Diverse networks in regional tourism: ties that foster and hinder the development of rural-urban tourism collaboration in Estonia

This study investigates the collaborative ties between rural and urban tourism enterprises and their networks in the post-communist Pärnu region of Estonia. For that, two research questions were answered: what is the nature of collaborative ties between entrepreneurs and sub-networks in a tourist region and how different ties between regional tourism networks foster and hinder the development process of rural-urban tourism network? This study bases on stakeholder theory, actor-network theory, social network analysis, and social network theory, uses multigrounded theory as a methodology, and qualitative data from semi-structured interviews with the regional tourism actors. Results indicate mutually beneficial relationships between prominent and small urban enterprises. Prominent urban enterprises gain by services offered by small enterprises, the latter benefit by the visitors brought in the town by the big ones. Regular collaboration is a common practice among rural enterprises. However, a collaboration between rural and urban entrepreneurs and networks at the regional level is minimal. However, some collaborative ties between tourism enterprises in a regional network exist between urban and rural entrepreneurs in different collaboration levels. The urban tourism network influences regional tourism much more than rural networks does.


Introduction
In tourism, collaboration networks help to allocate resources, offer engaging experiences to visitors, and increase entrepreneurs' resilience to shocks (Jesus & Franco, 2016;Luthe & Wyss, 2014;Pechlaner & Volgger, 2012; Van der Zee et al., 2017). According to Wood and Gray (1991, p. 146), "Collaboration occurs when a group of autonomous stakeholders of a problem domain engage in an interactive process, using shared rules, norms and structures, to act or decide on issues related to that domain".
A tourism network is a complex system where nodes (e.g., organisations, groups, and individuals) and ties (e.g., communication, agreements, and relationships) (Wasserman & Faust, 1994) form a system in which different entrepreneurs, through their position, decisions, behaviour, or attitudes, affect each other and the whole network (Fyall et al., 2012;Wellman, 1988). Networking affects several complex processes: operation of value chains, governance, downscaling, outsourcing, and production chains (Porter, 1990). During the networking process, network members work together to achieve a collective aim but maintain control over their individual resources (Baggio et al., 2010a).
Large tourist regions usually comprise both, rural and urban settlements. Regionwide rural and urban networking helps to share knowledge, resources and information to fulfil strategic aims that are unreachable to single entrepreneurs (Beritelli, 2011) or small networks, alleviate fragmentation and geographical spread problems in tourism (Scott et al., 2008) and improve competitiveness in the global tourism market and attract potential investors (Tremblay, 1998). The tourist region can benefit from networking via economic growth and social benefits to local communities (Novelli et al., 2006). However, collaboration networks between tourism entrepreneurs often occur in small and local formations characterised by frequent socialisation and a high level of mutual trust. Mutually beneficial tourism collaboration is problematic even in small networks as tourism entrepreneurs that occasionally collaborate are simultaneously competing with each other (Van der Zee et al., 2017).
While regional collaborative networking is necessary to achieve strategic aims in the tourist regions, it is a challenge to launch and sustain a collaboration network (Saxena et al., 2007). Rural and urban entrepreneurs operate their business and socialise in different environments. Urban tourism typically has a larger scale, and it involves both large-and small-scale enterprises. Concurrently, rural tourism entrepreneurs often are small family-owned or lifestyle enterprises, which have strong ties within local communities (Bredvold & Skalen, 2016;Dimitrovski et al., 2012). Relationships between entrepreneurs play a central role in networks (Jesus & Franco, 2012;Timur & Getz, 2008) and determine the success of the collaboration (Raab & Kenis, 2009). Large tourism enterprises typically establish formalised relationships (Czernek et al., 2017), while a trust-based informal collaboration is usually present between small tourism enterprises (Pilving et al., 2019). In some (post-communist) countries, low trust in society complicates the formation of tourism collaboration networks (Bjørnskov, 2007). In this setting, lack of a collaborative experience and tradition may hinder networking (Czernek, 2013).
Studies on networks in tourism have used different theoretical approaches. According to Nguyen et al. (2019), stakeholder theory (ST), actor-network theory (ANT), and social network analysis (SNA) are the most widely used approaches in the tourism networks research. ST contends that the success of a tourist destination and entrepreneurs' involvement in tourism networks are mutually dependent. Entrepreneurs have different levels of salience (Mitchell et al., 1997), and the tourist destination management must understand and address their interests (Freeman, 1984;Jones & Wicks, 1999;Mowforth & Munt, 2016). However, ST does not focus on the interconnections and relationships between entrepreneurs and sub-networks (Beritelli, 2011;Nguyen et al., 2019). According to ANT, networks are heterogeneous and typically formed by a principal actor who activates others (Dedeke, 2017;Ren, 2010). ANT focuses on the mutual influences between actor groups and allows exploring problem identification and goal setting (Vicsek et al., 2016). However, the use of ANT in network research is subject to limitations. The actor-network is constantly changing, generating the risk of endless chains of associations (Nguyen et al., 2019). Several studies (e.g., Baggio et al., 2010b;Dredge, 2006;Hristov et al., 2018;Tinsley & Lynch, 2001) use SNA, "a strategy for investigating social structures" (Otte & Rousseau, 2002, p. 441) to explain how actors share information in the network via direct and indirect links. However, SNA has limitations in explaining the processes and dynamics inside the network (Albrecht, 2013) and disregards the role of the individual influences on the network if used quantitatively (Ahrens, 2018).
Social network theory (SNT) addresses the limitation of ST by identifying different actors and by explaining the nature of ties within a network (Granovetter, 1985). Jesus and Franco (2016) used SNT to study regional rural-urban tourism cooperation networks between hotels in the inland regions of Portugal, focusing on cooperation habits, the structure of relationships, and entrepreneurs' perception on the role of network structures. However, there is a gap in academic literature related to problems of the development of regional collaboration network between urban and rural tourism enterprises.
To address this gap, the study aims to answer the following research questions: (1) what is the nature of collaborative ties between entrepreneurs and sub-networks in a tourist region; (2) how different ties between regional tourism networks foster and hinder the development process of rural-urban tourism network?
Because of the complexity of the development of networks in rural and urban tourism, this study employs key principles of ST, ANT, SNA and SNT. To avoid limiting the study in the early research stages with a too narrow theoretical frame, an in-depth qualitative approach based on the multigrounded theory (MGT) is applied. MGT allows developing or complementing a theory using empirical data. The use of MGT includes inductivism and deductivism and theoretical, empirical, and internal grounding without allowing the existing theory to limit the use of the data (Cronholm & Goldkuhl, 2010). Therefore, this approach is particularly suitable for addressing complex networking phenomena. The research was carried out, and data were collected using semi-structured interviews with local tourism stakeholders in the Pärnu region of post-communist Estonia. In the region, stakeholders in tourism networks can belong to different (private, public and non-profit) sectors. However, this study focuses on networking between entrepreneurs from the private sector. Pärnu town and surrounding rural area are situated in western Estonia. Its neighbours are Rapla County in the north, Lääne County in the north-west, Viljandi County in the northeast, and Saare County in the west. The southern area of the county neighbours the Republic of Latvia. Pärnu area consists of seven municipalities from which Lääneranna, Põhja-Pärnumaa, Tori, Saarde, Häädemeeste and Kihnu are rural, and Pärnu town is an urban municipality (Local Authorities of Pärnu County, 2020). The area has a population of 86,165 (2020) residents (Pärnu County 2020). More than half of the residents live in the Pärnu town and others in the smaller establishments such as villages (over 200 villages in the Pärnu County) (Local Authorities of Pärnu County, 2020).
In terms of area, the Pärnu county is one of the largest in Estonia with 5419 km 2 , of which islands and islets cover 23 km 2 . Islands of Kihnu (17.1 km 2 ) and Manija (2.1 km 2 ) are inhabited. The length of the county coastline is 635 km. The strategy of the Pärnu county highlights as the strengths of the region a strong regional centre (Pärnu town), and reputable resort and recreation industry. The weaknesses of the region include population decline and ageing, and the inability of Pärnu town to sufficiently fulfil the role of the regional development centre (Development Strategy of Pärnumaa 2035+, 2018. This region was selected for the study because tourism in this region has been for a long time concentrated in Pärnu town. Rural tourism started to foster in the region in the early 2000s when Estonia became a member of the European Union. The region hosts different rural and urban tourism networks and strategic aim of local tourism policy is to foster networking between them (Development Strategy of Pärnumaa 2030+, 2014). ANT helps to explain the motivations of actors to join the network (Albrecht, 2013), and suggests that without performance network does not succeed (Ren, 2010). Employing both, ANT and SNA enables to investigate the network formation and visualise network structure and knowledge circulation (Wickramasinghe & Bali, 2009).
Density concept of SNT explains how tightly network members are connected. In a dense network, actors are tightly connected while in an isolated network they share few interconnections. In the dense network, actors share certain values, membership or common interests (Reagans & McEvily, 2003) that enable them to formulate, express and execute collective will (Chung & Crawford, 2016). Entrepreneurs who belong to a regional tourism network but are located far from its administrative centre can be left out from social and communication circles. Interaction between entrepreneurs can reduce isolation and increase network density by providing them with substantial knowledge and increasing their involvement in local tourism (Hatipoglu et al., 2016). Isolation is linked with different dimensions of distance (cognitive, communicative, organizational, functional, social, cultural and geographical) which all influence collaborative networking (Czernek-Marszałek, 2019). The second important concept of SNT is centrality where the closeness of actors influences their ability to reach and communicate with others in the network. Freeman (1978) points to three forms of centrality: (1) degree centrality, which is measured in various connections that lead to a certain actor and shows how active this actor is in the field of communication; (2) closeness centrality indicates how close an actor is to others in the network. It also explains actor's ability to reach other actors in the network and shows the efficiency of the communication with other actors; (3) betweenness centrality shows the shortest path between the actor and other actors and indicates the actor's ability to control the communication in the network. Entrepreneurs with high betweenness centrality can enjoy more benefits in the network because they are brokering information to different groups that are isolated from each other (Burt, 2001). A central position in the network signals the degree of power obtained, as well as the capacity to access information and different network members (Borgatti et al., 2018;Wasserman & Faust, 1994).
Different ties: similarities (membership, attitude or location); social relations (friendship or acquaintance); interactions (trade) and flows (recourses or information) connect actors in a network (Borgatti et al., 2009). "Strength of a weak tie" is a SNT concept that bases on the idea that in dense networks the information does not stay novel for long because everybody knows what others are doing. It means that for new ideas the information must come from the weak ties (Granovetter, 1973). However, strong ties between the actors are also important in the network because this help to solve complex problems (de Montjoye et al., 2014). Therefore, both strong and weak ties are valuable for tourism network development (Houghton et al., 2009).
Some of the recent studies combine ST and SNA to research density of the tourism networks (Del Chiappa & Presenza, 2013;Nogueira & Pinho, 2015). Another set of studies have focussed on the network structure: a leadership network (Hristov et al., 2018), a network which contains tourist attraction (Liu et al., 2017) and a marketing network (Nogueira & Pinho, 2015). Some studies focus on the formation of networks (Dedeke, 2017;Tribe, 2010).
Rural and urban environments differ and can be regarded as different spaces that are related to global and local networks, which configure particular places. According to ANT different spaces take their form inside the networks (Murdoch, 1998). Halfacree's (2007) model of rural spatiality has three overlapping elements, which influence and constitute each other: representations of the rural, rural localities, and lives of the rural. The first element shows how the rural exists in formal contexts, such as in different policies, planning documents, and industrial interests. The second element refers to rural localities which have certain characteristics (e.g., cultural and natural landscape) that relates to localities through spatial practices. The third element refers to people's reproduction of rural practices in everyday life.
Explaining urban space is a difficult task because the urban environment has more elements. However, a combination of synergistic and structural perspectives allows identifying social features associated with urban space which can explain through the diversity of social roles, relationships, institutional arrangements and social networks (Aguirre, 2007).

Challenges in the tourism network development
Development of formal regional network raises the challenge of how to involve those informal and formal networks, which already exist in the tourism region, as well as entrepreneurs who do not belong to existing networks. This poses one of the key questions: should the network development incorporate a narrow or wide range of interests, and to which extent there is a common interest between entrepreneurs (Philipson et al., 2006)? According to Brandão et al. (2019), for tourism innovation, networks should be as diverse as possible. A tourism network can be self-organised, thus operating with minimal interference from the organising entity, or the opposite, where an organisational body, which implements regional tourism policy coordinates every aspect of the network with minimum interaction between the stakeholders (Provan & Kenis, 2008). However, the case by Phillipson et al. (2006) showed that when the foundation of the network is highly informal, external intervention could destabilise the network. Therefore, the formalisation of collaboration will result in a different network compared to the informal one.
Regular social interactions between the entrepreneurs in a network help to facilitate the collaborative process. Regular interactions help to develop and sustain trust, as well as social, intellectual, and political capital. A case from Poland shows that considering the potential benefits for all involved collaboration participants helps to build trust (Czernek & Czakon, 2016). Trust building in a network happens through information and knowledge sharing and commitment (Presenza & Cipollina, 2010). However, when there are different individuals and groups involved in the cross-regional network, overcoming the aforementioned challenges can be a difficult task.
Another challenge is the facilitation of networking between diverse regional tourism entrepreneurs. While regular interactions between the entrepreneurs increase social capital and trust, Beritelli (2011) points out that regular meetings of regional tourism entrepreneurs not always initiate beneficial collective action. However, sometimes, ad hoc meetings that focus on a certain problem can start a collaborative relationship between the members of a network (Parker, 2000). The facilitation of tourism collaboration networking requires a neutral arbiter. The arbiter must execute collective aims and ambitions (Fyall et al., 2012). Local DMO generally plays the role of neutral arbiter in facilitating regional tourism collaboration (Caffyn, 2000). However, trust towards the DMO may vary across the entrepreneurs, setting limits on extending networking. Different types of networks need different governance (Beaumont & Dredge, 2010;Provan & Kenis, 2008;Van der Zee et al., 2017). Large networks that lack consensus about their scope and exhibit low levels of trust typically need high management efforts to be successful. Lead organisation-governed networks typically have a top-down management approach and have a centralised structure. In contrast, participant-governed networks are managed by their members, are based on informal relationships, and have a decentralised structure (Provan & Kenis, 2008).
The evolvement of collaboration can lead networks into stagnation (Caffyn, 2000;Pilving et al., 2019). To avoid this, new knowledge must constantly reach the network (Brandão et al., 2019). Social capital in a community must link to the networks outside of one community not only derive from the resources that exist inside of that community (Woolcock, 1998). This can be difficult to achieve in the context of diverse regional tourism networks.
Various obstacles can hinder networking: disagreements, slow decision making, different needs and vision, insufficient dialogue and leadership. Jesus and Franco (2016) showed that networking helps to fulfil objectives which are impossible to achieve in isolation. It is possible to create synergies when collaboration benefits are clear to all entrepreneurs and there are enough interactions between the actors involved. Martínez-Pérez et al. (2019) add that diverse networks must include the widest possible range of relationships because a different type of knowledge will foster local tourism development and increase the sustainability of tourist destination.

Rural and urban tourism
Understanding the behaviour of the stakeholders in tourism networks implies understanding their context (Granovetter, 1985). Suppliers, travel intermediaries, and customers are tied together in the tourism distribution network through which tourism is developed (Pearce, 2009). A tourist destination is a complex dynamic system, where one large destination can comprise several smaller destinations (Framke, 2001). Therefore, rural and urban tourism networks can comprise several destinations in a tourist region. In the rural-urban tourism network, entrepreneurs must share a high level of interdependency. A lack of interdependency decreases collective strategic cognition, and individual actors end up focusing on their primary activities (Fyall et al., 2012). In this context, interdependencies are beneficial for the partners to combine resources and achieve competitive advantages (Teng & Das, 2008). Hence, the establishment of a regional rural-urban tourism network presumes that common ground and interests are found between different destinations and entrepreneurs. Jesus and Franco (2016, p.167) highlight that "the distinction between rural and urban enterprises is arbitrary". It is not easy to distinguish between rural and urban tourism entrepreneurs except for the environment (rural and urban space) where they operate. Urban tourism is usually concentrated in cities where tourism is integrated into other urban activities (Ashworth & Page, 2011;Edwards et al., 2008;Rasoolimanesh et al., 2017). In contrast, rural tourism includes all activities that visitors do in rural areas, which relate to the culture, hospitality and traditions in rural communities. It takes place in rural areas, where the entrepreneurs are members of small local communities (Dimitrovski et al., 2012;Jaafar et al., 2015).
Rural entrepreneurs are in a disadvantaged position compared to urban counterparts where different resources are more easily accessible (Gavrilă-Paven et al., 2015). Another aspect that influences network development in a rural area is a lack of critical mass of entrepreneurs. This is especially hindering factor when there are several competing networks in a community which all need resources like people and their time (Phillipson et al., 2006). A study from Portugal showed that rural entrepreneurs do not have a definite plan regarding to whom they will collaborate. Urban entrepreneurs do better analyses on this matter and are more willing to establish collaborative relationships with a large variety of other actors. Rural entrepreneurs collaborate with other rural entrepreneurs and in peripheral regions, they can be disconnected from wider regional networks. For urban actors, different contacts are important, and they are more willing to adopt innovations while rural entrepreneurs are less advanced on these topics (Jesus & Franco, 2016).

Rural and urban areas in Estonia from the collaboration perspective
Rural area in Estonia consists of places outside urban centres and their urban hinterlands and is characterized by sparse or very sparse population. Urban area in Estonia is a predominantly densely populated area consisting of a central town and/or suburban areas functionally linked to each other. Larger urban areas consist of municipalities with a population of more than 50,000 residents (Ministry of the Interior, 2014).
Rural and urban environments in Estonia have been shaped by recent communist past and transformation period. During the first independence period , in Estonia, cooperatives and associations were the connecting structures, which helped to raise the feeling of belonging at the community and personal level and foster collaborative relationships. During the collective farming regime, because of the scarcity of everyday supplies, an informal network of consumer acquaintances developed alongside the collective farms. When the Soviet era ended this informal network disappeared, which created a gap in collaborative relationships. The transition period after restoring independence led to several negative changes in the rural area. The deterioration of the reputation of Estonian rural life and socio-economic conditions during the transition period led to a decrease in the interdependence of rural people. This development is characterised by hard-to-accept social inequality between rural residents. Different events in the village (parties and cultural events) reinforced, not exceeded, the experience of the difference between the people. This problem could not be solved at the local level and external assistance had to be called for help. Help was received through several development programs, which aimed to activate residents and build partnerships in rural communities (Annist, 2011).
Nowadays, rural areas are characterized by a shortage of attractive jobs, limited services and leisure opportunities and weak links between the regional centre and the surrounding area. Sparsely populated and more peripheral rural areas must inevitably be able to adapt to changed circumstances. As of 2010, there were 48 rural municipalities in Estonia with about 50,000 inhabitants in total, whose population has decreased by at least half during the last 50 years or by at least one per cent per year during the last decade. Activities to implement and support regional development at the regional level have been hampered by the limited capacity of county governments, local government associations and regional umbrella organizations largely due to the weak status of regional development plans and in implementing national policies. The lack of collaboration between municipalities, the private sector and non-governmental organizations in regional development and the provision of public services has hindered regional development in a situation where people's daily activities are not limited with the borders of one municipality. Smaller municipalities alone cannot carry out large-scale development and the weak status of county development plans in directing state investments has also hindered the collaboration of local governments (Ministry of the Interior, 2014). Also, rural life in Estonia is affected by the fact that younger and older people have different opinions about the effects of past complex events (Martínez, 2018).
During the transition period in towns, events oriented to build nationality were replaced by wider cultural consumption (Annist, 2011). Nowadays, as a result of internal migration, a general regional pattern of the population is primarily affected by the continued concentration of the population in larger urban areas -especially in Tallinn, Tartu and Pärnu, where there are the best opportunities for work, study and another self-fulfilment. The growth of larger urban areas in Estonia has primarily taken place in the form of suburbanisation and population growth in the hinterland of towns. In the vicinity of larger towns, the age groups of younger adults and children are generally over-represented compared to the rest of Estonia, and older people are under-represented. Based on the current development trends, the concentration of the population in larger cities and their surrounding municipalities will continue in the coming years. Similarly, with the population, more active economic activity is concentrated in larger urban areas, where there is a greater critical mass of people, institutions and competencies for business development. Strengthening the physical and functional links between urban centres and their rural hinterland is also becoming increasingly important in the context of both EU cohesion policy and many national regional policies (Ministry of the Interior, 2014).

Tourism in the Pärnu region
The modern tourism era in the Pärnu region ( Figure 1) began after Estonia restored independence in 1991. Estonia opened up to Western markets, which created new opportunities and increased the role of the hospitality sector (Unwin, 1996). However, tourism vocation traces back to the 19 th century, with the health resort development in Pärnu town. In Pärnu town, the building of a sanatorium network began after WW II and several summerhouse establishments were built in the countryside, which made the region popular among visitors during the communist period. Today, the region is a popular location for summer vacation (Kask, 2008).
The role of rural tourism began to grow in the early 2000s when Estonia became a part of the European Union (EU), and additional funding became available for non-agricultural rural development. Attaining the EU membership had a substantial effect on tourism in Estonia due to new investments and the opening of the borders (Jarvis & Kallas, 2008). However, in rural areas, tourism has not developed as fast as in the town. With 778,000 accommodated visitors each year (2018) (Statistics Estonia, 2020) 85% of all visitors in the region are accommodated in Pärnu town (Statistics Estonia, 2018). The accommodations in Pärnu county have more than 8,000 beds (Development Strategy of Pärnumaa 2035+, 2018).
Pärnu region has adopted a strategy to foster tourism development and collaboration. Three main rural destinations -Lotte Village theme park for children, Soomaa national park, and Kihnu island, are considered as highlights in the region's rural areas (Figure 1). The regional strategy supports the development of rural destinations focused on family and conference tourists, and aims at the comprehensive development of tourism services, combining rural and urban tourism attractions Several organisations are related to tourism development in the Pärnu region. Visit Pärnu is the main public sector organisation (DMO) responsible for tourism development and collaborative networking in the region. Also, the region's LEADER action groups support tourism and companies in rural areas. LEADER groups are non-profit realities that mainly comprise small rural enterprises, local municipalities, and village associations. The Estonian Rural Tourism Organisation is a non-profit umbrella organisation of Estonian rural tourism actors in the whole country. Enterprise Estonia is the body responsible for promoting the Pärnu region globally. Many large spa hotels from Pärnu town belong to the Estonian Spa Association and Estonian Hotel and Restaurant Association, non-profit organisations that represent the interests of their members domestically and abroad.

Data
Qualitative research design (Flick, 2014) was used in this study because it enables to explore the research object (tourism collaboration networking) in its natural context (rural-urban and post-soviet transition settings). Qualitative research postulates that individuals (e.g., tourism entrepreneurs), their experiences and behaviours are "unique, context-dependent and largely non-generalizable" (Yilmaz, 2013: 317). Therefore, as part of the qualitative design, the purposive sampling strategy was used in this study. The principles of purposeful sampling (Flick, 2014) are extensively used in qualitative research because this enables to effectively select participants in the studies which are rich in information (Patton, 2015). The strategy involves participants who are experienced in and have a high level of knowledge about the phenomenon under study (Cresswell & Plano Clark, 2011).
The data of this study were collected through two sets of semi-structured interviews: 10 interviews with tourism stakeholders from Pärnu town and 27 interviews with tourism stakeholders from rural areas of the Pärnu region (Table 1). Participants were chosen via personal contacts and with them, the snowball technique was used for including more study participants. This allowed minimising the negative effect of the purposive sampling that at the beginning of the study is difficult to determine how big the sample size should be (Palinkas et al., 2015). In this study the sample size emerged during the research process when it was clear how many different tourism collaboration networks in the study region are and the participant selection ended when the information from the interviews saturated itself, meaning that theinterviewees no longer provided any new information. This kind of sampling approach helped to attain information about the regional networks where one entrepreneur named different enterprises, stakeholders and networks to whom he/she is connected which helped to map different networks and ties between the entrepreneurs. The following principles and justifications to construct the purposive sample were used: network representatives and entrepreneurs from rural and urban tourism areas, different fields of tourism operations, and several municipalities and tourism organisations and networks across the Pärnu region, along with active/passive and new/old tourism entrepreneurs and network representatives were included. This was guided by the principle to identify similarities and differences (Palinkas et al., 2015) in the regional collaboration networks. This study focuses on the entrepreneurs, however, because of the complexity of networking phenomena, representatives from other sectors (public and non-profit) were also interviewed (Table 1). Another important aspect here is that in Estonia (especially in the rural areas) many non-profits or public sector organisations offer tourism services as well. For example, the Pärnu Guiding Association is a non-profit organisation or a manor in the rural Pärnu which belongs to the local government but offers accommodation and catering for visitors or village associations which offer adventure guiding. On one hand, the aim was to select interviewees who are or have been actively involved in the tourism networks of the study region. On the other hand, to get a better overview of regional tourism networks, those entrepreneurs who have been more passive in networking were also included in the sample.
Usually, social network research focuses either on formalism or relationalism. Formalism means that the focus is on an objective view of networks' structure and relationalism deals with the meaning and experience of relationships and the latter examines the network from an insider's perspective (Erikson, 2013) which is also the case with this study.
More participants from the rural Pärnu region were chosen because the rural area is larger than the urban area, and the location of the entrepreneurs and distance from each other and Pärnu town can be essential for understanding the collaborative ties between rural and urban entrepreneurs.
The interviewees from the non-profit and public sector comprise several tourism organisations, municipalities, and existing tourism networks in the region. The sample includes both large and small enterprises from rural and urban areas. Some interviewees combine (a mix of) several tourism-related services on the daily basis (Table 1). Many participants (especially from the rural area) had multiple occupations. Because of the high seasonality in Estonian tourism entrepreneurs also work beside their business and have parallel jobs in local government, destination management organisation, in some local association or LEADER group or they have worked before in different tourism settings and fields. For example, the representative of a large spa has previously worked in the rural tourism company and the representative of the theme park in a large spa hotel in the Pärnu town. This ensured their high awareness and knowledge in the matter of regional tourism networking. Interviews were also conducted with entrepreneurs who only run their business and are regular members of some collaboration network in the region or do not belong to any network for getting a more detailed overview of the phenomena. 10 (accommodation and food, handicraft and museums) Non-profit and public sector 1 (representative of local DMO, guide association)

(village association, representatives of the local governments and representatives of the LEADER group)
The interview questions covered the following themes: the local tourism environment, the role of rural and urban tourism destination management, networking, decision making, the roles of DMOs and different actors in regional tourism, collaborative relationships and ties between rural and urban enterprises, as well as entrepreneurial benefits, motives, and problems related to collaboration. During the interviews, participants were asked to name all networks to which they belong to and other entrepreneurs to whom they collaborate and explain the ties and relationships during the networking which allowed to map different networks (Figure 3). The interviews lasted between 45 minutes and 2 hours, and all interview sessions were carried out between 2017 and 2018. All collected data were transcribed and coded using the MGT analysing principles.

The multi-grounded analysis
The research design of this study relies on the MGT, existing theories (ST, ANT, SNA, and SNT). MGT (Figure 2) enables the formulation (multi-grounding) of theoretical statements from empirical data and the development of an emerging theory by combining induction and deduction and build a research design that overcomes the limitations of existing theories. This approach allows thorough use of empirical data without limiting the study with a narrow theoretical focus, which may prevent the emergence of new knowledge. The MGT also enables to overcome the limitations of grounded theory (mainly introvert theorising and over-generalisation) (Cronholm & Goldkuhl, 2010).
The analysis begins with collecting preliminary theoretical arguments about the phenomena of interest. The concepts that emerge in different phases of multi-grounding are constantly assessed in the research process. In the final research phase, the generated theory is connected with the different origins of the achieved knowledge to prove its validity (Cronholm & Goldkuhl, 2010). During the research process, the research design can constantly change because of continuous theoretical matching and empirical validation during the research process (Lind & Goldkuhl, 2006).

(authors' figure)
The generation of theoretical statements during the MGT process consists of four-steps (Figure 2) (Cronholm & Goldkuhl, 2010). After the transcription, data were inductively coded closely to the text, without any pre-conceptions to prevent the loss of emerging concepts. Then the primary categorisation of codes was created, without any predetermined theoretical categorisations (Step 1). During this phase, the entity of different networks, ties, relationships, connections and density between networks of rural and urban entrepreneurs started to emerge. In this phase all codes were collected from different interviews and categorised to different topics. The first step allowed to map different stakeholders and networks with the use of insider view and SNA.
In the conceptual refinement phase (Step 2), the codes were assessed for categorisation. Conceptual refinement creates a 'comprehensive definition of categories' (Cronholm & Goldkuhl, 2010, p. 195). The MGT process assumes that the concepts that emerge in different phases of MGT are constantly assessed. In this phase, the codes, categories, and research notes were taken and collected during the interviews and different research stages were compared with each other. Here, also all empirical concepts that emerged from the first phase were critically challenged, examined and assessed before the next level of categorisation. During this phase, codes were regrouped, renamed, aggregated and deleted.
In the pattern coding phase (Step 3), the categories and codes that emerged from previous phases were compared with the concepts of ST, ANT, and SNT, mainly with those that explain density, salience, centrality and strength of different ties and relationships, level of power and information and knowledge flow. In this phase, a new set of categories was created. In this phase inductive approach met with the deductive approach.
Theory condensation is the last phase (Step 4) of the MGT process. This phase was preceded by different explicit grounding processes, which included theoretical matching, explicit empirical validation, and evaluation of theoretical cohesion. In this phase, the evolving theory was grounded, the empirical validity of the emerging theoretical statements was tested and evolving theoretical statements were formulated. This was assured with constant assessing the emerging theory by all authors.

Results and discussion
Urban networking space Data gathered from the interviews enabled to identify and map (Figure3) existing networks in urban and rural space. For the characterisation of collaborative ties, it is useful to consider the urban space in two levels and explore collaborative ties between; 1) key urban enterprises; 2) small urban enterprises, and 3) key urban and small urban enterprises (Figure 3).
One key principle of ANT highlights that places with similar elements and relations are close to each other and the proximity of different networks is related to space and time (Murdoch, 1998). It was found that in the urban tourism space, the ties between key urban tourism enterprises represent the first central level of collaboration. Interviewees highlight that a significant share of the visitors arrives in the region through channels controlled by key enterprises who have established relationships with public sector and shape tourism in the town. The roots of this situation date back to the beginning of the 1990s when sanatoriums in Pärnu town were privatised. The acquisition of existing infrastructures and extensive marketing generated significant interest in the Finnish market, which is the main visitor segment in the area. This helped to turn former sanatoriums into spas and give them a dominant position in the area. Key enterprises are open to collaborative relationships with others. A representative of a large spa explains:

"We have our company newspaper for promotion. In every edition, we have one article about rural tourism in the Pärnu region and if our clients ask about where to go in the countryside, we always recommend some businesses. I think that we have taken the first step and now the ball is in their hands".
Municipality of Pärnu town is actively involved in shaping and implementing local tourism strategy. In rural areas, there are different municipalities which tourism development related activities differ to a great extent. However, communities in the tourist region distinguish themselves less by formal rules and more by their key enterprises (Beritelli, 2011). This study's results show that dominant enterprises are involved in different working groups responsible for tourism development (e.g., infrastructure, strategic aims, and collective marketing), and thereby shape the local tourism policy. Key urban enterprises are salient, and DMO involves them to roundtables to discuss and decide on the development of tourism in Pärnu town and region. Also, they are involved in different professional networks (e.g., Estonian Spa Association, Estonian Hotel and Restaurant Association) which have helped them to establish a social network between them and have a strong influence on shaping the local tourism policy. However, the local tourism policymaking does not have a clear formal structure and is activated when needed. A spa representative illustrates co-working as follows: "Usually, we gather around the table when we have to make some big decisions in the region like the expansion of the local airport".
Beside the key enterprises, there are many smaller ones in the Pärnu town. Relationships between those enterprises can be characterised by high competition between them and they have not formed a collaboration network with each other. Usually, these enterprises target the spa visitors meaning that they don't have to look for the clients themselves and depend on the clients that the spas bring to the town. A representative of an adventure park explains the situation: "We mostly want to offer our service to the spa clients and holidaymakers on the beach. Doing this, we compete with the other adventure park". Jesus and Franco (2016) show that the relationships between urban enterprises are highly planned and organised. This study adds to this that large tourism enterprises often need the services provided by smaller ones for motivating tourists to extend the length of stay. A spa representative remark that: "Recently, we contacted one entrepreneur who has a minizoo. We want to collaborate with her and offer minizoo tours to our family clients who spend more time in our hotel. We don't offer services like those ourselves and this kind of collaborations help to keep the visitors` longer with us".
In Pärnu town, where Finnish tourists represent the primary visitor segment, local spas depend on Finnish resellers to fill their rooms. Small urban tourism enterprises also target Finnish tourists, which creates a dependency relationship (Figure 3) between the key and small tourism enterprises. Bramwell and Lane (2000) highlight that entrepreneurs must remain independent in collaboration. However, in the case of Pärnu town, collaborative dependency can be described as mutualism with some weakness where entrepreneurs' independence is limited mainly because small enterprises do not have control over their clients. The relationships between large and small urban enterprises are not dyadic. The more powerful partner can always withdraw from the collaboration (Ford et al., 2012) especially when collaborative relationships take place informally. For example, when a client asks in the spa reception where to rent a bike from, the received suggestion will depend on the relationship between large and small enterprises.

Figure 3. Collaborative ties between different regional networks (authors' figure)
In this context, key urban enterprises are salient, but due to the dependency, their relationships with small ones do not create a tourism network. Ties are created between them, but they remain weak. Besides mutual gain, key and small enterprises do not share knowledge or information and have minimal social relations. In this context, key enterprises enjoy the benefits from high betweenness centrality because they possess and share information with others at their own will. Jesus and Franco (2016) point out that weak ties between actors are highly formalised and based on short-term relationships, while strong ties are based on close relationships and repeated transactions. The results of this study add that weak informal ties in this context, even in the presence of mutual benefits and strategic cognition, are based more on the business transaction than a collaboration network between two groups of different enterprises. According to ANT, besides social relations networks also include material objects (e.g., tourism infrastructure) which creates a certain space which refers to a type of network "where the links between actors and intermediaries are provisional and divergent, where norms are hard to establish and standards are frequently compromised" (Murdoch, 1998, p. 362). From ANT perspective, the situation between two groups of entrepreneurs indicates that the position of power is manifested by the key enterprises through those dependency relationships (Figure 3). Several representatives of small enterprises highlight the need to develop new services in the low season and try to step out from the dependency relationships. However, they face difficulties in finding clients because of the strong spa and beach holiday image. The adventure park representative explains: "Some time ago did a marketing campaign to raise awareness that there are more things in the Pärnu region than just beach and spas, but we did not receive that many customers. I think that the town and spas are doing a lot of work to keep this summer holiday image in the region and small enterprises cannot change that".
However, some events (summer festivals and the Pärnu Restaurant Week in April) occasionally connect large and small urban enterprises for a joint networking effort. Despite being annual, the one-off nature of these events implies that they do not have a substantial impact on creating a collaborative network, which is based on common values and trust not dependency between large and small urban enterprises.

Rural networking space
In rural space, there are several local collaborative networks of tourism entrepreneurs. Some of them are associated with LEADER local action groups, i.e., these networks have emerged based on the existing collaborative networks (Figure 3). Despite the organisational support (e.g., collective marketing and marketplace, the possibility to apply for EU funding, socialisation events and workshops aimed for business development) provided by those groups, regional rural tourism networks are mostly selforganised, and decisions are made by their members. However, ties between different rural collaborative networks are not well established (Figure 3). There are several reasons for this. There exist a couple of very dense rural tourism networks (Kihnu island, Soomaa national park) in the region where members are closely located to each other, and that base on trust and frequent socialisation and information sharing between the members. On the other hand, there are other rural networks where density is low, members have social and locational distance between them, and where closeness centrality is low. The ties between the entrepreneurs with high locational distance are characterized by trust, friendship, resource and information sharing but it is not enough for achieving higher density which is needed for the growth of the network and connecting different networks. A rural entrepreneur illustrates the current situation: "Many tourists visit Kihnu island or Soomaa national park; after that, they go back home without visiting other destinations in the area ".
Connecting different networks requires common aims which is difficult to achieve. Rural entrepreneur continues: "All destinations in the Pärnu region are different and they all have different aims and vision. For example, the Kihnu island is a UNESCO heritage site and they do not have any problem to find visitors. Question is, why do they have to collaborate with Soomaa? The answer could be that in the future the tourism should focus more on developing functional tourist routes that will also guide visitors to other destinations in the region. This allows us to keep visitors longer in the region, but this is difficult to achieve. But right now, this is more like a dream or long-term vision".
Collaborative networking between the rural entrepreneurs takes the form of combining services (e.g., active holiday, accommodation and catering), and socialising (e.g., workshops that focus on learning and community tourism development, study trips and festivals).
According to SNT, nodes and ties form a system where different actors affect each other or the whole system and the network (Fyall et al., 2012;Wellman, 1988). Freeman (1978) highlights that the closeness of a network participant influences his/her ability to reach and communicate with others in the network. Dependency on rural networks is high because of the limited resources of a single entrepreneur but compared to urban networks information and resource sharing is mutually beneficial. From a regional perspective, resource allocation is achieved through collaborative efforts for increasing competitiveness (Jesus & Franco, 2012). It is a complicated process involving various networks in large tourist regions. In the core-periphery context, the periphery is usually far (both geographically and socially) from the core, where the main markets are, and decisions are made. However, rural networks can also form a core (Kauppila et al., 2009) in a specific location, as some networks addressed by this study. This suggests that the urban network might not be the only central group of actors in the tourism region. Dense rural networks can also be local or regional cores because of the low dependency from other regional networks (Figure 3). Rural networks that stand out in the region do not depend on urban tourists, and their visitors are usually not interested in visiting surrounding sights. Therefore, it would be difficult to motivate small high-and low-density networks to join a large regional tourism network.

Regional networking space
The results confirm the complexity of rural-urban networking. A spa representative highlights that 'it is difficult to define collaboration in this context'. This suggests that in Pärnu region, the ties between the existing networks in urban and rural tourism space are not established. Therefore, as suggested by the ST, since the DMO does not involve and address the interests of the tourism entrepreneurs with different salience (Mitchell et al., 1997), regional tourism networking suffers.
The findings are in line with the ANT in that networks in the region are heterogeneous (Dedeke, 2017;Ren, 2010). There are different reasons why the region-wide network of tourism entrepreneurs has not formed. First, despite the common interests between the high-and low-density networks in the region, it is difficult to find clear aims on those interests on the regional level because the basis of regional rural and urban networks is different. Second, despite the position of power of key enterprises in the region, they have a low influence on dense rural networks. This correlates with the ANT in that without relationships between the actors, power is not a resource itself (Nguyen et al., 2019). High betweenness centrality in this context means that key enterprises broke information mostly to small urban enterprises. Dense rural networks do not need knowledge and information offered by the key enterprises because they can access and gather knowledge by themselves. However, tourism networks are constantly evolving and finally, they stagnate (Caffyn, 2000;Pilving et al., 2019). The results show that this influences especially highly dense networks. Diversity is the key to network innovation (Brandão et al., 2019), but it is difficult to create ties with low-density networks when they are socially far from dense networks.
Third, there is no principal agent in the region who can activate rural and urban entrepreneurs in setting the goals of the regional tourism strategy. Dense rural networks have their principal agents, and the same situation is in town. In the complex environment where urban and rural development has followed different paths for decades and where the regional tourism management of the area has focused on the town and LEADER groups have developed rural areas a common basis for regional tourism network development is not established. Interviewees highlight that current tourism organisational structure is not capable to fulfil a strategic aim of rural-urban common network and this requires new kind of institutional development through which a new principal agent can evolve. Murdoch (1998, p.363) highlights that "the most unified networks tend also to be the most formal". However, in the Pärnu region, tourism networks are informal and the reason behind this is the evolvement of post-communist turbulent environment where different reforms have hindered the development of principal agent who is capable to create ties between different regional networks through formal relationships. To illustrate this situation the spa representative points out that: "We have tried to contact different rural enterprises and networks, but usually we do not get many answers. So, we have taken imitative, but without the initiative from the rural side it is difficult to proceed with the regional networking".
However, from a rural side, there is a fear that more collaboration with urban entrepreneurs the urban interests will still dominate in the region. A rural entrepreneur points out that: "Urban entrepreneurs have enjoyed the benefits of Finnish visitors. In the rural areas we do not have had that and I think that it has been much more difficult for us. I think that more collaboration with them means that finally they will swallow us and secure their dominance over us".

Spa representative replies that:
"There is no basis for the fears of rural entrepreneurs. I think that they are having the inferiority complex of some kind".
The above-mentioned aspects show that there is a lot of misunderstanding between urban and rural entrepreneurs because of lack of socialisation and links between different regional networks which make finding a common ground a challenging task.
Some entrepreneurs in the region offer services targeted to urban visitors, but they also operate in the rural area and autonomously look for clients, which helps them combine benefits offered by rural and urban networks. SNT contends that network participants are more successful when they occupy a central position in the network (Freeman, 1978). This study shows that by being active in both the rural and urban market, an actor may participate in both networks, creating a positive outcome for the company but not necessarily creating a connection between rural and urban networks (Figure 3). The entrepreneurs that belong to two networks can access members and information (Wasserman & Faust, 1994) from both markets. The numbers of such enterprises are increasing because in recent times urban visitors are becoming more and more interested in rural areas.

Conclusions
By applying ST, ANT, SNA, SNT, MGT, and a qualitative approach, this study aimed to answer the following research questions: (1) what is the nature of collaborative ties between entrepreneurs and subnetworks in a tourist region; (2) how different ties between regional tourism networks foster and hinder the development of rural-urban tourism network? The main contribution of this study is new insights on how diverse networks of tourism entrepreneurs' function in urban and rural space, and by which ties the entrepreneurs are connected. The novel use of MGT in the network research has proven suitable for studying tourism networks allowing to overcome previous limitations in network research. Figure 4 summarises the aspects that explain the formation of ties between regional networks of tourism entrepreneurs. The regional tourism networking can be analysed from network-specific and regional aspects. The latter refers to urban and rural space, which may have differences in their: (1) entrepreneurial setting (e.g. size and competitiveness of enterprises, entrepreneurial skills); (2) existing networks and networking practices (e.g. networks lead by dominant enterprises, LEADER groups engaging different public and private stakeholders); (3) public sector involvement in tourism development (e.g. initiative of municipalities in formulating regional tourism strategies, and establishment of institutions to implement actions to achieve strategic aims), and finally; (4) visitors (e.g. spa visitors in the urban area, and nature tourists in rural areas).
Network-specific aspects are related to the specific aim of regional tourism network. Often, the aim of regional tourism networking is related to increasing the competitiveness of tourist region and enhancing the capacity of local tourism entrepreneur.

Figure 4. Aspects that influence the evolvement of regional tourism network (authors' figure)
Results show that it is important to find a common ground (Figure 4). This relates both to the capacity of the entrepreneurs, and the clarity of aims of collaboration. Also, the higher capacity of the entrepreneurs makes it easier to make agreements on regional collaboration.
There must be clarity about the aims of regional tourism networking through which diverse entrepreneurs can create mutual benefits (Figure 4). The results reveal two broad categories of benefits of collaborative networking. Improved competitiveness of a tourist region (joint marketing, ability to implement larger projects, a larger variety of services to visitors), and improved capacity of entrepreneurs (through bringing together dispersed entrepreneurs, and community development). Confusing aims hinder the formation of a joint rural-urban network. This suggests that more awareness about the potential and achievable benefits is needed together with specific and measurable goals of the collaboration.
The capacity of the entrepreneurs is positively associated with the competitiveness of tourist region. This suggests that the entrepreneurs in the study region acknowledge the relevance and potential of the strength of the weak ties that is a vital resource in solving the complex challenge of bringing the tourist region out from stagnation.
Results show that when a tourism entrepreneur is strongly connected with a certain community in a tourist region, he/she has stronger collaborative ties with other stakeholders in that community. However, strong collaborative ties inside a community do not necessarily create strong ties with other communities in the region. There are limited connections between rural and urban enterprises, but also between local networks of rural tourism enterprises. The mere physical distance between these enterprises and lack of social connections hinder networking. Missing ties between the entrepreneurs make information sharing in the (potential) network a mission impossible.
The results are in line with the suggestion of ANT that in the rural-urban setting the possible initiative of dominant enterprises can trigger the growth of collaborative ties between different entrepreneurs. At the same time, it appeared that rural enterprises are cautious and fear urban domination in regional tourism network. It also appeared that the differences between identities of sub-regional destinations and networks hinder regional tourism networking. This aspect has not gained much attention in the academic literature before.
Findings indicate that it is important to involve all entrepreneur groups to goal setting and take advantage of the strength of the weak ties. It is important to consider the contexts of urban and rural tourism space in regional tourism networking. Therefore, applying the SNT for exploring the nature of ties within a network is an important element of regional tourism network studies. In the case of Pärnu region, the urban and rural space are quite different in their entrepreneurial setting (Figure 4), networking between stakeholders, the public sector's involvement, and visitors.
Based on the results, the study has two recommendations for developing rural and urban networks. First, the DMO must involve entrepreneurs with different level of salience in tourist destination management. This helps to create ties between entrepreneurs and networks that operate in rural and urban tourism space. Second, the regional tourism networking needs to utilise the potential that lies in the strength of the weak ties which are often found in low-density networks. This is a valuable resource in bringing the tourist region out from stagnation.
The findings suggest that further research in this field is necessary. In the environment of this study rural and urban tourism entrepreneurs share mostly informal relationships, however, collaborative ties can be different if the relationships between the entrepreneurs are formalised. The study focussed on the relationships and ties between the entrepreneurs in the tourism networking context but focus on the public and non-profit sector in the rural-urban networking setting needs further research and discussion. Despite the extensive analysis of one tourist region, which has a unique cultural, historical, and geographical legacy, the results of this study may not be generalised to other environments. Future studies should address different social, economic, and cultural contexts to develop a more general theoretical understanding of rural-urban collaboration networks. This research focussed on different networks in certain timeframe. However, networks are constantly evolving which means that a longitudinal approach may open up new perspectives in understanding rural and urban networking phenomenon and the evolvement of ties between the stakeholders. This study indicates that the use of MGT and qualitative methods are suitable for studying rural and urban tourism networks at the regional level. However, other methodological approaches may add new insights and aspects which concern rural-urban networking on a regional level.