the inter-relationship between rural and mass tourism: … · case study the inter-relationship...

lable at ScienceDirect

Tourism Management 54 (2016) 43e57

Contents lists avai

Tourism Management

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

Case study

The inter-relationship between rural and mass tourism: The case ofCatalonia, Spain

Juan M. Hern�andez a, *, Rafael Su�arez-Vega a, Yolanda Santana-Jim�enez b

a Institute of Tourism and Sustainable Development (TIDES), University of Las Palmas de Gran Canaria, c/ Saulo Tor�on, s/n, 35017, Spainb Department of Quantitative Methods in Economy, University of Las Palmas de Gran Canaria, c/ Saulo Tor�on, s/n, 35017, Spain

h i g h l i g h t s

� A comparison of the main attractions in rural and mass tourism in Mediterranean destinations is made.� A hedonic price model is estimated for rural tourism in Catalonia.� Recommendations for rural tourism management in Catalonia are presented.

a r t i c l e i n f o

Article history:Received 30 October 2014Received in revised form28 October 2015Accepted 28 October 2015Available online 13 November 2015

Keywords:Rural tourismSun-and-sand tourismMediterraneanHedonic pricingGIS

* Corresponding author.E-mail addresses: [email protected] (J.M

ulpgc.es (R. Su�arez-Vega), [email protected] (Y.

http://dx.doi.org/10.1016/j.tourman.2015.10.0150261-5177/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

In certain regions of Southern Europe, mature coastal resorts are currently coexisting with rural tourismareas several kilometers inland. This paper analyzes the inter-relationship of these two types of tourismand the conditions for sustaining both in Mediterranean destinations. To do so, common and uncommoncharacteristics of the tourist product in rural and mass tourism are identified. The case study focuses onthe region of Catalonia, Spain. By applying a hedonic price model, the valuation of some traditional masstourism characteristics in rural tourism is tested. The results show that rural and mass tourism in theregion share several attributes, some of them with opposite effects. The general conclusion is that bothtypes of tourism are compatible, but should be developed and promoted independently to preserve theattractiveness of the destination. Some managerial recommendations for rural tourism in Catalonia arealso derived from this analysis.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Rural tourism has been promoted across several regions inSouthern Europe as an alternative to the traditional mass tourismindustry. For example, several programs, regulations and planshave been implemented in Spain (Barke, 2004), Cyprus (Sharpley,2002), Crete (Andriotis, 2006) and other regions of Greece (Kizos& Iosifides, 2007). Aside from pursuing the socioeconomic revi-talization in rural areas, the diversification strategy of the industryis justified by tourism, environmental and economic reasons. Fromthe tourist-policy aspects, there is, among othermotivations, a needto: a) Look for higher-spending markets which leave a higher shareof economic benefits to the local inland population; b) Reduceseasonality in the tourist industry; c) Adapt the supply to an

. Hern�andez), [email protected]�enez).

increasing demand for destinations with high environmentalquality. From an environmental perspective, the promotion of newattractions (cultural or nature-based inland) also pursued to lessenpressure on the environment in the concentrated coastal areas(Sharpley, 2002; Bramwell, 2004). Pure economic reasons can beadded. Unlike the classical Ricardian theory of international trade,several contemporary economists defend that production special-ization in open economies can be suboptimal if some uncertainties,such as those on trading price or technological coefficient, are highenough (Turnovsky, 1974). These kinds of uncertainties also applyto tourism-based economies in Southern Europe. Specifically, theseeconomies are subject to global competition with other destina-tions in different continents and are dependent on fragile envi-ronmental conditions, which influence on price and the quality ofthe tourist product, respectively. Other socioeconomic problemsderived from the massive affluence of visitors in limited regions(Sheng, 2011) give more arguments to support the implementationof diversification strategies in tourist areas.

mailto:[email protected]




http://crossmark.crossref.org/dialog/?doi=10.1016/j.tourman.2015.10.015&domain=pdf

www.sciencedirect.com/science/journal/02615177

http://www.elsevier.com/locate/tourman

http://dx.doi.org/10.1016/j.tourman.2015.10.015



J.M. Hern�andez et al. / Tourism Management 54 (2016) 43e5744

Although the necessity of alternatives to sun-and-sand tourismis justified in these destinations, the suitability of rural tourism toobtain this objective needs to be analyzed in depth. Rural and sun-and-sand tourism are two branches of the same industry and shareessential elements of the tourism supply. For example, the sametransport facilities (flight routes, roads and so on) and other in-frastructures can be used both by rural or mass tourism. Addi-tionally, they can also share some attractions. Roberts and Hall(2001) stated that the main characteristics of rural areas are: a)Low population density; b) Rural land use; and c) Traditional ruralculture. The typology of rural tourists is diverse, but these charac-teristics conform the main attractions of the destinations. Accord-ingly, Sharpley and Sharpley (1997) identify the need for peace,tranquility, high natural value and unspoilt scenery, but also theenjoyment of traditional culture and gastronomy as the main mo-tivations to visit the countryside. Regarding mass tourism, empir-ical studies carried out in Mediterranean Europe have discoveredthat, although themain valued attributes of the coastal destinationsare still those traditionally related to sun-and-sand tourism(climate, beach, scenery and accommodation quality), other factorssuch as cultural events, heritage and low congestion are alsoidentified as positive attractions of these destinations (Kozak,2002; Yoon & Uysal, 2005; Alegre & Cladera, 2006; Santana-Jim�enez & Hern�andez, 2011; Farmaki, 2012). Therefore, since bothtypes of tourism take place in close geographical areas and sharesome of the elements of the tourist supply, the success of ruraltourism in these destinations can be affected (favored or hindered)by the presence of mass tourism and vice versa.

The activities that beach and rural tourists do at the destinationare not clearly compartmentalized either. Specifically, coastaltourists can take day-trips into the hinterland. This phenomenonhas been observed in some destinations around the globe (Getz,1998) and particularly in Southern Europe (Farmaki, 2012). In thelatter case, the coastal tourists can represent a large proportion ofthe total visitors to hinterland areas. The influence of these trips torural areas has not been analyzed in depth in the scientific litera-ture. Some studies stress the economic benefits to rural areas inislands with mass tourism resorts on their coasts (Kizos & Iosifides,2007), but frequent visits from beach destination to rural areas canbe accompanied by noise pollution and disturbances of rural life,thus harming the rural tourist product. However, to the extent ofthe authors' knowledge, the potential amenity or disamenity ofbeach resorts to neighboring rural tourists is still unexplored.Nevertheless, the influence of potential visitors from rural areas tothe coast may be negligible, since rural tourists would represent asmall part of the total affluence to beach resorts.

The theoretical inter-relationships between the two branches ofthe industry described above have been formalized throughout adynamical model proposed by Hern�andez and Casimiro (2012). Thesimulation of the model, parameterized to explain the case oftourism in some regions of Spain, foresees a strong expansion ofrural tourism combinedwith a slight increase, ormaybe decrease insome cases, of traditional mass tourism.

This study analyses the question of coexistence of rural tourismwith coastal mass tourism in regions of Southern Europe wherelong-developed beach resorts are present. The analysis is con-ducted by comparing the main attractions of mass and ruraltourism in a specific region of Southern Europe. By doing so,common links between both branches of the industry are identi-fied, which allow for the exploration of positive and negative in-fluences of one tourist sector on the other. As an additional result ofthe analysis, some promotional recommendations are proposed toallow for the sustainability of both types of tourism throughouttime.

The identification of the kind of inter-relationship between the

two types of tourism has relevant management implications. Thus,the recognition of complementary factors between mass and ruraltourism underpins designing promotional campaigns that presentboth destinations as a whole. Nevertheless, other inter-dependences which result in a benefit of one sector to the detri-ment of the other should be avoided or dampened. Additionally, theexclusive attractions of rural tourism supply should be enhanced incontrast to those factors which exert a major influence on coastaltourism. Private rural stakeholders can use this information toinvest in inland regions which are endowed with many ruraltourism attributes and enhance the characteristics of their ruralhouses which differentiate them from coastal tourism.

The analysis will be conducted in two steps. Firstly, the majordeterminants and attractions in both rural and sun-and-sandtourism in Mediterranean Europe, found by means of the applica-tion of hedonic pricing models, will be reviewed. Secondly, a casestudy is presented, which is devoted to tourism in Catalonia, Spain.Coastal (mass) tourism in this region started in the 60s and sincethe 80s rural tourism has increased in importance. A hedonic pricemodel was built and applied to rural tourism in Catalonia. Theeconometric specification includes some common attributes of thetwo types of tourism found in the previous analysis. The estimationof parameters is carried out using spatial econometrics andgeographically weighted regression. The latter is a GIS econometricmethodology which allows for the weight measure of each attri-bute according to the geographical location of the rural house. Byusing this technique, the role of the attributes is identified locally,as opposed to the common econometric approach which obtainsaggregated results.

2. Attributes in sun-and-sand and rural tourism in theMediterranean area

Hedonic pricing models assume that a product is composed of abundle of characteristics which are implicitly valued by the marketand added to the final price of the product. Thus, the price p of theproduct (for example, a night in a hotel room) can be described bythe hedonic function p ¼ f(x1, x2, …, xk), where xi represents thelevel or amount of characteristic i (e.g. square meters of the room),i ¼ 1, 2 …,k. The term k represents the number of characteristics ofthe product that are valued by the market. Once the expression offunction f is known (or estimated), the implicit price of the char-acteristics (pi) is obtained by mathematical derivation of the he-donic function, that is, pi ¼ vp/vxi, i ¼ 1, 2, …,k. The economicfundaments of the hedonic pricing methodology were developedby Rosen (1974).

In this section, the most common significant attributes of thesun-and-sand and rural tourism in Southern Europe found in therecent literature by applying hedonic price models are reviewed.Several considerations need to be made before presenting theresults:

a) First, the attributes included in these hedonic pricing modelsrepresent attractions, services, facilities and informationwhich compound the supply side of the tourist system. So,they represent destination-specific or supply-side factors,whichmake the tourist choose the destination, as opposed toother demand-side factors, such as income, time available orother socio-cultural factors, which are not taken into accountby means of this methodology. For the sake of comparison,specific characteristics found by each particular reference aregrouped into general attributes (e.g. the availability of aparking site, breakfast included and tennis court, amongothers, are grouped into the attribute Services&Facilities).

J.M. Hern�andez et al. / Tourism Management 54 (2016) 43e57 45

Other non-destination-specific characteristics, such as theinfluence of the tour operator, are disregarded in this review.

b) Second, the tourist product in these models is usually rep-resented by the hotel room or rural house, whose rental priceper night includes the valuation of the rest of attributes ofthe destination. Therefore, only the preference for overnighttourism is considered in the study. Daily visitors are notincluded in the analysis, so their possible indirect influenceson the rental price of the accommodation unit are dis-regarded. So, in this paper, only the tourist staying in a hotelroom (or an apartment) in the coast and in a rural house isconsidered mass and rural tourist, respectively.

c) Third, it is necessary to bear in mind that the hedonic pricingmethodology assumes perfect competition and price isdetermined by the market equilibrium. So, differences inprices among accommodation units are not only due to thedemand forces in the market, but also to the supply ones,such as the costs or barriers in offering the product. Thishypothesis must be taken into account in the interpretationof results.

d) Finally, most of the studies analyzed in this review focuses ona specific area in Mediterranean Europe, so the attributes canonly be strictly applied to that specific destination. Never-theless, since the region shares some general supply char-acteristics (climate, environmental and socioculturalattributes) and also demand characteristics (mainly fromNorthern Europe), similarities among destinations areobserved.

2.1. Attributes of sun-and-sand tourism in Mediterraneandestinations

Table 1a includes a summary of previous studies which use thehedonic pricing approach to examine the most relevant attributesof accommodation units in the Mediterranean European destina-tions. The hotel room is the most frequent unit in these studies,although there are some studies which analyze apartments, secondhome or private accommodation. The Table does not include thewhole literature on this issue. The review is restricted to thosepapers which, in the authors' opinion, analyze new locations orintroduce new significant attributes or insights.

Following the usual classification made in the application ofhedonic pricing models to the housing market (Taylor, 2008), theattributes or determinants of the room price are divided into twocategories: structural and location variables. The structural vari-ables describe the physical characteristics of the room (squaremeters, number of beds, etc.) and facilities or services provided bythe hotel (swimming pool, breakfast included, amenities, etc.),while the location variables combine the characteristics of the areasurrounding the room, such as the security level, population den-sity or number of restaurants, and those factors related to itsgeographical position, such as the proximity to amenities or ser-vices (area or townwhere the room is located, sea view, distance tothe beach, airport, town center, etc.).

As illustrated in Table 1a, most of the sun-and-sand Mediter-ranean destinations analyzed by hedonic pricing models arelocated in Spain, mainly in Catalonia, but also in the BalearicIslands, Andalusia and the Canary Islands (although located in theAtlantic Ocean, the Canary Islands are included in the group, sincethey have similar socioeconomic and cultural characteristics to therest of Mediterranean Europe). There are only two authors whohave analyzed the Mediterranean region as a whole and anotherauthor who has analyzed Croatia. Therefore, the analysis may bebiased by the particular conditions in Spain. In addition, there are

also differences in the selection of explanatory variables, sinceevery study stresses some specific attribute of the tourist product.

Recognizing that the sample is restrictive, some general con-clusions can be extracted from the comparison of the differentstudies. First, regarding the structural variables, facilities and ser-vices in the hotel or resort (swimming pool, half or full board,amenities, etc.) increases the rental price in all studies. Whenincluded as an explanatory variable, the star rating also appears as adeterminant of the room price. This result was expected, since thisrating is considered a trustworthy indicator of the quality of hotelservices and facilities. Additionally, the increase in room capacity(number of beds) also raises the room price. It is noteworthy thatthe presence of artificial infrastructures, such as marinas, areappreciated in coastal tourism in Catalonia but this is not the casefor cultural facilities (Rigall-I-Torrent & Fluvi�a, 2011). This resultagrees with Papatheodorou (2002), which observes that built at-tractions are primed in Mediterranean destinations. Unlike previ-ous findings, in other studies cultural attractions appear as apositive attribute for Catalonia (Sal�o, Garriga, Rigall-I-Torrent, Vila,& Fluvi�a, 2014) and Dubrovnik (Portolan, 2013).

The sea view is the main location attribute of the hotel room andit is positively valued in all the studies which have included it in thebundle of explanatory variables. Beach proximity also increases therental price of the room in several studies, although it does not havea significant effect in some of them (Aguil�o, Alegre, & Sard, 2003;Rigall-I-Torrent et al. 2011). Alegre, Cladera, and Sard (2013) findsa nonlinear influence of the distance to the beach on the rentalprice of the hotel room in the island of Mallorca. Themajor negativeinfluence of the distance factor is shown when comparing hotelslocated relatively close to the beach (e.g. hotels on the beachfrontare more positively valued than hotels away from the beach) and isnegligible for those located at a further distance from the coast. Theopen coastal landscapes and ocean proximity are also positivelyvalued attributes in hedonic models for hotels in other regions,such as Hawaii (Cox & Vieth, 2003) or the Northern states of Ger-many (Hamilton, 2007). Therefore, as expected, the beach (its viewand proximity) is one of the main attractions of the coastal desti-nations and is usually implicitly valued in the rental price of theaccommodation units. Another significant attribute is the specificdestination or area where the hotel is located, which can be inter-preted as an added value to certain destinations with respect toothers. The proximity to a natural area is positively valued in someof the studies (Aguil�o et al., 2003; Rigall-I-Torrent & Fluvi�a, 2011).

Crowding or congestion (measured as population or rooms persquare km) is a regular location attribute which detracts value fromthe hotel room (Papatheodorou, 2002; Rigall-I-Torrent & Fluvi�a,2011; Rigall-I-Torrent et al., 2011; Sal�o et al., 2014). Nevertheless,this does not present a uniform value for all the destinationsanalyzed. Congestion was not found as an explanatory variable ofroom price in the island of Mallorca (Alegre et al., 2013). Otherlocation attributes positively valued by themarket are the adoptionby the hotel of environmentally sustainable measures, security orthe number of restaurants in the surroundings.

2.2. Attributes of rural tourism in Mediterranean destinations

The number of studies analyzing the implicit attributes of ruralaccommodation is still rather limited. Table 1b presents five con-tributions which focus on rural tourism in Southern Europe.Although the sample is short, some recurrences can be observedfrom a comparison among the studies. Thus, similar infrastructuresand services are positively valued in the different regions, such asthe existence of a swimming pool or the higher capacity (number ofbeds or rooms) of the house. Among the location variables, land-scape view and diversity, measured through the level of diverse

Table 1bSummary of the valued attributes of a rural house in Mediterranean destinations using the hedonic pricing approach. The influence of the attribute on the rental price of therural house is positive by default. Notation (�) indicates a negative influence of this factor. Description of variables: Activities, number of tourist activities offered by theoperator; Congestion, population density in the surroundings; Cultural attractions, whether a specific cultural attraction exists in the location; Distance to view the stars, distanceto the nearest viewing point of the stars; Distance to port, distance to the nearest port; Green landscape, forest landscape in the surroundings of the house; Landscape diversity,diversity of land uses in the surroundings of the rural house; Landscape view, outstanding landscape view from the house;Mild temperature, non-extreme temperatures in thespecific location; Number of beds/rooms, number of beds/rooms in the house; Restaurants, number of restaurants in the surroundings of the house; Rural area, house locatedoutside urban centers; S&F, the house includes fireplace, jacuzzi, restaurant, swimming pool, TV or other service/facility; Size, square meters of the house/accommodation.

Study Location Structural variables Location variables

Fleischer and Tchetchik (2005) a Israel S&F, Size Landscape view, Cultural attractions, ActivitiesOhe and Ciani (2011) Italy S&F, Number of beds Cultural attractionsb

Santana-Jim�enez et al. (2011) Gran Canaria (Canary Islands, Spain) S&F Landscape diversity, Congestion (�), Rural areaSantana-Jim�enez, Hern�andez,

and Su�arez-Vega (2013)La Gomera (Canary Islands, Spain) S&F, Number of beds Green landscape, Rural area, Mild temperature

Su�arez-Vega et al. (2013) La Palma (Canary Islands, Spain) S&F, Number of rooms Landscape diversity, Distance to view thestars (�), Distance to port, Mild temperature

a The accommodation units in this study are Bed&Breakfasts in rural areas.b Specifically, the number of UNESCO sites and wineries designation with origin.

Table 1aSummary of the valued attributes of a hotel room in Mediterranean destinations using the hedonic pricing approach. The influence of the attribute on the rental price of thehotel room is positive by default. Notation (�) indicates a negative influence, (þ,�) means that the sign depends on the specific variable included in this category. Description ofvariables: Built attractions, artificial attractions (such as marinas, thematic parks, etc.); Capacity, number of beds per room; Congestion, room or population density in thesurroundings; Cultural attractions, whether a specific cultural attraction exists in the location; Distance to the beach/coast, distance to the nearest beach/coast; Environmentalmeasures, the hotel includes environmentally sustainable measures; Distance to natural areas, distance to natural areas of interest (e.g. caves); Restaurants, number of res-taurants in the surroundings of the hotel; Sea view, whether the room is located in front of the beach or enjoys sea views; S&F, Service and Facilities, whether the hotel includesa swimming pool, parking availability, etc., breakfast included, sport amenities, etc.; Security, local police officers per hotel room; Size, number of rooms in the Hotel/House/Apartment; Stars, hotel category (number of stars); Specific destination (area/town), destination (area/town) where the hotel is located.

Study Location Structural variables Location variables

Papatheodorou (2002) Mediterranean Europe S&F, Stars, Built attractions Sea view, Distance to the beach (�),Congestion (�), Specific destination,

Aguil�o et al. (2003) Balearic Islands (Spain) S&F, Stars, Capacity Sea view, Distance to natural areas (�), Specific areaEspinet, Saez, Coenders,

and Fluvi�a (2003)Southern Catalonia (Spain) S&F, Stars, Size (�) Sea view, Specific town

Thrane (2005) Canary Islands (Spain) S&F, Stars Distance to the beach (�)Sal�o and Garriga (2011) a Catalonia, (Spain) S&F, Stars, Size Sea view, Distance to the beach (�), Specific townRigall-I-Torrent and Fluvi�a (2011) Catalonia (Spain) S&F, Stars, Built attractions Sea view, Distance to natural areas (�),

Congestion (�), Security, Restaurants, Specific townRigall-I-Torrent et al. (2011) Catalonia, (Spain) S&F, Stars Sea view, Congestion (�), Beach width (�), SecurityFleischer (2012) Mediterranean area S&F, Capacity Sea view, Specific destinationAlegre et al. (2013) Mallorca (Spain) S&F, Stars Sea view, Distance to the beach (�), Specific areaPortolan (2013) b Dubrovnik (Croatia) S&F Distance to the beach (�), Cultural attractionsSanchez-Ollero, Garcia-Pozo,

and Marchante-Mera (2013)Andalusia (Spain) S&F, Stars Distance to the coast (�), Environmental measures

Vilchez (2013) c Catalonia, Balearic Islands,Valencia (Spain),Languedoc-Rousillon (France)

S&F, Stars,Size (þ,�)e, Capacity

Specific destination

Sal�o et al. (2014) d Catalonia (Spain) S&F, Stars Sea view, Distance to the beach (�),Congestion (�), Security, Culturalattractions, Specific destination

a It centers on second home rental market.b It analyzes private accommodation units.c This study combines hotel rooms and tourist apartments.d This study combines hotel rooms and second home rental market.e Positive for tourist apartments and negative for hotel rooms.


land uses in the house's neighborhood, is a positively-valuedattribute of the rural house. Landscape diversity was also valuedpositively for tourism in recreational sites on the island of Mallorca(Bujosa-Bestard & Font, 2009).

Other location variables illustrate the characteristics of ruraltourism in the region. Congestion, measured as the populationdensity in the surroundings, is negatively valued by rural touristson the island of Gran Canaria, Spain (Santana-Jim�enez, Su�arez-Vega, & Hern�andez, 2011), while houses located outside urbancenters is significantly appreciated in this destination. Theremote location of the house also raises its rental price in theisland of La Palma (Su�arez-Vega, Acosta-Gonz�alez, Casimiro-Reina, & Hern�andez, 2013). Cultural or natural attributes have a

positive implicit price in Israel, Italy and the Canary Islands.Finally, mild temperatures are also a tourist's preference on theislands of La Palma and La Gomera.

2.3. Comparison of attributes in rural and coastal tourism forMediterranean destinations

In general, although all the destinations were in the Mediter-ranean environment, the destinations analyzed in the empiricalresearch for mass tourism are not the same as those for ruraltourism. With the necessary caution derived from these sampleconditions, the review above evidenced the existence of similar anddissimilar attributes of the room/house in the two branches of the


industry. They can be summarized in four groups:

� Structural variables: There are some facilities and services thatappear positively valued in the hedonic function of the twotourist products. For example, the existence of a swimming poolincreases the rental value of the hotel room and rural house.Moreover, the positive effect on rental price of other similarservices, such as the availability of a restaurant or breakfastincluded, are also shared by the two products. These resultsshow that rural and sun-and-sand consumers appreciate certaincommon attractions.

� Congestion: Although to a different degree, high population/room density usually depreciates the value of the tourist prod-uct in the two markets. In fact, congestion alleviation in coastalareas was one of the reasons for promoting rural tourism inmass tourism destinations. However, the empirical analysis in-dicates that an increase of visitors in rural areas also detractsvalue to rural lodgings. Therefore, excessive affluence is nowa-days a factor that limits profitability of the industry in bothmarkets.

� Coastal attractions: Sea view and distance to the beach are de-terminants of the room price in coastal tourism. The studies forrural tourism for now have not revealed a significant influenceof the beach proximity.

� Natural and cultural attractions: This is a preferred amenity forrural tourists, but also exerts a positive influence in some he-donic price models for coastal areas in Croatia, Catalonia and theBalearic Islands.

The comparative analysis above reveals clear evidences of linksbetween the two types of tourism. Nevertheless, it also showscertain gaps still to be covered. Firstly, except for the case of GranCanaria (Thrane, 2005; Santana-Jim�enez et al., 2011), the regionanalyzed in themass tourism studies does not usually coincidewiththe region selected in the rural tourism studies, which detractsvalue to the joint consideration of attributes. Secondly, relevantattributes considered in the majority of mass tourism papers inTable 1a were not taken into account in the bundle of characteris-tics used in rural accommodation units (as is the case of the dis-tance from the beach). The coast (not specifically the beach) canrepresent a natural amenity for rural tourists. In certain ruralMediterranean destinations, it may be also a disamenity, since it isthe location where mass tourism takes place. The influence of thisfactor in the hedonic price of rural houses tourism is still unknown.

In order to partially fill this gap, the case study will test how themarket values some of the mass tourism attributes described abovein rural tourism in Catalonia, Spain. The selection of this region isbased on two reasons: a) The availability of rural tourism infor-mation by private owners and public administration; b) Many he-donic pricing models for mass tourism in Table 1a were applied tocoastal areas in this region, but a value of rural tourism attributes inCatalonia using this methodology has not been undertaken up untilnow.

3. Case study

3.1. Study area

The study area is Catalonia, a Spanish region located in theNorth-east of the Iberian Peninsula (Fig. 1). Tourist evolutionthroughout the last half century in Catalonia presents a similarpattern to other Mediterranean areas. This is characterized by astrong growth in the sixties leaded by a massive affluence oftourists coming from Northern Europe looking for sun-and-sanddestinations, and followed by several phases of stagnation and

recovery (Garay & C�anoves, 2011). Nowadays, Catalonia receivesmore than thirty million tourists every year, which leave V16billion and sustain 11% of total employment in the region(Generalitat de Catalunya, 2014a). Tourists stay not only in thenumerous resorts throughout the coast, but also in other alterna-tive destinations, such as the town of Barcelona, and in natural andrural areas.

Rural tourism in Catalonia has grown in importance since the90's. Starting from then, rural accommodation in the region pre-sents an average annual growth of 13.2% from 2001 to 2013 (INE,2014). Most visitors are local inhabitants (83%), concentrated inthe summer season. The average length of stay is three days(Idestac, 2014). Competences in matters of tourism were trans-ferred to the local government at the beginning of the 80's.Therefore, rural tourism regulation is implemented by the Auton-omous Government and is different from the rest of regions inSpain. These political circumstances make rural tourism in Cata-lonia susceptible to be analyzed as a whole.

To illustrate and compare the recent evolution of the two typesof tourism in Catalonia, Table 2a and Table 2b show the number ofmass and rural tourists in Catalonia, respectively. The data is dis-aggregated by origin and provinces (Barcelona, Girona, Lleida andTarragona). Two years were selected, 2009 and 2013. As com-mented above, the Tables show that most rural tourists come fromSpain (87.81% in 2013), while mass tourism is mainly demanded byforeigners (63.17% in 2013). It is remarkable that domestic demandfor both types of tourism has strongly decreased in all provinceswith the exception of Barcelona, which has maintained stable.However, foreign tourists have grown in the two types of tourism,more intensively in the case of rural tourism, which starts from ashare of 7% in 2009 to 12% in 2013 (in overnight stays, it rises from11% to 27%).

The preferred province in Catalonia by mass tourists is Barce-lona (59.46% share in 2013), due mainly to the strong attraction ofthe town of Barcelona, followed by the province of Girona (20.72%share in 2013), where many beach resorts are located. These twoprovinces are also the favorite ones for rural tourists, althoughGirona is in the first place of preference followed by Barcelona(41.85% and 30.50% share in 2013, respectively). The empiricalanalysis of rural tourism followed in this paper will provide newinsights about the rationale of these facts, which will be discussedbelow.

3.2. Data and variables

The study uses 393 rural tourism accommodation units in Cat-alonia from a total of 1569 registered in 2009 (INE, 2014). Therewasnot an official and public record of the characteristics of every unitat that time, so the sample was extracted from the tourist agencywebsites specializing in rural tourism. In order to include a variedrepresentation of environmental and location characteristics of thehouses, the sample was selected to cover most of the official dis-tricts in Catalonia. Two intermediaries were sampled, Asociaci�onEspa~nola de Turismo Rural (ASETUR) and Turismo y Ocio enInternet (TURISPAIN). Initially, a percentage of 50% of the total ruralhouses offered by ASETUR, segmented by districts, was randomlysampled. The information for most of the rural houses in theprovince of Girona was extracted from ASETUR. Due to the fall ofpublished information from ASETUR's web-page during the samplecollection process, the rest was extracted from the-page of TURI-SPAIN, maintaining the same sample criteria.

Fig. 1 presents a map of the rural houses sample used in thisstudy. As observed in the Figure, rural tourism lodgings are het-erogeneously distributed throughout the geography of Catalonia. Alarge amount of the supply is concentrated in the Pyrenees and pre-

Fig. 1. Distribution of the sample of rural houses in Catalonia used in the case study. Thick lines delimit the provinces in Catalonia (Barcelona, Girona, Lleida and Tarragona), thinlines the districts. Population density is measured in inhabitants/km2 in every district; the darker area corresponds to the town of Barcelona. Points of interest included in theeconometric estimation of the hedonic price function are also geographically located.

Table 2aSpanish and foreign tourists staying in hotels and apartments in Catalonia.

2009 2013 %D2009e2013

Total %Catalonia Total %Catalonia

BarcelonaSpanish 3,101,328 21.02% 2,823,065 16.54% �8.97%Foreigners 5,302,911 35.95% 7,327,693 42.92% 38.18%

GironaSpanish 1,308,842 8.87% 1,484,103 8.69% 13.39%Foreigners 1,860,617 12.61% 2,054,527 12.03% 10.42%

LleidaSpanish 692,404 4.69% 623,740 3.65% �9.92%Foreigners 108,527 0.74% 120,175 0.70% 10.73%

TarragonaSpanish 1,436,800 9.74% 1,356,706 7.95% �5.57%Foreigners 939,661 6.37% 1,283,088 7.52% 36.55%

CataloniaSpanish 6,539,375 44.33% 6,287,614 36.83% �3.85%Foreigners 8,211,717 55.67% 10,785,483 63.17% 31.34%

Source: INE (2014).


Pyrenees zones, which are endowed with many natural attractions,followed by the province of Tarragona, in the South of the region.The medium latitude zone shows fewer accommodation units. Thisarea includes the Metropolitan region of Barcelona and the flat, dryand agriculturally intensive areas in the South of the province ofLleida, which is presented as a reason for their lesser rural tourismdevelopment (C�anoves, Herrera, & Cuesta, 2005). There are notmany rural houses in the coastal areas either, since they are alreadyoccupied by seaside resorts.

The variables included in the model estimation are shown inTable 3. They represent the attributes found in previous hedonicprice models plus some other location characteristics related tocoastal tourism. The variable Price, which represents the rental

price for two people during high season, was collected from thewebsite of every lodging, together with the number of rooms, beds,total surface occupied by the house and the existence of specificinfrastructures or services (Satellite TV, barbecue, swimming pool,Jacuzzi, fireplace, washing machine or pets allowed). Thegeographic coordinates were checked with Google Earth ©.

The GIS software has been applied in order to build the locationvariables. For this purpose, rural houses and several sites of in-terest (capitals of provinces, beaches, natural reserves and ruralpoints of interest) were geocoded. Natural reserves were obtainedfrom the Generalitat de Catalunya (2014b). Beaches and ruralpoints of interest were selected from those used in previous he-donic pricing studies (Rigall-I-Torrent & Fluvi�a, 2011; C�anoves

Table 2bSpanish and foreign tourists staying in rural houses in Catalonia.

2009 2013 %D2009e2013

Total %Catalonia Total %Catalonia

BarcelonaSpanish 74,947 22.66% 72,785 27.11% �2.88%Foreigners 7093 2.14% 9096 3.39% 28.24%

GironaSpanish 124,942 37.77% 94,970 35.37% �23.99%Foreigners 9713 2.94% 17,390 6.48% 79.04%

LleidaSpanish 78,538 23.74% 48,291 17.99% �38.51%Foreigners 3129 0.95% 2326 0.87% �25.66%

TarragonaSpanish 30,244 9.14% 19,710 7.34% �34.83%Foreigners 2203 0.67% 3923 1.46% 78.08%

CataloniaSpanish 308,671 93.31% 235,756 87.81% �23.62%Foreigners 22,138 6.69% 32,735 12.19% 47.87%

Source: INE (2014).


et al., 2005). From this information base, straight-line distancesfrom every rural house to the closest site of interest were calcu-lated (Dist_nearest_cap_prov_Km, Dist_nearest_beach_Km, Dis-t_nearest_rural_int_Km). The number of sites of interest within aspecific radius (r) from every rural house was also obtained(Npoints_rural_int_rKm, Ncap_prov_rKm, Nbeaches_rKm). Theprovince (Barcelona, Girona, Lleida and Tarragona) where the ruralhouse is located was also considered.

The diversity of land uses in the surroundings of a house wasincluded as an environmental attribute. The index proposed byBastian, McLeod, Germino, Reiners, and Blasko (2002) is used here,

Table 3Definition and descriptive statistics of the variables included in the estimation of the he

Variables Definition

Price Price per night for 2 people (euros)StructuralBeds Number of beds in the houseBarbecue Barbecue in the house (1 ¼ yes. 0 ¼ no)Satellite TV Satellite TV in the house (1 ¼ yes. 0 ¼ no)Pool Swimming pool in the house (1 ¼ yes. 0 ¼ no)Jacuzzi Jacuzzi in the house (1 ¼ yes. 0 ¼ no)Fireplace Fireplace in the house (1 ¼ yes. 0 ¼ no)Pets allowed Pets allowedResort Included in a resort (1 ¼ yes. 0 no)LocationBarcelona Located in Barcelona (1 ¼ yes. 0 ¼ no)Girona Located in Girona (1 ¼ yes. 0 ¼ no)Lleida Located in Lleida (1 ¼ yes. 0 ¼ no)Tarragona Located in Tarragona (1 ¼ yes. 0 ¼ no)Landscape Diversity Diversity Index within a radius of 1KmDist_nearest_cap_ prov_Km Distance to nearest capital of province (Kms)Dist_nearest_beach_Km Distance to nearest beach (Km)Dist_nearest_rural_int_Km Dist. to nearest rural interest point (Km)Npoints_rural_int_25Km No points of rural interest within 25KmsNpoints_rural_int_50Km No points of rural interest within 50KmsNpoints_rural_int_100Km No points of rural interest within 100KmsNcap_prov_25Km No capitals of provinces within 25KmsNcap_prov_50Km No capitals of provinces within 50KmsNcap_prov_100Km No capitals of provinces within 100KmsNbeaches_25Km No beaches within 25KmsNbeaches_50Km No beaches within 50KmsNbeaches_100Km No beaches within 100KmsPop_1Km Population within 1 KmPop_5Km Population within 5KmsPop_50Km Population within 50KmsPop_100Km Population within 100Kms

Landscape Diversity i ¼ 1�Xmh¼1

ð4hÞ2;

where 4h indicates the proportion of land use h in a radius of i kmsfrom the rural house andm represents the total amount of possibleland uses. The higher the value of Landscape Diversity_i is, thehigher the diversity of land uses in a radius of i kms from the houseis. These variables were calculated using the land uses map pro-vided by the Corine program of the European EnvironmentalAgency (Corine Land Cover, 2006).

Variables Pop_rKm show the population within an r-radius fromthe rural house. These variables have been obtained from the mapsprovided by the LandScan 2007™ High Resolution global Popula-tion Data Set (Bright, Coleman, King, & Rose, 2008). Radii of 1, 5, 50and 100 km have been considered.

3.3. The methodology used in estimating the hedonic price model

The hedonic price model (HPM) was chosen to represent therelationship between attributes and prices for rural houses inCatalonia. As commented above, the right application of thismethodology is based on the assumption that the good (e.g. ruralhouse) is commercialized in a unique market in perfect competi-tion, where producers and consumers have homogeneous infor-mation regarding the attributes of the good. These are rather strongconditions that cannot be radically verified in most of the hedonicpricing application. Nevertheless, there are arguments to acceptthem in the specific case presented in this paper. First, the hy-pothesis of single market is plausible, since Catalonia is a well-defined geographic and sociopolitical region, where most of po-tential rural tourists are local residents. Second, the hypothesis of

donic price model for rural houses in Catalonia. Data corresponds to year 2009.

Mean Std. Dev. Min. Max.

65.19 33.44 15 240

8.92 3.99 2 200.59 0.49 0 10.02 0.14 0 10.43 0.49 0 10.02 0.14 0 10.63 0.48 0 10.27 0.44 0 10.13 0.34 0 1

0.25 0.43 0 10.38 0.48 0 10.22 0.42 0 10.13 0.33 0 10.53 0.18 0.0002 0.8449.58 27.72 5.23 12257.5 37.67 0.55 16514.73 10.33 0.02 632.49 1.94 0 77.34 3.41 0 1117.6 5.73 3 260.14 0.35 0 10.36 0.48 0 10.43 0.49 0 10.48 1.14 0 82.14 2.92 0 117.47 4.34 0 13511 1072.6 0.1 67156.26 11,046 14 92,789901.94 1.02$106 44,492 4.55$106

4.16$106 1.72$106 0.76$106 5.91$106


absence of information barriers is justified by the way to collectdata. Rural houses are announced in few websites, where infor-mation (price and attributes) can be easily found and compared bytourists and owners. Third, the relatively large amount of ruralhouses in the market (1569) works also in favor to assume a situ-ation close to market equilibrium in a competitive market.

Given the extensive amount of attributes considered in themodel, a method for selecting explanatory variables has been fol-lowed in this study. It consists of three steps:

� Firstly, in order to select the attributes that optimize the fit ofthe econometric estimation, stepwise has been appliedtogether with the Genetic Algorithms guided by the SchwarzInformation Criterion (GASIC). GASIC (Acosta-Gonz�alez &Fern�andez-Rodríguez, 2007) is a procedure for automatic se-lection of factors in the regression model. Guided only by data,GASIC was constructed using a genetic algorithm where thelost function is the Schwartz Information Criterion (SIC). Thismethodology avoids the tendency to over-identify modelsdetected in several popular heuristic methods for selectingmodels (such as stepwise, see Lovell, 1983). Moreover, byallowing all possible combinations of potential regressors, itrenders results that are robust to any specification. Su�arez-Vega et al. (2013) show an application of this method to theselection of implicit attributes of rural houses in the island ofLa Palma (Spain).

� Secondly, results from both GASIC and stepwise procedureshave been crossed and finer selections of attributes have beenmade based on the comparison of outputs obtained from bothprocedures.

� Thirdly, the spatial dependence of the data is tested andincluded in the model. The existence of spatial dependence maybe taken into account in two different ways:a) By using the spatial econometric approach, if spatial

autocorrelation is assumed, which considers spatialdependence in the endogenous variable or in the error(Anselin, 2005). In the first case the value of the dependentvariable is affected by the value of the variable corre-sponding to the neighboring units, while in the secondcase the error captures the existence of omitted variablesthat induce certain spatial dependence. Ignoring the exis-tence of spatial dependence in the endogenous variableentails biased and inconsistent ordinary least squares (OLS)estimators, while ignoring spatial dependence in the errorinvolves unbiased but inefficient OLS estimators. A generalspatial model including spatial dependence in the endog-enous variable (price) and in the error was tested in theempirical analysis.

b) By estimating geographically weighted regression models(Brunsdon, Fotheringham, & Charlton, 1996), if spatial non-stationarity is assumed, which uses distance-decay-weighted sub-samples of the data to produce locally linearestimates for every point in space. This model captures theidea that parameters vary according to the spatial location ofthe data. The theoretical justification for the use of thesemodels depends upon the application. In this specific casewhere data corresponds to rural houses and factors thataffect their rental price in Catalonia, it is reasonable to thinkthat parameters of the model could vary locally, and thishypothesis should be checked. Geographically weightedregression models do not fully ignore spatial autocorrelation,but explain it as part of the local analysis (Fotheringhamet al., 2003, p. 114e115).

The details of both methodologies are shown below.

3.3.1. Spatial econometric modelThe general specification of the spatial econometric model for

hedonic prices proposed for the rental houses in Catalonia is thefollowing:

lnðpÞ ¼ a1þ rWlnðpÞ þ Xbþ u; (1)

Where u ¼ lWuþε, ε~N(0,s2In). The dependent variable ln(p)represents the n-vector of logarithm of prices of the n houses; a isthe intercept (scalar), and 1 is an n rows vector of ones. MatrixWnxnrepresents the influence of the price of houses in a neighborhood ofa given house. The determination of which observations areconsidered neighbors is crucial for the model estimation. Coeffi-cient r measures the degree of dependence among the rental pri-ces. A positive sign of r indicates that the rental price of each houseis positively relatedwith the rental price of nearby houses. Vector b,with m rows, is the parameter corresponding to the total number ofm variables (X) considered in the model. If l ¼ 0, the spatial lagmodel is obtained, while if r ¼ 0, the resulting model is the spatialerror model. The general spatial model is the model defined in (1).

All the independent variables are expressed in linear terms,except Pop5Km, Pop100Km and Capacity, which are inverted due tooptimum fit reasons. Finally, u is the vector of errors that follows aspatial error model.

The existence of spatial correlation is tested by applying theLagrange Multiplier (error) test and its robust version (RobustLagrange Multiplier (error) test) and the Lagrange Multiplier(spatial lag) test and its robust version (Robust Lagrange Multiplier(spatial lag)) to the classical OLS model.

Finally, there exist additional tests aimed at valuing whether thespatial specification significantly improves the classical model ornot: the Likelihood ratio test, that studies the suitability of theproposed spatial lag model (or the spatial error model) and theLagrange multiplier test, which allows for the verification thatother significant spatial structures in the proposed spatial modelhave not been omitted. Further details on these tests are availablein Anselin (2005).

3.3.2. Geographically weighted regression (GWR) modelThis methodology is based on the assumption that geographi-

cally closer elements must present closer values. Hence, the re-gressors for a particular house i, located at geographical coordinates(ui,vi), must be estimated under the assumption that nearbyobserved data is more influential than that located further away.The estimator for the GWR model is similar to the weighted leastsquares global model, but considers that weights are associated tothe location of the observed values. A specific GWR model adaptedto our problem can be written as

lnðpÞ ¼ aðu; vÞ þ Xbðu; vÞ þ ε; (2)

whereε~N(0,s2In) and regressors b(u,v) are allowed to vary overspace. The intercept a(u,v) is an nx1 vector and is estimated forevery specific house.

Let X1¼ (1,X) be the matrix of mþ1 columns of explanatoryvariables X together with the vector of ones and n rows corre-sponding to each individual house. Analogously, letb1(ui,vi) ¼ (a(ui,vi),b(ui,vi)) be the combination of the intercepta(ui,vi) and regressors b(ui,vi) for the specific house i. The estima-tion of the vector of coefficient regressors b1 is:

bb1ðui; viÞ ¼�X10

Wðui; viÞX1��1

X10Wðui;viÞlnðpÞ; (3)

where W(ui,vi) is a diagonal matrix. The element wj(ui,vi) in thediagonal (j ¼ 1,2,…,n) represents the weight of the observation j in

Table 4Goodness of fit measures and model validity tests. The models are: Ordinary leastsquares (OLS) model; spatial error (SE) and spatial lag (SL) model, both consideringW the spatial weight matrix that connects the nearest neighbor from each house;and mixed geographically weighted regression (MGWR) model.

OLS SE SL MGWRd

Adjusted R2 0.21 0.204 0.246 0.244Normality test 8.42** 8.7** 8.2**

BreuschePagan test 23.6** 22.08* 22.39*

Adjusted Akaike Inf. C. (AICC) 391.74 363.4 365.8 378.82

Diagnostic forspatial dependence

Lagrange Multiplier (error) 21.52***

Robust LM (error) 0.50Lagrange Multiplier

(spatial lag)21.28***

Robust LM (spatial lag) 0.26Likelihood ratio testa 22.6*** 22.23***

Lagrange Multipliers testb 0.005 0.12Improvement F-testc 3.62***

Super-indexes ***, ** and * denote significance at 1%, 5% and 10%, respectively.


the estimations of coefficient bb1ðui; viÞ for the house i. The specificmethod for calculating the different weights W(ui,vi) is calledkernel. The most frequently used kernels are the Gaussian one

wjðui; viÞ ¼ e��

dijh

�2

and the Bi-square one,

wj ui; við Þ ¼

8><>: 1� dijh

� �2 !2

; if dij <h;

0 ; if dij � h;

where dij is the Euclidean distance between location (ui,vi) andobservation j, and h is called the bandwidth (written in the samecoordinate units used in the dataset). While the specific kernelshape does not usually influence a lot on the estimation results, thechoice of the bandwidth may cause significant changes in theestimated coefficients. For larger values of h, the weights wj(ui,vi)are close to one and GWR's estimations tend towards those ob-tained by the OLS.When the sample is regularly spaced in the studyarea, a kernel with fixed bandwidth is recommended; otherwise anadaptive bandwidth may be indicated. In the adaptive form, aminimumnumber of observations or amaximum distance are fixedin order to select the subsample used for calculating the weights.

In order to obtain a more flexible model, Brunsdon,Fotheringham, and Charlton (1999) proposed the mixedgeographically weighted regression (MGWR) model. In this newmodel, some parameters are considered fixed but others areallowed to vary over space. The specific MGWR model used herecan be written as

lnðpÞ ¼ aðu; vÞ þ XNCbNCðu; vÞ þ XCbC þ ε; (4)

Where ε ~ N(0,s2In). There are two types of variables: XNC, whosecoefficients bNC (u,v) are allowed to vary over space; XC, whosecoefficients bC are fixed for all the sample. The intercept a(u,v) is annx1 vector and is estimated for every specific house.

In this paper, we have used a bi-square kernel with adaptivebandwidth. There are different ways to calculate the optimalbandwidth. Here the bandwidth that minimizes the Akaike Infor-mation Criterion (AIC) was selected, following the rule proposed byHurvich, Simonoff, and Tsai (1998).

Additionally, the F-test proposed by Brunsdon et al. (1999) wasapplied to test the fit improvement obtained by MGWR.

Finally, in order to select the global and local variables, a step-wise procedure was performed in the regression. Firstly, all vari-ables were considered local and a spatial variability test wasapplied to each local coefficient (Nakaya, Fotheringham, Brunsdon,& Charlton, 2005). The null hypothesis of this test states that thereis no variability for the parameter across the study region. If therewere any coefficient for which the null hypothesis were notrejected, the corresponding variable would be transformed intoglobal variable and a new MGWR estimation is performed. Theprocess is repeated until the null hypothesis is rejected for all localcoefficients.

a The likelihood ratio test considers as an alternative hypothesis the existence ofdifferences between the proposed spatial model (spatial lag or lag error) and theclassical model.

b The Lagrange multiplier test proposes as an alternative hypothesis for the SEmodel the existence of spatial lag trends from the proposed lag error model; in thecase of the SL model, it proposes the existence of error lag tendencies from theproposed spatial lag model.

c The null hypothesis is that the MGWR model does not improve the OLS model.d Normality and BreuschePagan tests cannot be applied to MGWR models.

3.4. Results

Following the methodology described above, the stepwise andGASIC procedure were applied and compared in order to select thesignificant variables of Table 3 in the hedonic pricing function. Mostchosen variables were identical using either algorithm. The

estimation with the common and uncommon variables obtainedfrom the two algorithms did not improve the fit. The chosen vari-ables obtained from the stepwise procedure were considered, sinceit showed better fit than those from GASIC.

Once the attributes were selected, the econometric estimationof the hedonic price function by using spatial and geographicallyweighted regression model was carried out. The results aredescribed below.

3.4.1. OLS and spatial dependent modelsThe statistical tests of the OLS model are shown in the first

column of Table 4. In order to test the sensitivity of the modelestimation, several weight matrices W were considered assumingdifferent areas where the rental price of a rural house is influencedby the rental price of neighbors. These areas were one, two, five, tennearest neighbors, threshold distance, inverse distance andsquared inverse distance. According to the Lagrange Multipliertests, significant spatial correlation was found with all thesematrices.

The general spatial model was estimated for the different op-tions of weight matrices considered (one, two, five, ten nearestneighbors, threshold distance, inverse distance and squared inversedistance). The first case (one nearest neighbor) included a singularmatrix in the calculations, so the parameters could not be esti-mated. The rest of options showed that the coefficients for thespatial lag and/or the spatial error model were significant, but notsimultaneously in the general spatial model. Therefore, the crite-rion was to estimate the spatial error (SE) and spatial lag (SL)models for all weight matrices above and select the model that bestfitted the data according to the Adjusted Akaike InformationCriteria (AICC). The models with lowest AICC were the spatial lagand the spatial error model with theweight matrix considering onenearest neighbor. This means that price of the nearest house to thehouse in question positively influences the rental price of the lod-ging unit. In general, the sensitivity analysis showed that the largerthe area considered in the weight matrix, the weaker the spatialcorrelation.


Specifically, when analyzing the results obtained by consideringone nearest neighbor, the Lagrange Multiplier tests correspondingboth to the error lag and spatial lag model reject the null hypoth-esis, that is, both models improve the OLS estimation. In this case,the robust versions of the tests need to be considered. Nevertheless,none of these robust tests can be rejected. This result shows thatthere may be other factors that invalidate the asymptotic tests.Therefore, since there exists spatial dependence, there is evidencesupporting the existence of spatial autocorrelation.

The second and third column of Table 4 present the results of thestatistical tests for the SL and SE model with W matrix builtconsidering the influence of one nearest neighbor. As it is shown inthe Table, the Normality assumption is satisfied at 1% of significancefor both models and OLS model as well, while the Homo-skedasticity assumption is satisfied at 5% for the SE and SL models,as compared with 1% for the OLS model. Residuals of both modelsdid not show additional spatial structure in the data.

Table 5 shows the estimation results for the models. The SEmodel presents a positive and significant l coefficient, showing apositive spatial relationship among the errors. The significant co-efficients of the model are the same as those in the classical modeland have similar values. The log likelihood test rejects the nullhypothesis, revealing that the proposed model significantly im-proves the classical one (see Table 4). Additionally, the non-rejection of the null hypothesis from the Lagrange multiplier testshows that there are no alternative spatial structures excluded fromthe model. Regarding the goodness of fit, the AICC shows that theSEmodel improves the classical one. To sum up, taking into accountthe joint results, one may conclude that the SE model offers animprovement of fit with respect to the OLS model.

The SL model presents the spatial lag coefficient r as positiveand significant (see Table 5), showing that neighboring houses havesimilar rental prices. The significant variables are identical to theother models, and the diagnostic tests satisfy the required standardhypotheses. The AICC shows that there exists a minor differencebetween SL and SE models, but that the latter seems to be slightlysuperior.

Table 5Estimation of the parameters in: OLS model (r ¼ l ¼ 0); SE (ls 0, r ¼ 0) and SL model (lneighbor from each house; and MGWR model.

Dependent variable: Log(Price) OLS SE

r

l 0.19D334a 0.98** 1.03D196 1.81 1.81D184 �1.67 �1.43D182 1.06 0.97Intercept 4.07 3.68

StructuralPool 0.12 0.141/Capacity 0.60** 0.84

LocationBarcelona 0.46 0.43Girona 0.18* 0.17*

Npoints_Rural_int_100Km 0.047 0.049Nbeaches_100Km �0.07 �0.07Dist_nearest_beach_Km �0.01 �0.011/Pop5Km 12.73 11.191/Pop100Km 809,515.2 851,225.4

Coefficients are significant at 1%, except those with ** and * which are significant at 5 ana Di indicates a dummy variable for individual i (i ¼ 334, 196, 184, 182). It indicates tb Local variables show confidence intervals for their coefficients.c D%Price when the Geometric mean of prices increases by 1%.d D%Price due to the existence of the proposed category of the qualitative variable wite Semielasticities are calculated through the formula (D%Price/DX).f Elasticities are calculated through the formula (D%Price/D%X) where D% means perce

3.4.2. The MGWR model estimationThe characteristics of the data analyzed may support the hy-

pothesis that parameters of the model vary locally. If thisassumption is considered, the use of the geographically weightedregression is justified. Therefore, this method has been applied tothe same selected variables as the OLS and spatial dependentmodels in order to analyze how each regressor behaves over thestudy area. This analysis has been performed with GWR 4.0 soft-ware. The adaptive bi-squared kernel with a bandwidth of 378individuals minimized AICC and was chosen in the model estima-tion. Therefore, the closest 378 neighbors (out of 393) wereconsidered to estimate the local regressors for each rural house,under the assumption that closer neighbors have more influencethan those located further away.

The stepwise procedure to determine the local and globalvariables, described in the methodology section, was applied. Thespatial non-stationarity of each independent variable wasassessed to determine whether the GWR model improves theglobal regression model. The results showed that the intercept andvariables D334, D196, D184, D182, Barcelona, Girona,1/Pop5Km, and1/Pop100 Km are global and their coefficients estimated by theMGWR are similar to those estimated with the OLS model. Thecoefficients for the rest of attributes are locally estimated with theMGWR model and the results can be found in the last column ofTable 5 and Fig. 2. Since each individual has its own estimations,Table 5 includes the mean value and the standard deviation of thesample of coefficients for each variable, while the graphs in Fig. 2show the range of values of the coefficients located in the map ofCatalonia.

The signs of the estimated coefficients coincide with those ob-tained by the OLS model. From the comparison of AICC, it can bededuced that the MGWR model improves the OLS model estima-tions. This conclusion is further reinforced by the improvement F-test results (Brunsdon et al. 1999) for the MGWR model. The esti-mation results from the SE and MGWR models are used jointly toextract information of how the attributes are valued in the ruralmarket.

¼ 0, rs 0), both considering W the spatial weight matrix that connects the nearest

SL MGWRb Elasticities/semielasticities SE

0.181 0.181%c

0.93** 1.07 180.1%d

1.82 1.92 511.04%d

�1.53 �1.66 �76.06%d

1.01 1.13 163.7%d

3.00 3.74

0.12 0.15 ± 0.08 15%d

0.71 0.64 ± 0.9 �1%e

0.37 0.47 53%d

0.16* 0.29 18%d

0.041 0.039 ± 0.014 4.9%e

�0.06 �0.07 ± 0.018 �7%e

�0.009 �0.01 ± 0.002 �0.57%f

10.83 13.87 �0.0018%f

729,705.8 887,318.2 �0.20%f

d 10%, respectively.he outliers in the sample.

h respect to the category of reference.

ntage increase and X refers to the specific explanatory variable.

Fig. 2. Range of values for coefficients of five local (specific to every individual) variables of the hedonic price model using MGWR estimation. Local R2 is also represented. Thepoints indicate the location of rural houses.


3.4.3. Interpretation of resultsThe estimation of the four models in Table 5 presents identical

significant explanatory variables, with identical sign and similarmagnitudes. These results show that the empirical findings arestable with respect to the use of different econometric approaches.The interpretation of results will be based on the specific estima-tion of the SE model, which exhibits the lowest AICC,

complemented with the local estimations obtained with theMGWR model.

Thus, given the coefficients corresponding to the significantvariables obtained with the SE model, houses located in Barcelonaor Girona have a higher rental prices than those located in the otherprovinces in Catalonia. Specifically, rental price of rural houses inBarcelona are 53% more expensive than those in Lleida and


Tarragona, and rural houses located in Girona are 18% moreexpensive than those from the reference provinces. In the case ofBarcelona, this can be interpreted as an added value to the prox-imity of the main source of demand. As commented in the previoussection, almost 90% of rural tourism demand is local and Barcelonais by far the most populated province of Catalonia (73% of totalpopulation in 2009). The results show that distance from home istranslated into a cost for potential tourists to rural areas. This idea isreinforced by coefficients for other location attributes, which arecommented below.

Girona hosts the larger percentage or rural tourists in Catalonia(40.71% in 2009). The province is highly endowed with natural andcultural attractions, since it combines mountain forests withtraditional agrarian landscapes. Additionally, it is connected to themain transportation routes in Catalonia. In the last decades, theseconditions have motivated urban residents in Catalonia to acquiresecond residences inland, preferring traditional isolated buildings(masias) or village houses (Solana-Solana, 2010). The result is astronger pressure on the housing market. The increase of secondresidences in rural areas is not exclusive for Girona, but it is moremarked here than in the rest of provinces in Catalonia. This phe-nomenon and the specific attractions of Girona commented abovemake the rental price of a rural house higher here than in the otherprovinces. Nevertheless, the distribution of second residences is notuniform across districts in Girona, varying from 4.77% (Pla de l’Es-tany district) to 45.06% (Alt Empord�a district) of total residences(Idestac, 2014). Consequently, it is expected that their effect onrental price of rural houses is heterogeneous, which can at leastpartially explain the low values of local R2 obtained in this province(Fig. 2f).

Regarding the structural variables, Pool and 1/Capacity show theexpected sign according to the previous literature. The rentalhouses with swimming pool are 15% more expensive than thosewithout it, (e0.14�1)100 z 15, and capacity slightly reduces therental price per two people of the house.

The negative sign of the coefficients for Nbeaches_100Km andDist_nearest_beach_Km in principle looks contradictory. This resultindicates that the excess of beaches within a certain radius from thehouse is negatively valued by the market, but on the contrary theproximity to the nearest beach is positively valued. The latterfinding can be interpreted in the same line as the coefficients forBarcelona. The demand comes mainly from the coastal areas ofCatalonia, where most of the inhabitants reside (see Fig. 1). So,distance to the nearest beach is a proxy to the distance from themajority of visitors' homes, which is penalized by tourists. Thus, thecloser the home the higher the rental price of the rural lodging. Thisinterpretation is strengthened by the estimation of the local co-efficients of Dist_nearest_beach_Km, shown in Fig. 2e. The effect ofdistance is more pronounced for rural houses in the populatedprovince of Barcelona and the coastal areas of Girona than in Lleidaand the interior of Tarragona.

The influence of the number of beaches in a certain radius fromthe house on the house's rental price connects rural and coastaltourism. The negative sign points to opposite roles of the beach inboth types of tourism in Catalonia. While coastal accommodationunits increase their rental prices when they are close to the beach,rural houses aremore positively valued if they are located far from anumber of beaches.

Each additional rural point of interest within a 100 Km radiusincreases the rental price by 4.9%. This result indicates thatrural attractions are explicitly valued by the market, whichagrees with previous analysis of rural tourism (Sharpley &Sharpley, 1997). Two types of rural tourism are found fromobserving the local coefficients of Npoints_rural_int_100Km(Fig. 2c) and Nbeaches_100 Km (Fig. 2d), one in the Northern and

another in the Southern part of Catalonia. Both have the samesign for the valuation of rural attractions and the number ofbeaches in their surroundings. However, the former appreciatesthe rural attractions in a higher degree and dislikes beaches in ahigher degree than the latter. Swimming pools are also moreappreciated for rural tourists in the Southern part of Catalonia, asopposed to the North.

Finally, congestion detracts value to the rental price of ruralhouses. This is deduced by the coefficients of the inverse variable ofpopulation within 5 and 100 Km radii. In particular, an increase of1% of population within the radius of 5 Km from a house reducesthe rental price by 0.0018%, while the decrease is 0.20% when theradius is 100 Km.

4. Discussion and implications

4.1. Discussion

The empirical analysis above sought to identify specific char-acteristics of rural tourism in a region of the Mediterranean area(Catalonia) where mature beach resorts are present. New attributesfor rural tourism were found, some of them not previously un-covered by applying hedonic pricing methodology. Some of theseattributes are similar or related to others found in hedonic modelsfor coastal destinations in the same region (Table 1a). From thecomparison of results in the case study with the determinantsidentified for mass tourism in the previous literature, common anduncommon links between the two branches of the industry inCatalonia can be detected. The attributes are summarized below,grouped in the same four classes presented in Section 2.3.

� Structural variables: Similar to other regions in SouthernEurope, some facilities traditionally valued by sun-and-sandtourist, such as the availability of a swimming pool, are alsoappreciated by rural tourists. This finding was rather expected,since the high season for rural andmass tourism is summertime,where the highest temperatures throughout the year are felt.Thus, the result leads to think that modern rural tourists inCatalonia and throughout the whole area look for a kind of non-conventional rural comfort.

� Congestion: The negative influence of population/room densityappears again as a common link between the two markets. Thisresult indicates that both coastal and rural tourists appreciate acertain level of isolation, although to a different extent. It isexpected that the tolerance to congestion be much higher inmass destinations than in rural areas. Nevertheless, the findingemphasizes the limited capacity of rural tourism to increase thetourist affluence.

� Coastal attractions: The distance to the nearest beach is also acommon attribute in both markets, although with oppositevaluation. The previous analyses applied to coastal resorts in theSpanish region have found that the proximity to the beach is oneof the most appreciated characteristics of a hotel room. How-ever, in the case study, although rural houses near the coast aremore valued than those located inland, beaches appear as adisamenity for rural tourists. These observations reveal that thetwo types of tourism are directed to different market segments,with opposite consumer preferences for the same attribute.

� Natural and cultural attractions: The results agree with previousfindings and show that these attractions are positive amenitiesfor rural tourism. Several previous studies for coastal tourism inCatalonia have found that tourists also appreciate accommo-dation units near natural areas and other cultural attractions.

Regarding the two last points, the case study shows that the


valuation of these attributes is not homogenous throughout thewhole region. Specifically, two areas in Catalonia are identifiedwhere rural tourism presents different characteristics: One in theNorthern region and another in the Southern region. The former ischaracterized by higher valuation of specific rural attractions thanthe latter, which gives a greater value to the availability of swim-ming pools and dislike to a lesser extent the proximity to beaches.There are marked climatic differences between these two areas.The Pyrenees are in the North of Catalonia, so this area presentshigher rainfall index andmilder weather in summer than the Southof Catalonia. Therefore, the weather and orography seem todetermine the valuation of attributes in the two areas.

The comparison of attributes also provides information aboutthe inter-relationship between both types of tourism. In otherwords, since the two activities take place in nearby areas and sharesome characteristics, tourist behavior or general functioning in onetourist sector can disturb the performance of the other. The influ-ence can be exerted in both directions, from mass to rural tourismand vice versa.

Some insights about the influence of mass tourism on ruraltourism in Catalonia can be extracted from the previous analysis.The third point above states that rural houses close to the coastalareas are more appreciated than those located far from these.Taking into account that the most rural tourists come mainly fromthe coastal areas of the region, which concentrates most of the localpopulation, the finding can be explained from the consumers' side.Rural tourists take into account the travel cost to get to the touristdestination when planning their rural vacations and this is trans-lated to rental prices. The capital of Catalonia (Barcelona) is on thecoast but a significant amount of population live in the surround-ings of beach resorts outside the district of Barcelona (darker areain Fig. 1), which are sources of employment for the region. Thus,given this interpretation of the results, the existence of masstourism in the region has favored the development of rural tourismin some areas with respect to others. In particular, rural tourisminland of beach resorts enjoys a comparative advantage withrespect to rural tourism in more remote areas, due precisely to theexistence of mass tourism resorts on the coast.

Other point of inter-relationship between sun-and-sand andrural tourism is derived from the fourth point above, the appreci-ation of natural and cultural attractions, which is shared by the twotypes of tourism. Unlike other attributes which can be reproducedin different locations (e.g. swimming pool), these attractions arenecessarily linked to a specific geographic location, such as a nat-ural park, or inclusive to a location andmoment, such as an event orfestival. Consequently, all tourists must necessarily share the sameplace (maybe at the same time) to enjoy the attraction. Since theamount of potential visitors from the coastal areas is high, thequality of the product can be reduced in several aspects: one, due tothe degradation of the natural or esthetic values of the attractions,which usually arises from frequent visits or massive events; andtwo, due to the increase of population density, which is an unde-sirable factor in both types of tourism. The latter effect would harmmostly rural tourism, which is sensible to congestion, as theempirical analysis has revealed.

In connection with the last point, rural tourism developmentcan also exert some influence onmass tourism.When implementedin a sustainable way, rural tourism helps to preserve the environ-ment and the traditional architecture. This benefit would produce apositive externality for visitors from the coastal areas, who canenjoy the improvements in the rural landscape derived from ruraltourism activity.

4.2. Implications

The findings above can be transferred into some policy andmanagerial recommendations for rural and mass tourism in Cata-lonia. Firstly, the negative effect of congestion in both types oftourism should lead to consider the existence of limits to tourismgrowth in the region. Rural tourism is an economic opportunity tobring income derived from the tourism industry inland. Alongthese lines, it should remain a low-affluence type of tourism withhigh spending rates per person. Policy regulations in the futureshould be laid down in order to follow this strategy.

Secondly, the analysis has found that some amenities for coastaltourists can be the opposite for rural tourists, as is the case of thebeach proximity. Nevertheless, the proximity to the coast, wheremost of transport or urban centers are located, is also positivelyvalued by rural tourists. These apparently paradoxical preferencesshould be managed by local authorities and stakeholders. Thepromotional campaigns for rural tourism can include both attrac-tions by highlighting the major attractions of the countryside,which are, isolation, tranquility, wilderness and rural life, avoidingthe reference to crowded beaches and services on them, while alsoindicating the proximity to urban centers in geographic distance ortime.

From the two areas in Catalonia where rural tourism presentsclear differences, the Southern part highly values some non-ruralspecific facilities, such as swimming pools, while the reluctanceto coastal resorts is not so remarkable than in the Northern part.Tarragona is located in the South and has experienced a strongincrease in foreign visitors in the last five years, as observed inTable 5. The origin and specific profile of rural tourists, evidencedby the results (preference for some non-rural facilities, closeness tothe coast), can also be taken into account in the promotional stra-tegies for local stakeholders.

5. Conclusions

The aim of this paper was to find common and uncommon at-tributes between the traditional coastal tourism and the recentalternative rural tourism developed in some destinations ofSouthern Europe. The literature review of hedonic price modelsapplied to both types of tourism reveals that similar characteristicsare positively valued in both markets, such as some facilities in theaccommodation unit (swimming pool), natural/cultural attractionsor the low population density in the surroundings. Nevertheless,these conclusions have to be taken with some caution, since thecomparison is made from empirical analysis applied to differentspecific destinations in the Mediterranean area. The extrapolationof the particular results to the whole region is based on similaritiesshown by the tourist industry in the area.

In order to obtain more in depth insights on the inter-relationship among both types of tourism, the case study applieda hedonic price model to rural tourism in Catalonia. In this context,rural tourist is restricted to those visitors staying in a rural house.The destination Catalonia was chosen due to the availability ofdisaggregated data about the characteristics of rural houses,together with the existence of several previous studies applying thesame methodology to mass coastal tourism in this Spanish region.The use of spatial econometrics and geographical weightedregression allowed the obtainment of location-specific estimationson the influence of the attributes on the rental price of rural houses.

The results for the case study reinforce some of the conclusionsderived from the literature review. Specifically, swimming pools areclearly appreciated by both rural and coastal tourism. Congestion,measured as population density in the surroundings, appears againas a disamenity for rural tourists in Catalonia, which is also shared


with mass tourists. One of the novelties is the significant influenceof beaches near rural houses on their rental price. Unlike coastaltourists, the beach is not an attraction for rural tourists in Catalonia.Instead, locations far from mass tourism centers are preferred.However, results also reveal that the proximity to visitors' homeand transport services, which have been developed near the masstourist destinations (the town of Barcelona and beach resorts), isappreciated by rural tourists. These common links makes the twosectors influence each other, positively in certain aspects andnegatively in others.

Some general conclusions about the coexistence between massand rural tourism in Catalonia can be also derived from the analysis.The opposite valuation of the proximity to the beach by rural andmass tourism indicates that the two branches of the industry do notcompete for this attribute and therefore can be developed withoutconfrontation in this point. Rural tourism is not affected by theaffluence to the beach and vice-versa. However, other attributes,such as cultural or natural attractions, are positively valued by bothmarkets. Consequently, the common exploitation of the resource(affluence of visitors, construction of infrastructures) can lead tonegative externalities suffered by one or the two branches. In thiscase, rural tourism can be more deeply harmed ought to it is highlysensible to congestion. The general conclusion is that, more than asimple coexistence, a kind of symbiosis between both types oftourism is possible, although it can be turned into predator/preyrelationship, with rural tourism being the prey, if not carefullymanaged.

The study presented in this paper also includes several limi-tations which must necessarily be mentioned. The chosenmethodology to find the characteristics of both types of tourism(hedonic price model) assumes very restrictive hypotheses(perfect competition in equilibrium, no information barriers).Although rural tourism market in Catalonia presents character-istics to support theses hypotheses, the results must be naturallytaken with caution, as usual in any application of hedonic pricingapproach. Additionally, the methodology only allows quantitativeattributes. However, some relevant attractions of mass and ruraltourism cannot be easily converted into quantitative variables.For example, this is the case of beach quality, heritage or events.These and other similar factors are not considered in the hedonicprice estimation of the case study. They may explain partially thelow adjustment level of the estimated econometric models.

Although the case study has focused on tourism in Catalonia, thegeneral conclusions above may apply to other regions in theMediterranean, which present similar characteristics to the casestudy, these being traditional mature coastal tourism and incipientrural tourism development. Nevertheless, new empirical studiesthat compare both types of tourism in other parts of the regionwould help to increase knowledge on this matter.

Acknowledgments

The authors thank three anonymous referees for their valuablecomments to a previous version of the paper and also Rosy andSoraya Quinga for their help in the data collection. The work waspartially financed by project ECO2008-05589/ECON and projectSEJ2006-15408/ECON from the Ministry of Education, Science andTechnology of the Government of Spain.

References

Acosta-Gonz�alez, E., & Fern�andez-Rodríguez, F. (2007). Model selection via geneticalgorithms illustrated with cross-country growth data. Empirical Economics,33(2), 313e337. http://dx.doi.org/10.1007/s00181-006-0104-3.

Aguil�o, E., Alegre, J., & Sard, M. (2003). Examining the market structure of theGerman and UK tour operating industries through an analysis of package

holiday prices. Tourism Economics, 9(3), 255e278. http://dx.doi.org/10.5367/000000003101298385.

Alegre, J., & Cladera, M. (2006). Repeat visitation in mature sun and sand Holidaydestinations. Journal of Travel Research, 44(3), 288e297. http://dx.doi.org/10.1177/0047287505279005.

Alegre, J., Cladera, M., & Sard, M. (2013). Tourist areas: examining the effects oflocation attributes on tour-operator package holiday prices. Tourism Manage-ment, 38, 131e141. http://dx.doi.org/10.1016/j.tourman.2013.02.011.

Andriotis, K. (2006). Researching the development gap between the hinterland andthe coastdevidence from the island of Crete. Tourism Management, 27(4),629e639.

Anselin, L. (2005). Exploring spatial data with GeoDa: A workbook. Revised versionMarch 6, 2005. Center for Spatially Integrated Social Science. University of Illi-nois, Urbana-Champaign. Available in http://www.csiss.org.

Barke, M. (2004). Rural tourism in Spain. International Journal of Tourism Research,6(3), 137e149.

Bastian, C. T., McLeod, D. M., Germino, M. J., Reiners, W. A., & Blasko, B. J. (2002).Environmental amenities and agricultural land values: a hedonic model usinggeographic information systems data. Ecological Economics, 40(3), 337e349.http://dx.doi.org/10.1016/S0921-8009(01)00278-6.

Bramwell, B. (2004). Mass tourism, diversification and sustainability in SouthernEurope's coastal regions. In B. Bramwell (Ed.), Coastal mass tourism: Diversifi-cation and sustainable development in Southern Europe (pp. 1e21). Channel ViewPublications.

Bright, E., Coleman, P., King, A., & Rose, A. (2008). LandScan 2007. Oak Ridge NationalLaboratory. http://www.ornl.gov/landscan/.

Brunsdon, C., Fotheringham, A. S., & Charlton, M. E. (1996). Geographically weightedregression: a method for exploring spatial nonstationarity. Geographical Anal-ysis, 28(4), 281e298. http://dx.doi.org/10.1111/j.1538-4632.1996.tb00936.x.

Brunsdon, C., Fotheringham, A. S., & Charlton, M. (1999). Some notes on parametricsignificance tests for geographically weighted regression. Journal of RegionalScience, 39(3), 497e524. http://dx.doi.org/10.1111/0022-4146.00146.

Bujosa-Bestard, A., & Font, A. R. (2009). Environmental diversity in recreationalchoice modelling. Ecological Economics, 68(11), 2743e2750. http://dx.doi.org/10.1016/j.ecolecon.2009.05.016.

C�anoves, G., Herrera, L., & Cuesta, L. (2005). Turismo rural en Catalu~na: con-dicionantes de la oferta y la demanda. Investigaciones Geogr�aficas, 37, 5e26.

Corine Land Cover. (2006). Coordination of information on the environment. Euro-pean Environment Agency. http://www.eea.europa.eu/publications/COR0-landcover.

Cox, L. J., & Vieth, G. R. (2003). Hotel investment in open area. Annals of TourismResearch, 30(2), 342e352. http://dx.doi.org/10.1016/S0160-7383(02)00059-2.

Espinet, J. M., Saez, M., Coenders, G., & Fluvi�a, M. (2003). Effect on prices of theattributes of holiday hotels: a hedonic prices approach. Tourism Economics, 9(2),165e177. http://dx.doi.org/10.5367/000000003101298330.

Farmaki, A. (2012). An exploration of tourist motivation in rural settings: the case ofTroodos, Cyprus. Tourism Management Perspectives, 2e3, 72e78. http://dx.doi.org/10.1016/j.tmp.2012.03.007.

Fleischer, A. (2012). A room with a viewdA valuation of the Mediterranean Seaview. Tourism Management, 33(3), 598e602. http://dx.doi.org/10.1016/j.tourman.2011.06.016.

Fleischer, A., & Tchetchik, A. (2005). Does rural tourism benefit from agriculture?Tourism Management, 26(4), 493e501. http://dx.doi.org/10.1016/j.tourman.2003.10.003.

Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2003). Geographically weightedregression: The analysis of spatially varying relationships. John Wiley & Sons.

Garay, L., & C�anoves, G. (2011). Life cycles, stages and tourism history: the Catalonia(Spain) experience. Annals of Tourism Research, 38(2), 651e671. http://dx.doi.org/10.1016/j.annals.2010.12.006.

Generalitat de Catalunya. (2014a). Resum de resultats anuals del turisme. Any 2013http://www20.gencat.cat/portal/site/observatoritreball.

Generalitat de Catalunya. (2014b). Parcs de Catalunya. http://parcsnaturals.gencat.cat/ca/.

Getz, D. (1998). Developing rural tourism: the potential of beach resort hinterlands.In B. Faulkner, C. Tidswell, & D. Weaver (Eds.), Progress in tourism and hospitalityresearch, Proceedings of the eighth Australian hospitality and research conference(pp. 700e714). Gold Coast, Australia Bureau of Tourism Research.

Hamilton, J. M. (2007). Coastal landscape and the hedonic price of accommodation.Ecological Economics, 62(3e4), 594e602. http://dx.doi.org/10.1016/j.ecolecon.2006.08.001.

Hern�andez, J. M., & Casimiro, L. A. (2012). Simulation model for a joint mass/ruraltourism system. Tourism and Hospitality Research, 12(1), 5e14.

Hurvich, C. M., Simonoff, J. S., & Tsai, C.-L. (1998). Smoothing parameter selection innonparametric regression using an improved Akaike information criterion.Journal of the Royal Statistical Society: Series B (Statistical Methodology), 60(2),271e293. http://dx.doi.org/10.1111/1467-9868.00125.

Idestac. (2014). Turismo rural. Viajeros y pernoctaciones por procedencia, 2013. http://www.idescat.cat/es/economia/ecotur.html.

INE. (2014). Encuesta de ocupaci�on de alojamientos turísticos, 2009, 2013. http://www.ine.es/inebmenu/mnu_hosteleria.htm.

Kizos, T., & Iosifides, T. (2007). The contradictions of agrotourism development inGreece: evidence from three case studies. South European Society & Politics,12(1), 59e77.

Kozak, M. (2002). Comparative analysis of tourist motivations by nationality anddestinations. Tourism Management, 23(3), 221e232. http://dx.doi.org/10.1016/

http://dx.doi.org/10.1007/s00181-006-0104-3

http://dx.doi.org/10.5367/000000003101298385

http://dx.doi.org/10.5367/000000003101298385

http://dx.doi.org/10.1177/0047287505279005

http://dx.doi.org/10.1177/0047287505279005


http://refhub.elsevier.com/S0261-5177(15)30035-2/sref5





http://www.csiss.org














http://www.ornl.gov/landscan/

http://dx.doi.org/10.1111/j.1538-4632.1996.tb00936.x

http://dx.doi.org/10.1111/0022-4146.00146

http://dx.doi.org/10.1016/j.ecolecon.2009.05.016








http://www.eea.europa.eu/publications/COR0-landcover

http://www.eea.europa.eu/publications/COR0-landcover

http://dx.doi.org/10.1016/S0160-7383(02)00059-2

http://dx.doi.org/10.5367/000000003101298330

http://dx.doi.org/10.1016/j.tmp.2012.03.007

http://dx.doi.org/10.1016/j.tmp.2012.03.007








http://dx.doi.org/10.1016/j.annals.2010.12.006

http://dx.doi.org/10.1016/j.annals.2010.12.006

http://www20.gencat.cat/portal/site/observatoritreball

http://parcsnaturals.gencat.cat/ca/

http://parcsnaturals.gencat.cat/ca/












http://dx.doi.org/10.1111/1467-9868.00125

http://www.idescat.cat/es/economia/ecotur.html

http://www.idescat.cat/es/economia/ecotur.html

http://www.ine.es/inebmenu/mnu_hosteleria.htm

http://www.ine.es/inebmenu/mnu_hosteleria.htm






http://dx.doi.org/10.1016/s0261-5177(01)00090-5


s0261-5177(01)00090-5.Lovell, M. (1983). Data mining. Review of Economics and Statistics, 65, 1e12.Nakaya, T., Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2005). Geographically

weighted Poisson regression for disease association mapping. Statistics inMedicine, 24(17), 2695e2717. http://dx.doi.org/10.1002/sim.2129.

Ohe, Y., & Ciani, A. (2011). Evaluation of agritourism activity in Italy: facility based orlocal culture based? Tourism Economics, 17(3), 581e601. http://dx.doi.org/10.5367/te.2011.0048.

Papatheodorou, A. (2002). Exploring competitiveness in Mediterranean resorts.Tourism Economics, 8(2), 133e150. http://dx.doi.org/10.5367/000000002101298034.

Portolan, A. (2013). Impact of the attributes of private tourist accommodation fa-cilities onto prices: a hedonic price approach. European Journal of TourismResearch, 6(1), 74e82.

Rigall-I-Torrent, R., & Fluvi�a, M. (2011). Managing tourism products and destina-tions embedding public good components: a hedonic approach. Tourism Man-agement, 32(2), 244e255. http://dx.doi.org/10.1016/j.tourman.2009.12.009.

Rigall-I-Torrent, R., Fluvi�a, M., Ballester, R., Sal�o, A., Ariza, E., & Espinet, J.-M. (2011).The effects of beach characteristics and location with respect to hotel prices.Tourism Management, 32(5), 1150e1158. http://dx.doi.org/10.1016/j.tourman.2010.10.005.

Roberts, L., & Hall, D. R. (2001). Rural tourism and recreation: Principles to practice.Wallingford, Oxon, UK ; New York, NY: CABI Pub.

Rosen, S. (1974). Hedonic prices and implicit markets: product differentiation inpure competition. Journal of Political Economy, 82(1), 34e55. http://dx.doi.org/10.2307/1830899.

Sal�o, A., & Garriga, A. (2011). The second-home rental market: a hedonic analysis ofthe effect of different characteristics and a high-market-share intermediary onprice. Tourism Economics, 17(5), 1017e1033. http://dx.doi.org/10.5367/te.2011.0074.

Sal�o, A., Garriga, A., Rigall-I-Torrent, R., Vila, M., & Fluvi�a, M. (2014). Do implicitprices for hotels and second homes show differences in tourists' valuation forpublic attributes for each type of accommodation facility? International Journalof Hospitality Management, 36, 120e129. http://dx.doi.org/10.1016/j.ijhm.2013.08.011.

Sanchez-Ollero, J. L., Garcia-Pozo, A., & Marchante-Mera, A. (2013). How doesrespect for the environment affect final prices in the hospitality sector? A he-donic pricing approach. Cornell Hospitality Quarterly, 55(1), 31e39. http://dx.doi.org/10.1177/1938965513500709.

Santana-Jim�enez, Y., & Hern�andez, J. M. (2011). Estimating the effect of over-crowding on tourist attraction: the case of Canary Islands. Tourism Management,32(2), 415e425. http://dx.doi.org/10.1016/j.tourman.2010.03.013.

Santana-Jim�enez, Y., Hern�andez, J. M., & Su�arez-Vega, R. (2013). Estimating theenvironmental attractions of rural tourism lodging units in La Gomera island,Spain. European Journal of Tourism Research, 6(2), 132e140.

Santana-Jim�enez, Y., Su�arez-Vega, R., & Hern�andez, J. M. (2011). Spatial and envi-ronmental characteristics of rural tourism lodging units. Anatolia, 22(1),89e101. http://dx.doi.org/10.1080/13032917.2011.556223.

Sharpley, R. (2002). Rural tourism and the challenge of tourism diversification: thecase of Cyprus. Tourism Management, 23(3), 233e244. http://dx.doi.org/10.1016/s0261-5177(01)00078-4.

Sharpley, R., & Sharpley, J. (1997). Rural tourism: An introduction. London etc. In-ternational Thomson Business Press

Sheng, L. (2011). Specialisation versus diversification: a simple model for touristcities. Tourism Management, 32(5), 1229e1231. http://dx.doi.org/10.1016/j.tourman.2010.09.012.

Solana-Solana, M. (2010). Rural gentrification in Catalonia, Spain: a case study ofmigration, social change and conflicts in the Empordanet area. Geoforum, 41(3),508e517. http://dx.doi.org/10.1016/j.geoforum.2010.01.005.

Su�arez-Vega, R., Acosta-Gonz�alez, E., Casimiro-Reina, L., & Hern�andez, J. (2013).Assessing the spatial and environmental characteristics of rural tourismlodging units using a geographical weighted regression model. In �A. Matias,

P. Nijkamp, & M. Sarmento (Eds.), Quantitative methods in tourism economics(pp. 195e212). Physica-Verlag HD.

Taylor, L. (2008). Theoretical foundations and empirical developments in hedonicmodeling. In A. Baranzini, J. Ramirez, C. Schaerer, & P. Thalmann (Eds.), Hedonicmethods in housing markets (pp. 15e37). New York: Springer.

Thrane, C. (2005). Hedonic price models and sun-and-beach package tours: theNorwegian case. Journal of Travel Research, 43(3), 302e308. http://dx.doi.org/10.1177/0047287504272034.

Turnovsky, S. J. (1974). Technological and price uncertainty in a ricardian model ofinternational trade. The Review of Economic Studies, 41(2), 201e217. http://dx.doi.org/10.2307/2296713.

Vilchez, J. R. (2013). Destination and seasonality valuations: a quantile approach.Tourism Economics, 19(4), 835e853. http://dx.doi.org/10.5367/te.2013.0297.

Yoon, Y., & Uysal, M. (2005). An examination of the effects of motivation andsatisfaction on destination loyalty: a structural model. Tourism Management,26(1), 45e56. http://dx.doi.org/10.1016/j.tourman.2003.08.016.

Dr. Juan M. Hern�andez is an Associate Professor at theUniversity of Las Palmas de Gran Canaria, Spain, and at theUniversity Institute of Tourism and Sustainable Develop-ment (TIDES). His main line of research is the developmentand application of quantitative methods to naturalresource economics. In tourism research, his major in-terests are the economic growth of tourism economies andthe valuation of environmental attractions in touristdestinations.

Dr. Rafael Su�arez-Vega is an Associate Professor at theUniversity of Las Palmas de Gran Canaria, Spain, and theUniversity Institute of Tourism and Sustainable Develop-ment (TIDES). His main research areas are competitivelocation problems and the application of Geographic In-formation Systems (GIS) to solve these problems. He hasalso applied GIS in combination with spatial econometricstechniques to determine the most valuable characteristicsof rural tourism lodging units.

Dr. Yolanda Santana-Jim�enez is an Associate Professor ofEconometrics at the University of Las Palmas de GranCanaria. Her research interests are Financial Econometrics,with some published works on exchange rate risk. In thelast years she is more oriented towards Tourism, dealingwith topics such as overcrowding or rural tourism and theapplication of spatial econometric techniques andGeographic Information System programs.

http://dx.doi.org/10.1016/s0261-5177(01)00090-5



http://dx.doi.org/10.1002/sim.2129

http://dx.doi.org/10.5367/te.2011.0048

http://dx.doi.org/10.5367/te.2011.0048

http://dx.doi.org/10.5367/000000002101298034

http://dx.doi.org/10.5367/000000002101298034










http://dx.doi.org/10.2307/1830899

http://dx.doi.org/10.2307/1830899

http://dx.doi.org/10.5367/te.2011.0074

http://dx.doi.org/10.5367/te.2011.0074

http://dx.doi.org/10.1016/j.ijhm.2013.08.011

http://dx.doi.org/10.1016/j.ijhm.2013.08.011

http://dx.doi.org/10.1177/1938965513500709

http://dx.doi.org/10.1177/1938965513500709









http://dx.doi.org/10.1080/13032917.2011.556223

http://dx.doi.org/10.1016/s0261-5177(01)00078-4

http://dx.doi.org/10.1016/s0261-5177(01)00078-4





http://dx.doi.org/10.1016/j.geoforum.2010.01.005















http://dx.doi.org/10.1177/0047287504272034

http://dx.doi.org/10.1177/0047287504272034

http://dx.doi.org/10.2307/2296713

http://dx.doi.org/10.2307/2296713

http://dx.doi.org/10.5367/te.2013.0297


the inter-relationship between rural and mass tourism: … · case study the inter-relationship...

Documents