level level level level

scale scale scale scale scale scale

a A French départment is an administrative unit comprise between municipalitiy and Aquitaine region levels. It refers to the NUTS Level III of the EU legal statistical framework.

Table 2 Annual costs of beach supervision per beach (operating costs, in 2010 Euros). Level of use

Ocean

Lake

High Medium Low

€90,000 €65,000 €20,000

N/A €20,000 €20,000

Table 3 Annual costs of beach clean-up per beach by administrative département (operating and capital costs, in 2010 Euros). Level of use Administrative département & clean-up technique

High Medium Low

Gironde

Landes

Mechanical and manual Annual costs

Mainly Mechanical Annual costs

€87,500 €27,500 €27,500

€97,500 €97,500 €67,500

Pyrénées Atlantiques Manual Annual costs N/A €67,500 €32,500

Table 4 Annual costs of the management of natural areas and recreational facilities under adequate maintenance per beach (operating costs, in 2010 Euros). Level of use

Natural surroundings and variety of services Natural

Wide variety of services High €160,000 Medium €85,000 Low €35,000

Low variety of services

Wide variety of services

Low variety of services

€130,000 €67,500 €20,000

€80,000 €40,000 €30,000

€65,000 €30,000 €20,000

Lake

N/A €35,000 €20,000

Table 5 Annual costs of the management of natural areas and recreational facilities per beach (investment and capital costs, in 2010 Euros). Level of use Initial outlay

Ocean

same variable as the one used in multivariable analysis, whereas the “variety of services” relates to the variables picnic areas, shops and other services. The variety was considered to be “wide” if at least two out of the three services were observed on the site.

Tourist facilities and semi-urban

High Medium Low a b

Lake

Inadequate maintenancea Capital costs

Adequate maintenanceb Capital costs

Ocean

Lake

Ocean

Lake

N/A €13,000 €2,000

€48,000 €15,000 €2,000

N/A €6,000 €1,000

€1,600,000 N/A €108,000 €500,000 €200,000 €34,000 €60,000 €30,000 €4,000

Inadequate maintenance is associated with a 20-year lifetime of facilities. Adequate maintenance is associated with an infinite lifetime of facilities.

80

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Another important cost variable relates to management efforts. On many beaches, management inadequacies have been observed to lead to environmental damage or a deterioration of recreational facilities. On beaches where management has been judged appropriate, the cost variation is approximated to be, ceteris paribus, 20% higher on average. Finally, capital costs also appear to be much higher. They are impacted by inadequate maintenance that actually reduces infrastructure lifetime and thus raises annual costs (Table 5). Capital costs have also been related to the level of use and the type of beach (ocean or lake). For instance, large parking lots are typically associated with the most heavily used sites. Once again, the annual costs for ocean beaches appear to be higher than the annual costs for lake beaches.

3. Results 3.1. Multivariate analysis \The MCA identifies 28 factors, the first two of which (hereafter called F1 and F2) account for 18.86% of total inertia (i.e. variance).6 This percentage is quite low but not problematic: traditionally, the percentages of inertia with this approach are severely underestimated because of the MCA algorithm. The definition of Factors F1 and F2 calls for interesting interpretations (Appendix A and Fig. 2). Factor F1 can be interpreted simultaneously on the basis of a group of variables reflecting the overall recreational quality of the beach (i.e. access by cars, adequacy of facilities, safety and shops) and the “level of use” of the site (i.e. number of visits variable). On the one end, we find the beaches that have a variety of services, big facilities and parking lots, and a good level of security. These beaches frequently have high levels of use. On the other end, we find beaches with few, basically inadequate services and facilities. The level of use for these beaches is often low. Given the fact that this analysis reveals relationships between variables only, it is impossible to conclude whether managers have concentrated efforts on the attractive and heavily used sites (i.e. “supply” has followed “demand”) or whether users have been attracted by high-standard sites (i.e. “demand” has followed “supply”). Factor F2 summarizes land reservation and site area variables. Small beaches tend to have smaller tracts of natural areas and consequently a low level of land reservation, and vice-versa. This particular aspect of environmental management appears to be a critical point.7 The ascendant hierarchical clustering method is applied to the first 12 factors of the MCA (which measure 71% of total variance).8 Both the analysis of the cluster tree and the histogram of the level indices of the AHC suggest four clusters of beaches (Fig. 3). The 24 beaches of Cluster 1 (i.e. 26% of the sample) offer large parking lots, large natural spaces, “light” facilities and few pedestrian accesses (Appendix B). Fire prevention is maximal (fire risks are generally low due to the importance of fire fighting). Most of these sites are located along the northern part of the coast (e.g. Gironde département) where the coastline is mainly composed of vast sandy beaches close to areas with high population densities but still at a distance from big cities. As a consequence, the level of use is intermediate or high. Despite this high level of use, the majority of indicators referring to the adequacy of management obtain good values. Very few lakes belong to this cluster. Cluster 2 (30 beaches; 33% of the sample) are much more closely connected with coastal 6

It rises to 26.15% if the third factor is included. Another reason may be that land reservation is a zoning practice aimed at preventing the intrusion of users. For technical as well as social reasons (e.g. social acceptability or political will) it is not necessarily applied everywhere. As such, it can be seen as discriminating. 8 The analogue of the Kaiser criterion used in PCA is to retain only the axes whose Eigenvalues are greater than 1/p (where p is the number of active variables). In our case 12 eigenvalues are greater than 1/13 (i.e. 0.077) 7

cities than beaches in Cluster 1. Here again, population density and the level of touristic activities are very high. Most of them are located in the south, where the natural environment consists mainly of cliffs and rocks. As a consequence, risks also vary (compared to Cluster 1, landslides are more frequent than fire risks). Natural spaces and parking lots are smaller than those of beaches in Cluster 1. Here, land is not as available as in the north, and cities already provide parking facilities. Like beaches in Cluster 1, the management effort also seems appropriate, and some lakes also feature in this group. In Cluster 3 (21 sites; 23% of the sample), the beaches show some management deficiencies compared to the earlier clusters. Problems may be associated with the maintenance of the facilities (parking size and safety variables, for instance), and environmental management (land reservation variable, in particular). These sites are generally small, with a low or medium level of use (respectively 38.10% and 61.90% of the cluster). They are frequently located in the Landes département (intermediate position between the north and the south of the coast). In this area, tourism is the main use of the beaches (this is a low populationdensity area) with significant seasonal variations (peak periods in summer). A large proportion of the beaches are on lakes, although not all the lakes are in this class (8 lakes out of the 22, see Table 6). Cluster 4 contains 16 beaches (i.e. 18% of the sample). Every beach included in Cluster 4 has some management problems. More than 60% of them have the lowest level of security. They offer few facilities and their use levels are low, but the natural areas are often large. In this cluster most land reservation practices are applied. These beaches are scattered all along the coastline, mainly at some distance from the coastal cities. Finally, the projection of the partition into 4 clusters and the levels of variables on the first factorial plane, defined by the axis F1 and F2, is illustrated in Fig. 2. Cluster 1 and 4 sites are distinguished predominantly along the axis F1. The north-west quadrant contains the sites of Cluster 1 with the highest recreational quality and most intensive use, whereas the sites of Cluster 4 are positioned far on the right side of the north-east quadrant. The sites of Cluster 2 and Cluster 3 reveal a much lower discrimination along F1. Clusters are also distinguished along the F2 axis: at the bottom, Clusters 2 and 3 are characterized by smaller tracts of natural areas and few land reservation practices, whereas Clusters 1 and 4 sites are located at the top. In Table 6 we compare both distributions of beaches according to the Plan (rows) and our cluster analysis (columns). As one can see, the two distributions do not coincide perfectly. A Chi square test confirms that the distributions are not correlated at the 1% level (Chi-square¼ 16.97, df¼ 9, p-value ¼0.049)9. In our opinion, it illustrates the difference between an assessment provided by a sound statistical analysis based on the one hand and an arbitrary classification motivated by political decisions on the other hand. Thus, the results of the multivariate analysis show that the earlier classification of the Plan does not seem be based on groups of sites characterized by homogeneous functioning, as measured by our variables. The first major difference appears in the absence of a dedicated class for lake beaches in the cluster analysis. The multivariate analysis shows that the lake beaches are not that different from seashore beaches, contradicting a main assumption of the typology in the Plan. The second difference concerns the influence of the management adequacy, which was not included in the typology of the Plan because the authorities automatically strive for the highest possible quality standard. In practice, management efforts (whatever aspect is examined, e.g. environmental or recreational quality) differ substantially between the

9 Although the distributions seem to be correlated at the 5% level, we think that the p-value of 0.049 indicates that the relationship would not be very clear. This issue is return to below.

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

81

Fig. 2. Two-dimensional plot of the levels of categorical variables and the centroids of the four beach clusters. The size of the symbols indicates their quality of representation as measured by the squared cosines (see Appendix A).

Fig. 3. Cluster tree of the beaches based on the Ward method.

Table 6 Distribution of the beaches according to the Beach Plan typology and the cluster analysis. Cluster analysis

Cluster Cluster Cluster Cluster Total

1 2 3 4

New Beach Plan typology Extended recreation

Recreation and nature

Nature Lakeside Total

4 14 8 5 31

5 3 2 1 11

12 5 3 7 27

3 8 8 3 22

24 30 21 16 91

sites. The third difference relates to the environment of the beach. In the cluster analysis, the environment of the site is not always a discriminating factor (unlike the above “gradient” of “wilderness” depicted in Section 2.1). As noted above, beaches in Cluster 1 are often located in the north, while beaches in Cluster 2 are located in

the south. Beaches in Cluster 3 are generally located on rural lands. Nevertheless, some connections between the two typologies can be observed. Table 6 illustrates this point. First, it looks like the beaches in Cluster 1 and Cluster 4 may be associated mostly with the “Nature” type of the Plan (19 sites out of the 27). This could be seen as a way to preserve some valuable existing natural characteristics. Because beaches in Cluster 1 have good management practices, the shift from one class to the other may not be problematic. On the other hand, in order to attain the objectives of “Nature” sites, changes in the management of beaches in Cluster 4 imply significant improvements which may not be that easy to implement. About 50% of the beaches in Cluster 2 coincide with the new “Extended recreation” type. In the cluster analysis, Cluster 2 sites already have appropriate facilities as required by the new “Extended Recreation” class. For this reason, the move from one class to the other may not be problematic either. However, the 8 beaches of Cluster 3 (out of the 21) have been identified as possible candidates for the “Extended recreation” type. Here, significant changes (regarding safety for instance) are needed. Overall, despite these similarities, the question of the

82

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Table 7 Annual management costs of the beaches in Aquitaine (in 2010 Euros). Annual management costs

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Combined a

Min

Max

Meana

Totala

€106,000 €41,000 €38,000 €0 €0

€385,500 €315,500 €250,500 €97,500 €385,500

€211,000 €147,000 €137,000 €47,000 €144,000

€5,054,000 €4,397,000 €2,879,000 €758,000 €13,087,000

The figures are rounded to the nearest thousand.

Table 8 Municipal budgets per beach (in 2010).

beaches in Cluster 3 and 4 belong to municipalities with the lowest revenues. At the very end of the list, beaches in Cluster 4 are, on average, located in the municipalities with the lowest budgets. These are the beaches with the most severe management deficiencies. In this sense, management deficiency may be connected with the municipality's ability to provide financing. The comparison of Tables 7 and 8 provides additional details concerning the sites which are supposed to be managed appropriately. The beaches in Cluster 1 (with the highest management costs of €1,851,000) are not necessarily found in municipalities with the most resources. In other words, the municipalities which are responsible for the management of the most expensive sites are not those with the highest level of resources. Hence, exogenous cost factors are still of importance. These points raise some equity issues that will be of importance during the implementation phase of the Plan.

Municipal budgets per beach in Euros

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Combined

Minimum

Maximum

Mean

€425,200 €295,500 €425,200 €317,500 €295,500

€7,296,000 €9,244,000 €7,296,000 €1,776,500 €9,244,000

€1,851,000 €2,010,000 €1,572,000 €1,067,000 €1,701,000

implementation of the Plan is not solved. In a world of scarce resources, some strategies may be better than others. Therefore cost analysis matters. 3.2. The current management costs of the beaches Given the cost estimates, an approximation of the current annual management costs of each of the 91 beaches can now be made. Results are grouped together in Table 7 according to the four classes produced by the cluster analysis. Currently, the beaches in Cluster 1 accrue the highest management costs of the sample, with a mean of €211,000 per beach per year. In comparison, the mean value for the beaches in Cluster 2 is estimated at €147,000 per year. Such a difference illustrates the greater complexity of managing environmental resources compared to the effort required for the maintenance of the recreational facilities. The lower costs of beaches in Clusters 3 and 4 reflect their smaller size as well as the management deficiencies observed in the sample. Note that the lowest costs are found to be €0 for beaches without any management. In contrast, the highest costs of the sample (€385,500) are found in the first class. Once again, this illustrates the challenge of managing recreational natural areas while dealing with a high level of use. The hierarchy between the four classes remains the same for both the minimum or the maximum values of costs, which confirms the homogeneity within the clusters of our cluster analysis. The total annual cost of the whole sample is evaluated at €13,087,000. During the collection of the data, the idea that the costs may also be connected with the managers' “capacity to finance” emerged. Although no clear correlation emerged, Table 8 contains some trends into that direction. In Table 8, the financing capacities of the municipalities are approximated by the municipality's total annual budget divided by the total number of beaches (included or excluded from the sample) under its responsibility. Here, the beaches with the highest management costs (i.e. Cluster 1 and 2) are frequently located in the municipalities which have the largest financial resources. These beaches have few or no management deficiencies. In contrast,

3.3. The cost of implementing the Beach Plan 3.3.1. The incremental costs of including a beach in the Plan The costs of including a beach in the Plan are defined as the difference between the estimated management costs obtained through the application of the new standards, and the costs of current management (as estimated previously). For each beach i of the sample, they are called the “incremental costs” (hereafter IC) and are denoted as IC i ¼ C i2 ðY Þ  C i1 ðY Þ where Ci2 are the simulated management costs based on the new standards, Ci1 the costs of current management and Y the vector of cost variables. By definition, Ci2 could not be observed and remains hypothetical (as the new management practices are still to be applied). To approximate Ci2, we based our estimates on the new quality standards which are described in the directives of the Plan (Table 9).

Table 9 New standards for beach management defined in the Beach Plan. New standards/ operations

Typology of the Beach Plan Extended recreation

Beach supervision Beach clean up

Yes Mainly mechanical Services High Management effort Complete

Recreation and nature

Nature

Lakeside

Yes Mechanical and manual High Complete

Yes Manual

Yes N/A

Low N/A Complete Complete

Table 10 Distribution of incremental costs regarding both typologies (mean value, in 2010 Euros)a. Cluster analysis

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Combined a

Typology of the Beach Plan Extended recreation

Recreation and nature

Nature

þ €15,000 þ€23,000 þ €29,500 þ €90,500 þ €34,500

 €18,000  €13,000  €73,000 þ€120,000  €14,000

€0 þ €21,500 þ €9,500 €0  €12,000 þ€7,000 þ€2,500  €17,000 þ€3,000 þ €41,000 þ €61,000 þ €70,000 þ€6,500 þ€21,000 þ €18,000

Some figures are rounded to the nearest thousand.

Lakeside

Combined

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

83

Fig. 4. Annual incremental costs by beach (in 2010 Euros).

Every time previous directives refer to cost variables, values have been modified accordingly. Others factors remain constant. Table 10 provides an overview of the distribution for IC values. The IC values depend on the current management practices (as defined in the CA) and on the objectives of the Plan. On average, beaches in Cluster 3 have the lowest incremental costs. Including a beach of this particular class into the Plan will cost €3000 on average. The costs rise to €29,500 when the site has been selected as a future “Extended recreation” beach. On the other hand, beaches in Cluster 4 demand the highest costs (e.g. €70,000 on average). In many cases, ICs are negative, which illustrates a decrease in annual management costs. Such cost decreases come from two sources. First, the new directives may push towards less intensive techniques, which in turn lead to a decrease in costs (for instance mechanical cleaning proves not to be the most efficient strategy everywhere). Second, cost decreases may also reflect a decrease in capital costs. In many places, the rise in operational costs is more than offset by the decrease in capital costs. The greatest decrease concerns two beaches in Cluster 3 (when “transformed” to “Extended recreation” sites) and was estimated to be €73,000 per year. Fig. 4 provides a more complete view of the distribution of the estimated ICs, sorted in increasing order. For 18 beaches, the incremental costs are negative. These beaches belong to Cluster 1 (5 sites), Cluster 2 (5 sites) and Cluster 3 (8 sites). The incremental costs are zero in 21 cases and positive for the 52 others. Zero ICs are mainly associated with beaches belonging to Cluster 2 (14 sites) and Cluster 1 (6 sites). In the latter, the increase in total annual operating costs equals the decrease in annual capital costs. The highest cost increases occur with beaches in Cluster 4 (four of the 10 biggest increases). In these cases, the highest costs are associated with the cost of developing beach supervision (no supervision is currently applied). The greatest cost increase is estimated at €165,000 per year. In the latter, the new management practices include beach supervision and the installation of new facilities as well as the restoration and the management of natural areas.

3.3.2. The total costs of implementing the Beach Plan As noted above, the Plan does not contain a clear implementation strategy. Considering the fact that such important changes

take time and will cost money, defining rules for site selection may be useful. In this section we look at the effects of various implementation strategies on costs. However, the definition of the objectives remains a key issue. Depending on the situation, objectives may take various forms, e.g. the global size of an area open to the public, the number of users affected by recreational policies, or the length of new trails. In the present case, it is given by the new Plan directives, i.e. the applications of the new recreation standards on the 91 selected sites. The “recreational site” is commonly referred to as the decision unit by decision makers,10 and we assume it is appropriately applied at a regional level. Nevertheless, we agree that it suffers from severe limits because the sites may differ substantially, causing the comparison of the two ICs to be biased. Therefore we compare several planning processes in this paper. The aggregation of the previous IC values leads to the definition of a total incremental cost curve for the implementation of the Plan. This curve illustrates the financial consequence of the implementation of the Plan on a step by step basis. At the same time, it develops an implicit “hierarchy” of the sites (Fig. 5). Now, the total cost function11 first starts with a decreasing phase which corresponds to the beaches whose incremental costs are negative. In other words, implementing the Plan on the basis of this principle would save money during the first stage. The aggregate costs are then constant and rise again after the introduction of the 40th beach. The costs become positive for the 67th beach. From a constant-cost perspective, this means that 66 beaches could be introduced in the Plan at a “zero” overall cost. Such an approach may be appealing for decision makers and illustrates the full “power” of the efficiency criterion. However, it tends to underestimate some important social effects. First, the previous selection rule leads to many beaches with safety or environmental problems being excluded (i.e. beaches in Cluster 4 mainly) at the end of the aggregation process. Although these management deficiencies have no financial outlays, they may have 10 Moreover, the new Plan is defined not in terms of lengths of beaches or acreage open to the public, but instead in terms of sites and number of beaches offered. 11 Considering the fact that the ICs are sorted by increasing order, this first cost function may be seen as a standard micro-economic curve (Varian, 1992).

84

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Fig. 5. The total incremental costs of implementing the Beach Plan following the efficiency criterion (in 2010 Euros).

Fig. 6. The total incremental costs of implementing the Beach Plan following the “no social costs” criterion (in 2010 Euros).

other impacts whose full value can be estimated by specific economic valuation methods. These may be referred to as congestion or environmental costs. Since the pioneering work of Fisher and Krutilla (1972), we know that such social costs must be included in economic analysis and methodology to give them monetary value. They have constantly been developed ever since (Bockstael & McConnell, 2007; Hanley et al., 2003). Second, the aggregation process described in Fig. 5 pays little attention to equity issues. Many of the sites which appear at the end of the total costs curve are located in municipalities with fewer financial resources (see above). In addition, these sites also tend to be located in particular areas (Landes département) where tourism is less developed. In this sense, inequalities also tend to cumulate in

the recreational sector. In order to take account of such social effects, another aggregation path is examined. We call it the “no social costs” curve. The new total incremental costs curve is built after having ranked beaches in Cluster 4 (with safety and environmental problems) first, by order of increasing incremental costs, and beaches in Cluster 3 second, also by order of increasing incremental costs. The same principle is used for the 54 remaining sites, regardless of their cluster. The resulting cost function exhibits a rather different shape compared to the previous one (Fig. 6). First, the total incremental costs are now positive from the beginning and they strongly increase until the introduction of Cluster 3 beaches. Because the incremental costs of these beaches

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

are frequently lower than the costs of Cluster 4 beaches, the total costs decrease in a first step before the most expensive ones are included in the Plan. After having aggregated the incremental costs of every beach in Cluster 4 and 3, the introduction of the remaining ones (i.e. sites of Cluster 1 and 2) induces another cost decrease. At this stage, the total cost decreases even more than it did in the previous phase. Regarding budget, a series of growth and decline phases must be anticipated. The end of the curve shows an increasing trend again. In this model, efficiency is not rejected but it is balanced with other objectives.

4. Discussions and conclusion In this article we set out to demonstrate the usefulness of combining multivariate and cost analyses in recreation planning. This combination dovetails with the need for developing recreation research from a far more interdisciplinary perspective (Haider, 2006). To some degree, it also offers an example of a “framework for making the trade-offs inherent to the management issue on a conceptual basis” (Manning, 2004), by integrating issues such as equity, efficiency, environmental damages or recreation quality. We apply it to the case of a regional beach planning policy, namely the Beach Plan in Aquitaine (Schéma Plan Plage Littoral Aquitain – Stratégie régionale, 2010). Multivariate analysis in the form of a combined multiple correspondence analysis (MCA) and cluster analysis (CA) provided useful insights into the inventory of ecological and social conditions. In this paper, we use multiple correspondence analysis instead of principal components because we deal with categorical variables. As far as we know, such an application is rare in the outdoor recreation literature and thus constitutes another added value of the paper. Various aspects of recreational management were examined: natural value of surroundings and administrative zoning, natural hazards, physical alterations, the level of use, facilities and services, and the quality of access. MCA identified relationships between the “size” of the site and the management effort (both illustrated by several variables). The subsequent cluster analysis identified four homogeneous clusters of sites which did not coincide perfectly with the classification of the Plan. This lead us to question its consistency. It also confirmed that outdoor planning is necessarily an iterative process, with a feedback loop between inventory and objectives (Hammitt & Cole, 1998; Pröbstl et al., 2010). Another result provided by the multivariate analysis lies in the identification of possible social and economic inequalities, depending on the managers' funding capacities, as well as the spatial distribution of the sites along the coast. Finally, an ambitious analysis of the costs of recreation for the 91 beaches of the sample lead to the estimation of 42 values explained by 6 cost drivers (level of use, type of site, administrative département, natural surroundings, services provided, and management efforts) which fully illustrated the variability of management costs at the site level. A dedicated methodology for the cost analysis is developed, given the various empirical constraints. Although the final total costs of the Plan will remain constant, the process of implementation is far from neutral. Following this idea, we showed that, despite its relatively appealing properties (i.e. the introduction of 66 sites at a “zero overall cost”), the strategy based on a “pure” efficiency criterion may have undesirable social impacts. We therefore offered another

85

option, the “no social costs” aggregation process, to try to alleviate some of these. In our opinion, this paper makes several contributions to the decision process of recreation planning. In the case of the Beach Plan, it provides a better understanding of the current management practices as well as a more extensive view of the effects of its implementation. Based on our results, decision makers can support one of the two “aggregation paths” depending on the chosen priority. The “no social costs” strategy can be fulfilled by subsidizing some conversions of beaches or by operating financial transfers between the municipalities (for example, by reallocating subsidies from managers who “save” money to those who face increases in management costs). It also stresses the necessity for a closer coordination among the various stakeholders, i.e. the municipalities and the organisms in charge of the planning process. Multivariate analysis can be applied in other recreational planning processes, especially on large scales. Diverse use patterns or natural environments (e.g. beaches, mountains, urban, etc.) can also be analyzed with multivariate analysis. Furthermore, the introduction of economic criteria offers a better insight about the social consequences of recreational planning. This is consistent with the need for greater awareness of equity and “social justice” in tourism, as emphasized by Bramwell and Lane (2008). Multivariate methods may also contribute to the widely known the widely known frameworks such as the LAC or ROS (Haider, 2006; Stein, 2013) which do not necessarily account for these planning and management concerns. In addition to our promising results in this case, several improvements may be expected in the future. First, additional research may be undertaken on the nature of the relationships between the variables. Multivariate analysis is fundamentally based on the study of the correlational ratios and provides no information on causality. Such improvements would definitely enrich the inventory stage. Second, the costs of the new management practices are still hypothetical and the recent local implementation of the Beach Plan has to be carefully examined to check whether our estimates are correct or not. Third, one could also base the cost analysis on other economic ratios. In this paper we used what we called the incremental costs per site as the basic unit, but other indicators such as the costs per visit could also be appropriate. In the absence of a robust estimate of the number of visits for the whole sample, we were not able to calculate such an indicator. Nevertheless, we posit that the use of the cost per visit might strengthen our results because the sites with the highest ICs (i.e. Cluster 3 and 4) frequently have the lowest levels of use (as demonstrated in Section 3.1). However, this has still to be proven. Some kinds of recreational benefits may also be introduced. Beaches in Cluster 1 or 2 have strong advantages because of their current high level of use (which implies a high level of aggregated benefits), although the “wilderness” of beaches in Classes 3 and 4 could generate large individual benefits for specific visitors interested in naturebased tourism (Tyrväinen, Buchecker, Degenhardt, & Vuletic, 2009). A distinctive analysis of the structure of users' preferences should definitely be of great help. Lastly, our study also calls for a sound monitoring strategy in recreation planning. In addition to data on environmental conditions and recreational equipment, planners need information on users, institutions, managers' practices and economic drivers. Producing such information will undoubtedly require enhanced interdisciplinary collaboration (Sievänen et al., 2008).

86

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Appendix A. Factor scores (coordinates), contributions and squared cosines of the 42 categories on the first two factors of MCA

Variables/modalities

Access bicycles conditions BIKE_0 BIKE_1 BIKE_2 BIKE_3 Land reservation-access restriction LAND_0 LAND_1 LAND_2 LAND_3 LAND_4 Adequacy of facilities FACIL_0 FACIL_1 FACIL_2 Safety SAFE_0 SAFE_1 SAFE_2 Others services SERV_0 SERV_1 Picnic areas PICN_0 PICN_1 Site area SITE_ o22Ha SITE_22-37Ha SITE_37-75Ha SITE_ 475 Ha Shops SHOP_0 SHOP_1 SHOP_2 SHOP_ 43 Number of visits VISIT_0 VISIT_1 VISIT_2 Overcrowding OVER_0 OVER_2 Access cars conditions CAR_0 CAR_1 CAR_2 Adequacy of parking size PKG_0 PKG_1 PKG_2 Access walkers conditions WALK_0 WALK_1 WALK_2

Factor scores

Contributions

F1

F2

0.50  0.22  0.58  0.24

 0.51  0.09 0.82 0.27

0.05  0.18 0.05  0.12 0.10

 0.66  0.71 0.15 0.18 1.44

2.03 0.01  0.28

0.55  0.87 0.28

1.47  0.33  0.22

0.68  0.26  0.07

0.56  0.25

0.11  0.05

0.36  0.42

 0.28 0.33

 0.11 0.47  0.40 0.04

 0.63  0.45  0.06 1.12

0.92 0.19  0.19  0.79

0.66  0.11  0.49 0.38

0.60  0.16  1.15

 0.53 0.15 0.98

0.74  0.17

0.07  0.02

1.96 0.75  0.51

0.61 0.07  0.09

1.18  0.09  0.23

 0.02  0.56 0.09

0.16  0.19  0.02

0.47  0.43  0.40

Squared cosines

F1

F2

5.49 3.21 0.19 1.26 0.83 0.29 0.04 0.06 0.01 0.13 0.05 14.06 12.35 0.00 1.71 12.30 10.49 0.77 1.04 4.79 3.32 1.47 5.14 2.37 2.77 3.33 0.10 1.87 1.34 0.01 11.51 6.57 0.18 0.53 4.22 9.71 4.30 0.45 4.96 4.24 3.44 0.79 19.61 8.57 5.25 5.79 8.64 7.26 0.03 1.35 0.90 0.42 0.48 0.00

8.60 4.12 0.04 3.18 1.26 21.20 6.79 1.20 0.15 0.38 12.68 11.84 1.12 8.55 2.18 3.58 2.84 0.62 0.12 0.21 0.15 0.06 3.94 1.82 2.12 19.83 4.08 2.21 0.05 13.50 10.04 4.25 0.08 4.46 1.25 9.22 4.22 0.49 4.50 0.05 0.04 0.01 1.34 1.05 0.06 0.23 1.72 0.00 1.47 0.24 8.43 4.45 2.99 1.00

F1

F2

0.15 0.01 0.04 0.04

0.15 0.00 0.08 0.05

0.00 0.00 0.00 0.01 0.00

0.25 0.03 0.00 0.01 0.35

0.40 0.00 0.14

0.03 0.27 0.14

0.36 0.03 0.09

0.08 0.02 0.01

0.14 0.14

0.00 0.00

0.15 0.15

0.09 0.09

0.00 0.07 0.05 0.00

0.13 0.07 0.00 0.42

0.25 0.01 0.03 0.15

0.13 0.00 0.18 0.04

0.20 0.03 0.16

0.15 0.02 0.12

0.12 0.12

0.00 0.00

0.27 0.21 0.50

0.03 0.00 0.02

0.25 0.00 0.15

0.00 0.04 0.02

0.02 0.02 0.00

0.20 0.11 0.03

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

87

Appendix B. Characterization of the beach clusters

Cluster Discrete attributes 1/4

Test value(a) Cluster Discrete attributes

Parking lot size (vehicles): 41000 4.30 Adequacy of facilities: Good 3.77 Access bicycles conditions: Good 3.76 Picnic areas: Yes 3.59 Site area: 475 Ha 3.40 Département: Gironde 2.84 Fire Risks: Low 2.77 Land reservation-access restriction: Large 2.63 Natural surroundings: Natural 2.53 Access walkers conditions: Poor 2.47 Density of population: Bordeaux & Arcachon Bay 2.39 Touristic area: Arcachon Bay 2.24 Ground movements: None 2.22 Number of visits (summer): Intermediate 2.21 Fire risks: Intermediate 2.19 Adequacy of parking size: Good 2.17 Number of visits (summer): High 2.08 Access bicycles conditions: Intermediate 2.08

2/4

Access walkers conditions: Intermediate Département: Pyrénées Atlantiques Touristic area: Basque coastline Adequacy of parking size: Good Safety: Good Density of population: Basque coastline Access cars conditions: Good Natural surroundings: Semi-urban Adequacy of facilities: Good Ground movements: Rockfalls high Fire risks: None Ground movements: Landslide low Site area: o 22 Ha Shops: 2–3

Test value(a) 5.06 3.95 3.95 3.54 3.46 3.37 3.32 3.03 2.93 2.74 2.60 2.32 2.18 2.09

Cluster

Discrete attributes

Test value(a)

Cluster

Discrete attributes

Test value(a)

3/4

Adequacy of facilities: Intermediate Adequacy of parking size: Intermediate Safety: Intermediate Signs of physical alterations: None Land reservation-access restriction: Ineffective

6.55 3.78 2.97 2.64 2.35

4/4

5.00 4.26 4.22 4.14 2.76

Density of population: Landes coastline

2.03

Safety: None Adequacy of facilities: Poor Access cars conditions: Poor Shops: 0 Number of visits (summer): Low Access cars conditions: Intermediate Picnic areas: No Adequacy of parking size: Poor Other services: No Parking lot size (vehicles): o 50 Overcrowding: None Access bicycles conditions: None

2.44 2.20 2.17 2.08 2.01 2.01 1.98

(a)

The test value (V-Test) is used to detect which level of variables played a statistically significant part in the characterization of the cluster. Only the categories with an absolute value of the V-Test greater than 2 are retained to describe the cluster.

References Abdi, H., & Valentin, D. (2007). Multiple correspondence analysis. In: N. Salkind (Ed.), Encyclopedia of measurement and statistics (pp. 652–658). Thousand Oaks, CA: Sage Publications, Inc. Abdi, H., & Williams, L. J. (2010). Principal component analysis. Wiley Interdisciplinary Reviews: Computational Statistics, 2(4), 433–459. Armsworth, P. R., Cantu-Salazar, L., Parnell, M., & Stoneman, R. (2011). Management costs for small protected areas and economies of scale in habitat conservation. Biological Conservation, 144, 423–429. Bockstael, N. E., & McConnell, K. E. (2007). Environmental and resource valuation with revealed preference. A theoritical guide to empirical models. Dordrecht: Springer Verlag. Bramwell, B., & Lane, B. (2008). Priorities in sustainable tourism research. Journal of Sustainable Tourism, 16(1), 1–14. Chen, W. Y., & Jim, C. Y. (2008). Cost-benefit analysis of the leisure value of urban greening in the new Chinese city of Zhuhai. Cities, 25, 298–309. Cheshire, P., & Sheppard, S. (2002). The welfares economics of land use planning. Journal of Urban Economics, 52, 242–269. Conseil Régional du Tourisme d'Aquitaine (2011). La fréquentation touristique de l'Aquitaine, hors série, édition 2011. Bordeaux: Conseil Régional du Tourisme d'Aquitaine (in French).

Daubet, B., Dehez, J., Figura, S. (2010). Analyse prospective pour un schéma des plans plages sur le littoral Aquitain, rapport pour le GIP Littoral, ONF, CEMAGREF, ICABE, 6 volumesþ annexes, Bordeaux (in French). Escobedo, F. J., Wagner, J. E., Nowak, D. J., Luz De la Maza, C., Rodriguez, M., & Crane, D. E. (2008). Analyzing the cost effectiveness of Santiago, Chiles policy of using urban forests to improve air quality. Journal of Environmental Management, 86, 148–157. Fisher, A., & Krutilla, J. V. (1972). Determination of optimal capacity of resourcebased recreation facilities. Natural Resources Journal, 12(July), 417–444. Fyhri, A., Jacobsen, J. K. S., & Tommervik, H. (2009). Tourists' landscape perceptions and preferences in a Scandinavian coastal region. Landscape and Urban Planning, 91, 202–211. Haider, W. (2006). North American idols: Personal observations on visitor management frameworks and recreation research. In: D. Siegrist, C. Clivaz, M. Hunziker, & S. Iten (Eds.), Exploring the nature of management, Proceedings of the third international conference on monitoring and management of visitors flows in recreational and protected areas (pp. 16–22). Rapperswill, Switzerland: University of Applied Sciences. Hammitt, W. E., & Cole, D. N. (1998). Wildland recreation: Ecology and management (2nd ed.). New York: John Wiley and Sons. Hanley, N., Shaw, W. D., & Wright, R. E. (2003). The new economics of outdoor recreation. Cheltenham, U.K; Northampton, MA, USA: Edward Elgar.

88

J. Dehez, S. Lyser / Journal of Outdoor Recreation and Tourism 7-8 (2014) 75–88

Kaplunosky, A. S. (2005). Factor analysis in environmental studies. Journal of Science and Engineering B, 2(1–2), 54–94. Lebart, L., Piron, M., Morineau, A. (2006). Statistique exploratoire multidimensionnelle. Visualisation et inférence en fouilles de données. Paris: Dunod. Leung, Y., & Marion, J. L. (1999). Characterizing backcountry camping impact in Great Smoky Mountains National Park, USA. Journal of Environmental Management, 57, 193–203. Loomis, J. B., & Walsh, R. G. (1997). Recreation Economic Decisions: Comparing Benefits and Costs (2nd ed.). State College, Pennsylvania: Venture Publishing, Inc. Manning, R. (2004).“Recreation Planning Frameworks”, Society and Natural Resources – A Summary of Knowledge. Prepared for the 10th International Symposium on Society and Resource Management. Jefferson. Mirkin, B. (2005). Clustering for Data Mining: A Data Recovery Approach. Chapman and Hall/CRC. Montes, F., Sarmiento, O. L., Zarama, R., Pratt, M., Wang, G., Jacoby, E., et al. (2011). Do health benefits outweigh the costs of mass recreational programs? An economic analysis of four ciclovia programs. Journal of Urban Health: Bulletin of the New York Academy of Medicine, 89(1), 153–170. Monz, C. A., & Twardock, P. (2010). A classification of backcountry campsites in Prince William Sound, Alaska, USA. Journal of Environmental Management, 91 (2010), 1566–1572. Naidoo, R., Balmford, A., Ferraro, P. J., Polasky, S., Riketts, T. H., & Rouget, M. (2006). Integrating economic costs into conservation planning. Trends in Ecology and Evolution, 21(12), 681–687.

Pröbstl, U., Wirth, V., Elands, B., & Bell, S. (Eds.). (2010). Management of recreation and nature based tourism in European forests. Berlin: Springer. Roca, E., Villares, M., & Ortego, M. I. (2009). Assessing public perceptions on beach quality according to beach users' profile: a case study in the Costa Brava (Spain). Tourism Management, 30, 598–607. Roig, F. X., Comas, E., Rodriguez-Perea, A., & Martin-Prieto, J. A. (2005). Management of beaches on the Island of Menorca (Balearic Islands): The tension between tourism and conservation. Journal of Coastal Research, Special Issue, 49, 89–93. Schéma Plan Plage Littoral Aquitain - Stratégie régionale (2010). GIP Littoral, Bordeaux. Sievänen, T., Arnberger, A., Dehez, J., Grant, N., Jensen, F. S., Skov-Petersen, H. (Eds) (2008). Forest recreation monitoring – a European perspective. Working Papers of the Finish Forest Research Institute, 79, METLA, Helsinki. Stein, T. V. (2013). Planning for the many benefits of nature-based recreation. School of Forest and Conservation Departement, Institute of Food and Agricultural Sciences, University of Florida. Tyrväinen, L., Buchecker, M., Degenhardt, B., & Vuletic, D. (2009). Evaluating the economic and social benefits of forest recreation and nature tourism. In: S. Bell, M. Simpson, L. Tyrväinen, T. Sievänen, & U. Pröbstl (Eds.), European forest recreation and tourism – A handbook (pp. 35–63). Abington, Oxon: Taylor and Francis. Varian, H. L. (1992). Microeconomic analysis (3rd ed.). New York, USA: Norton & Company Inc.