The Social Impact of Events in Social Media Conversation Alessandro Inversini, Rogan Sage, Nigel Williams, and Dimitrios Buhalis Abstract Events often support social causes. In addition to altruistic reasons, this association may bring also commercial benefits. However, to date, it is not entirely clear the extent to which event stakeholders engage in socially related discussions, making it difficult to evaluate the degree to which events act as a platform for social awareness. Using archived online narratives from Twitter.com, this study seeks to examine the extent to which event stakeholders engage in discussions of social causes. Results show that there is a scarce interest in socially motivated discussion by events attendees on social media. Keywords Events technology • Social impact • Social media 1 Introduction There is a vast array of studies measuring the impact of events. Many of these publications focus largely on the economic impacts (e.g. Bagiran and Kurgun 2013). Events can also generate intangible, social impacts among their communities (Kania 2013). Schulenkorf and Edwards (2012), for example, noted the capacity of events to facilitate the crossing of social boundaries like racism and prejudice through positive social interactions. However, there is much discrepancy over how to accurately measure intangible impacts (Fredline et al. 2003). The growth of social media may provide an avenue to resolve this challenge. These sites currently host real time discussions on a number of issues, including events. They can therefore be analyzed to understand the extent to which socially motivated discussions are a part of the discourse on a given event to understand it is possible social impact. This research uses twitter.com as medium to investigate socially motivated discussions. It examines its Communities of Interest (Williams et al. 2014) to understand the level of engagement of attendees on socially motivated discussions. A. Inversini (*) • R. Sage • N. Williams • D. Buhalis School of Tourism, Bournemouth University, Poole, UK e-mail: ainversini@bournemouth.ac.uk; rsage@bournemouth.ac.uk; nwilliams@bournemouth.ac.uk; dbuhalis@bournemouth.ac.uk © Springer International Publishing Switzerland 2015 I. Tussyadiah, A. Inversini (eds.), Information and Communication Technologies in Tourism 2015, DOI 10.1007/978-3-319-14343-9_21 283 284 A. Inversini et al. Data was collected from the twitter conversations around Glastonbury music festival (glastonburyfestivals.co.uk). This event was suitable as it is not only one of the largest music festivals in United Kingdom, but maintains social values and is often associated with social causes (glastonburyfestivals.co.uk/worthy-causes/). Twitter conversations about the festival were archived and analyzed using social network analysis and content analysis in order to understand the nature and extent of social cause related conversations. 2 Literature Review 2.1 Festivals and Events The term ‘festival’ has been used traditionally to signify a time of celebration, relaxation and recuperation which often followed a period of hard physical labour, typically the sowing and harvesting of crops. In the field of tourism and the related disciplines of event management and event tourism, festivals are described as “. . . public, themed celebrations. . .” (Getz 2005, p. 21). Festivals are distinguished from other types of special events by their purpose, which is the celebration or expression of the historical, social or cultural aspects of a particular host community (Getz 2008). While this is still true for many festivals, an increasing number of festivals incorporate economic and promotional objectives to justify the costs of organizing for the taxpayers (Gold and Gold 2008). Kania (2013) found that events generate social capital among their communities. Sharpley and Stone 2011) specified three ways in which events generate positive social capital. Schulenkorf and Edwards (2012) noted the capacity of events to facilitate the crossing of social boundaries like racism and prejudice through positive social interactions. Their study focused on interactions between children surrounding sports events and suggested that shared experiences united the event participants. Most importantly, they noted the likelihood of influence on a wider range of people by expression of acquired socially responsible views to others through word of mouth. 2.2 Social Impacts of Hallmark Events Most of the studies looking at events’ impacts do focus on economic impacts (Bagiran and Kurgun 2013). Researchers have developed an increasing interest in the more intangible impacts of events and a number of studies have been conducted over the past decade including many focusing on social impacts (Fredline and Faulkner 2000; O’Brien 2007; Schulenkorf and Edwards 2012). Although the research is evolving, there remains much discrepancy over how to accurately measure these often-intangible impacts (Fredline et al. 2003). As a plethora of studies emerge in attempt to measure and quantify social impacts of events (Picard The Social Impact of Events in Social Media Conversation 285 and Robinson 2006), a strong definition of social impact is necessary to realise this objective. Social impact is rarely allocated a comprehensive definition in the literature with authors opting to simply list the effects under headings such as ‘community pride, ‘participation’, and ‘expanding community perspectives’ etc. (Swart and Bob 2005). However the following definition from (Latane 1981) is commonly accepted throughout the literature (e.g. Page and Connell 2011): . . .Any of the great variety of changes in psychological states and subjective feelings, motives and emotions, cognitions and beliefs, values and behaviour, that occur in an individual, human, or animal as a result of the real, implied, or imagined presence of action of other individuals—(Latane 1981) Lin (2012), who employed a very broad scope regarding the social impact of events, supported the comprehensive nature of Latane’s definition and its propensity for negative and/or positive orientation of an event’s social impact. Moving from Latane’s (1981) definition of social impact this research explores socially motivated discussions via the social media within the communities formed on Twitter.com. 2.3 Social Media Communities of Interest to Understand Social Impact In order to understand if social media conversation related to events can be vehicle of social capital and socially motivated discussion the analysis of the narratives created online communities of interest around the event (Williams et al. 2014) is here proposed. In order to understand its nature, top down deductive methods may not be sufficient. While inductive methods can provide a deep understanding of the topic, it is difficult to apply them to the volume of stakeholders involved in a major event. Actually, the disruptive rise of the internet present an incredible opportunity for tourism and events researchers as many of the discussions about travel, destinations and events now occur online in a form that can be archived and analysed (Neuhofer et al. 2013). Due to the number of individuals using these online platforms, it is possible to compare a number of perspectives on the issue (Zaglia 2013). Since the emergence of communities based on interest, information and affection, researchers have sought ways of classifying them. These network communities have been defined by the structured social relationships created by fans, customers or admirers (Muniz and O’Guinn 2001). Communities of Interest (COI) can be online or offline (Muniz and O’Guinn 2001), small (Bagozzi and Dholakia 2006) or large. Communities of interest are agglomerations of individuals with a shared interest in a domain or area. Members may also share distinct values, behaviours patterns of language and signals (Muniz and O’Guinn 2001). Further, beyond common beliefs, members may feel moral responsibility for supporting other members and integrating new members into the community. This is the core of an online community as these categorizations define the nature and extent of their activities, allowing them to identify members and non members (Bagozzi and Dholakia 2006). It defines and structures the community 286 A. Inversini et al. experience and allows members to assign meaning to their activities that they then communicate to others (Casal o et al. 2008). Several factors influence the nature of the interaction that members will have in these communities. The size of the group can positively influence the amount of content created or shared and hence the benefit that individuals will gain from membership. Group heterogeneity also positively influences the amount of contributions and benefits to members (Oliver et al. 1985). In addition to ubiquity, customer narratives on COI hosted on social media, are based on the perceived experience of individuals (Inversini et al. 2009) and may be seen as more authentic than media or in general promotional materials. By presenting a consumer influenced narrative (Inversini and Buhalis 2009) about the event, these discussions can generate “eWord of Mouth” (Hennig-Thurau et al. 2004), creating impressions that can influence consumers’ actions (Godes and Mayzlin 2004). 3 Methodology The main aim of this research is to develop an understanding of the socially motivated discussions of festivals on social media communities of interest. In order to understand this two main research objectives have been designed: • To understand the extent to which social media has a propensity to facilitate Socially Motivated Discussion • To establish a correlation between social media users centrality in the network and the propensity to facilitate Socially Motivated Discussion 3.1 Research Design The analysis was conducted on a body of secondary data collected from the microblogging site Twitter.com about the Glastonbury Music Festival (GMF), one of the major music festivals in UK. Every year GMF publishes on its website a series of social causes supported by the festival (please see: glastonburyfestivals. co.uk/worthy-causes/). The data was compiled of tweets from the time period of Data was collected for the time period: 00:00:00 am (GMT + 0) on 27/06/2013– 11:59:59 pm (GMT + 0) on 04/07/2013. Tweets were collected based on them containing at least one of the following words: ‘Glastonbury’, ‘Glasto’, ‘#Glastonbury’ or ‘#Glasto’. The Tweets were then collated into a spreadsheet for sorting, filtering, and analysis. The date boundaries were based on the festival schedule. The opening date represents the start of the performances at the festival and the closing date represents 1 week after this date and the 3 days after the final day of the festival. Each tweet was collected with the following information: Username, Text, Language, Location, Time Zone, Hashtags and User Mentions. The Social Impact of Events in Social Media Conversation 3.2 287 Data Filtering In Honey and Herring (2009, p. 7) study found that Tweets including the @ were “more likely to provide information for others and more likely to exhort others to do something”. Therefore tweets not including a user mention (signified by an ‘@’ sign) were removed because they do not constitute conversation or interaction and are just seen as ‘noise’ (Denzin 2008). Tweets that were sent from a GMT + 0 time zone were separated from those within a GMT + 0 time zone to differentiate between tweeters who could have been directly influenced by the event in other ways; such as direct contacts with socially activist tents or expositions at the festival; and those who could not. With the separated dataset, it was possible to compare the international impact across the online platform with the impact in the local time zone across the same platform. Time zone variation was used instead of ‘location’ because time zone is allocated automatically and location is given by the user and therefore may not be reliable. This was confirmed in the data where users often left the location blank or put in alternatives such as ‘earth’ and ‘here’ which reduces to the reliability of all ‘location’ information. Given the window of time that the dataset represents, it is unlikely that many of the tweets from outside the GMT + 0 time zone would be from attendees, however it is possible that some may have left the time zone before tweeting about the event in the proceeding days up until the 4th of July. The final data set consisted of a total of 106,650 individual tweets, all of which contained either the words ‘#Glastonbury’, ‘#Glasto’, ‘Glastonbury’, or ‘Glasto’. 3.3 Social Network Analysis Process The data was then put through a SNA tool (NodeXL—V.1.0.1.245); a program which sorts the data into nodes and edges in order to group tweets by topic of discussion and establish the individuals who are most visible within the groups. A participation/relational method was used to define key actors within the network (Knoke and Kuklinski 1991). Those with a higher ‘betweenness centrality’ (BCen) score were those who bridged the gap between two or more other nodes most frequently; and those with a higher ‘closeness centrality’ (CCen) score were those who were more closely linked to all members and hence are in a good position to monitor what goes on within the network (Scott et al. 2008). Both figures indicate positions of higher influence in those network members (nodes) with higher scores. Centrality measures enable researchers to understand the role of nodes (in this research, twitter users) in networks (Borgatti 2005). Degree centrality is defined as the number of nodes that are directly connected to a given node. For this research, it would be the number of twitter users that have retweeted, mentioned and replied to a given user during the period of study. A high degree centrality indicates that a user is active in the network as they have the most ties to other people in the network. 288 A. Inversini et al. Closeness centrality identifies the overall influence of a node in a network. It is defined as how close a node is to other nodes in a network. In the case of this research, closeness indicates the number of times a given node lies on the shortest path from one side of the network to another. The higher the closeness centrality, the more control the node has over communication between nodes on different sides of the network. 3.4 Content Analysis Computerised content analysis is an objective and systematic process allowing inferences to be made from attitudes, intentions and values of individuals (Morris 1994). The text was extracted from the dataset and put through the text analysis tool, Voyant (http://voyant-tools.org) for syntactic analysis to see how frequently each word was used. Topical Word Uses (TWUs) were identified through word frequency analysis and then contextually analysed to determine whether they constituted Socially Motivated Discussion (SMD). When running the content analysis there were a number of stop-words removed from the results including: ‘Glastonbury’, ‘Glasto’, ‘http’, ‘t.co’, ‘2013’, ‘@youtube’, and ‘RT’. When trigger words such as ‘feminism’ or ‘peace’ were identified, these tweets were then analysed semantically to reveal the context of the usage to define relativity to the research context (Zhang et al. 2013). For example: the word ‘peace’ could be used in a social context or in reference to the artists whose band are named ‘peace’. To remove insignificant date, TWUs with a count of 2 and above were recorded whilst single count word uses were overlooked. 4 Results The total number of tweets gathered was 106,650. Of those tweets there were a total of 58,406 edges, which connected users through direct interaction. The interactions between each data point were used to create groups where users were connected directly and indirectly between each other. The top 20 groups were identified in both the GMT and NGMT networks; collectively they accounted for 32.03 % (18,710) of the total edges. Topical word uses were identified within these groups through word frequency analysis and then contextually analysed to determine whether they constituted Socially Motivated Discussion (SMD). A total of 164 edges (interactions) made up the 19 SMDs identified through context analysis. It is here important to note that for this research, a socially motivated discussion is an online interaction in the form of a retweet, reply or mention of a tweet that is focused on a domestic or international social issue. The average edges per SMD were 8.632, whilst the maximum was only 38, which shows that the SMDs were generally not adopted by the networks to any significant The Social Impact of Events in Social Media Conversation 289 Table 1 Network comparison, GMT + 0 vs. non-GMT + 0 Tweets % of total Edges (interactions) % of total Vertices (users) % of total Groups % of total Edges in top 20 groups % of network edges SMDs in top 20 groups % of total SMDs Total SMD edges % of total edges % of top 20 groups Max edges in SMD Avg. edges per group GMT + 0 Non-GMT + 0 Total 33,025 31.0 % 17,475 29.9 % 21,002 30.2 % 5,170 31.2 % 5,714 32.7 % 9 47.4 % 81 0.464 % 1.418 % 28 4.062 73,625 69.0 % 40,931 70.1 % 48,480 69.8 % 11,390 68.8 % 12,996 31.8 % 10 52.6 % 83 0.203 % 0.639 % 38 3.594 106,650 – 58,406 – 69,482 – 16,560 – 18,710 32.03 % 19 – 164 – – – – degree. Given the wealth of data collected between the set dates, it is clear that the volume of socially motivated conversation generated on the social media site ‘Twitter.com’, regarding Glastonbury Music Festival, was minimal. This shows that the social conscience, which the organisation articulates through its company objectives, is not replicated in the public social media domain. A substantial amount of the total conversation (Table 1—69 %) was generated outside of the GMT + 0 time zone. This implied that there is a vast potential audience for GMF, and hence a strong potential outside of the host time zone for significant social impact connected to conversation associated with the event (Durst et al. 2013). The network analysis separated the GMT network into a total of 5,170 distinguishable groups based on 17,475 interactions. The average number of interactions (edges) per group was 3.59; when compared to the number of edges in Group G1 (713), Group G2 (674), and Group G3 (412), this indicates that the majority of GMT discussion surrounding GMF on the social media site Twitter.com was nucleated around a small number of larger groups, with many small groups connecting much smaller quantities of people. The NGMT network was separated into a total of 11,390 distinguishable groups based on 40,931 interactions. The average number of interactions (edges) per group was 3.38. The relationship between the total network edges and the total edges in Group 1 (1,868), Group 2 (1,475), and Group 3 (1,185), indicated that the discussions were nucleated as they were in the GMT network. In the top 20 GMT groups identified through content analysis, there were only 81 interactions contributing to nine (9) SMDs that were identified to be contextually 290 A. Inversini et al. Table 2 GMT + 0 socially motivated discussions (SMDs) and lead user centrality SMD Count Lead user BC CC Groups MP Resigns Ageism Volunteers Throwaway society #lovesyria #wowpetition Class priorities Glastonbury University Crime Total 28 18 11 10 4 3 3 2 2 81 NoiseyMusic HadleyFreeman EmilyEavis Redorbital Oxfamgb WOWpetition arbolioto crowsnestcrew GlastoWatch 378,579.4995 66,949.84464 326,887.6952 173,882 69,574 0 52,188 52,186 23,693,015.61 0.00002 0.000021 0.000022 0.000017 0.000023 1 0.000015 0.000014 0.000029 10,12 4,12,17 1,7 14,17 1 1 13 5 9 relevant to the research (see Table 2). Given the size of the network as a whole, the amount of social discussion generated in the GMT network was very low. In the top 20 NGMT groups identified through data analysis, there were only 83 interactions among ten (10) SMDs that were identified to be contextually relevant to the research; 38 of which could be attributed to one SMD (see Table 3). Given the size of the network as a whole, the amount of social discussion generated in the NGMT networks was very low. Only 0.639 % of interactions within the top 20 NGMT groups constituted SMD compared with 1.418 % of top 20 GMT group interactions. More than double the percentage of conversation was socially motivated in the GMT network compared to the NGMT network. This implies that the degree of social impact generated by GMF varies depending on location. However this could also be due to a lack of international transferability regarding some of the social topics discussed. When country-specific SMDs which are not considered transferable; in this case ‘MP resignation’ & ‘war on welfare petition’ (combined at 31 vertices); are removed from the GMT top 20 findings then the percentage of interactions constituting SMD is reduced to 0.875 % (compared with 0.639 %—NGMT). This decreases the implication that the degree of social impact generated by GMF varies significantly depending on location. 4.1 Betweenness Centrality and Socially Motivated Discussions The betweenness centrality rating of the lead user in each SMD was plotted against the edge frequency within the respective SMD. The graph exposed two anomalous results (‘#savethearctic’ & ‘Crime’), which had significantly larger BCen scores. Therefore these two data points were removed from the graph to give a clearer representation of the trends within the data (Rosenthal 2011—see Fig. 1). Figure 1 shows no discernable correlations between the x and y axis, which indicates that The Social Impact of Events in Social Media Conversation 291 Table 3 Non-GMT + 0 socially motivated discussions (SMDs) and lead user centrality SMD Count Lead user BCen CCen Groups #savethearctic Feminism Legal highs Girl power Social class in the music industry #stopthecull Charity event promotion Corporate glastonbury Ageism Horse & Pony Sanctuary Total 38 21 8 4 3 alt_j UK_Feminista ukhomeoffice AllanaSouthgate JohnRMulvey 4,810,121.821 161,140 456,059.2792 322,218 81,471.4481 0.00001 0.000007 0.000007 0.000007 0.000009 9 1,3 18 6 3 2 2 2 2 1 83 veggiesnottm coldplaying GuardianUS HadleyFreeman veggiesnottm 80,574 619,615.517 161,140 201,420 80,574 0.000008 0.000009 0.000007 0.000006 0.000008 4 9 16 19 4 800000 600000 400000 200000 0 0 5 10 Series1 15 20 25 30 Linear (Series1) Fig. 1 SMD edge frequency vs. betweenness centrality rating of lead user BCen rating of the lead user and the adoption of SMDs in the network have no significant impact upon each other. 4.2 Closeness Centrality and Socially Motivated Discussions The closeness centrality rating of the lead user in each SMD was plotted against the edge frequency within the respective SMD. The data included one anomalous result (‘#wowpetitiion’), which had a significantly higher CCen rating (1). The data point was removed from the graph to give a clearer representation of the trends within the data (Rosenthal 2011—see Fig. 2). Figure 2 shows no discernable correlation between CCen score of the lead user and the adoption of SMDs. This indicates that the CCen rating of the lead user and the adoption of SMDs in the network have no significant impact upon each other. 292 A. Inversini et al. 0.00004 0.00002 0 0 5 10 15 Series1 20 25 30 35 40 Linear (Series1) Fig. 2 SMD edge frequency vs. betweenness centrality rating of lead user 5 Discussion and Conclusions The results of this study clearly highlight the importance of studying community of interests on social media when it comes to socially motivated discussion. Social motivated discussion can be a string driver for event organizers to generate awareness towards a given topic and involve event attendees in meaningful discussions thus fostering the relevance of the social cause. Events such as the one studied which claim to support a series of social causes may like to increase the online and offline discussion about those in order to generate much more impact. The indication from the literature was that social media networks had strong potential as a platform for the advocacy of social issues, and potential therein as a catalyst for social change. However the adoption of SMDs throughout both networks was very limited which suggests that this is not the case. Two reasons for this lack of SMD adoption in the networks could be lack of advocacy to inspire discussions or simply a lack of willingness from the network to adopt these discussions. The findings for the two networks showed that there was a difference in the percentage adoption of SMDs; however this could be due to a range of factors including anomalous results, given the relatively small dataset. Additionally, the lack of significant differentiation between GMT and NGMT adoption of SMDs was expected given that geographical location is not a prerequisite of membership within the network, and nor does it affect the ways in which users can interact with the network. This indicates that the conversation surrounding events can transcend across time zones via social media irrespective of geography. Events organisers must therefore consider their social impacts as global rather than domestic. Additionally, the research expected to identify that SMDs garnered more support when users with higher centrality (BCen/CCen) were involved in the discussion, however the findings showed no correlation between the centrality of the lead user and the adoption of SMDs. It is important to note that the BCen and CCen graphs show a total of 17 and 18 data points respectively, and therefore do not give substantial evidence to make a conclusive inference. Further research into this phenomenon is needed. Future research may be able to examine this issue. In the case of lack of advocacy, a comparative approach examining the online narratives explicitly The Social Impact of Events in Social Media Conversation 293 cause driven events and music festivals that support causes may be able to illustrate the nature of advocacy in this domain. It may also provide rationales for the relative presence or absence of such advocacy. For what concerns lack of willingness, it may be necessary to examine the users online narratives outside of this event to understand their level of engagement with social causes any why they may or may not continue such engagement during an event. References Bagiran, D., & Kurgun, H. (2013). A research on social impacts of the Foça rock festival: The validity of the festival social impact attitude scale. Current Issues in Tourism, 16(1), 1–19. Bagozzi, R. P., & Dholakia, U. M. (2006). Antecedents and purchase consequences of customer participation in small group brand communities. International Journal of Research in Marketing, 23(1), 45–61. Borgatti, S. P. (2005). Centrality and network flow. Social Networks, 27(1), 55–71. Casaló, L. V., Flavián, C., & Guinaliu, M. (2008). Promoting consumer’s participation in virtual brand communities: A new paradigm in branding strategy. Journal of Marketing Communications, 14(1), 19–36. Denzin, N. K. (2008). Studies in symbolic interaction. Emerald: Bingley. Durst, C., Viol, J., & Wickramasinghe, N. (2013). Online social networks, social capital and health-related behaviors: A state-of-the-art analysis. Communications of the Association for Information Systems, 32, 5. Fredline, E., & Faulkner, B. (2000). Host community reactions: A cluster analysis. Annals of Tourism Research, 27(3), 763–784. Fredline, L., Jago, L., & Derry, M. (2003). The development of a generic scale to measure the social impacts of events. Event Management, 8(1), 23–37. Getz, D. (2005). Event management and event tourism. New York: Cognizant Communication Corporation. Getz, D. (2008). Event tourism: Definition, evolution, and research. Tourism Management, 29(3), 403–428. Godes, D., & Mayzlin, D. (2004). Using online conversations to study word-of-mouth communication. Marketing Science, 23(4), 545–560. Gold, J. R., & Gold, M. M. (2008). Olympic cities: Regeneration, city rebranding and changing urban agendas. Geography Compass, 2(1), 300–318. Hennig-Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004). Electronic word-ofmouth via consumer-opinion platforms: What motivates consumers to articulate themselves on the Internet? Journal of Interactive Marketing, 18(1), 38–52. Honey, C., & Herring, S. C. (2009). Beyond microblogging: Conversation and collaboration via Twitter. In 42nd Hawaii international conference on system sciences, 2009. HICSS’09 (pp. 1–10). Inversini, A., & Buhalis, D. (2009). Information convergence in the long tail: The case of tourism destination information. In W. Höpken, U. Gretzel, & R. Law (Eds.), Information and communication technologies in tourism 2009 (pp. 381–392). Vienna: Springer. Inversini, A., Cantoni, L., & Buhalis, D. (2009). Destinations’ information competition and Web reputation. Information Technology & Tourism, 11(3), 221–234. Kania, L. (2013). Social capital in the metropolis BrabantStad. In exploring the social impacts of events. London: Routledge. Knoke, D., & Kuklinski, J. (1991). Network analysis: Basic concepts. In Markets, hierarchies and networks: The coordination of social life (pp. 173–182). London: Sage. 294 A. Inversini et al. Latane, B. (1981). The psychology of social impact. American Psychologist, 36(4), 343. Lin, Y. (2012). A critical review of social impacts of mega-events. International Journal of Sport & Society, 3(3), 57. Morris, R. (1994). Computerized content analysis in management research: A demonstration of advantages & limitations. Journal of Management, 20(4), 903–931. Muniz, A. M., Jr., & O’guinn, T. C. (2001). Brand community. Journal of Consumer Research, 27(4), 412–432. Neuhofer, B., Buhalis, D., & Ladkin, A. (2013). A typology of technology-enhanced tourism experiences. International Journal of Tourism Research. doi:10.1002/jtr.1958. O’Brien, D. (2007). Points of leverage: Maximizing host community benefit from a regional surfing festival. European Sport Management Quarterly, 7(2), 141–165. Page, S., & Connell, J. (2011). The Routledge handbook of events. London: Routledge. Picard, D., & Robinson, M. (2006). Remaking worlds: Festivals, tourism and change. In D. Picard & M. Robinson (Eds.), Festivals, tourism and social change: Remaking worlds (pp. 1–31). Buffalo: Channel View. Rosenthal, J. A. (2011). Statistics and data interpretation for social work. New York: Springer. Schulenkorf, N., & Edwards, D. (2012). Maximizing positive social impacts: Strategies for sustaining and leveraging the benefits of intercommunity sport events in divided societies. Journal of Sport Management, 26(5), 379–390. Scott, N., Cooper, C., & Baggio, R. (2008). Network analysis and tourism. Buffalo: Channel View. Sharpley, R., & Stone, P. R. (2011). Socio-cultural impacts of events: Meanings, authorized transgression, and social capital. In S. In Page & J. Connell (Eds.), The Routledge handbook of events. London: Routledge. Swart, K., & Bob, U. (2004). Measuring the social impacts of events: Resident reactions to the North Sea Jazz Festival. In Proceedings of the 2005 IIPT (International Institute for Peace through Tourism) Africa Conference (pp. 1–5). Williams, N., Inversini, A., Buhalis, D., & Ferdinand, N. (2014). Social media and festivals as destination marketing tool: A study on Twitter Conversations. In Presented at the Academy of Marketing Conference 2014, Bournemouth, July 2014. Zaglia, M. E. (2013). Brand communities embedded in social networks. Journal of Business Research, 66(2), 216–223. Zhang, G., Ding, Y., & Milojević, S. (2013). Citation content analysis (CCA): A framework for syntactic and semantic analysis of citation content. Journal of the American Society for Information Science and Technology, 64(7), 1490–1503.