Using social network analysis to explain communication characteristics of travel-related electronic word-of-mouth on social networking sites

Tourism Management 46 (2015) 274e282

Contents lists available at ScienceDirect

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

Using social network analysis to explain communication characteristics of travel-related electronic word-of-mouth on social networking sites Qiuju Luo a, b, *, Dixi Zhong a, b, 1 a b

School of Tourism Management, Sun Yat-sen University, Building 329, 135 Xingangxi Road, Guangzhou 510275, PR China Center for Tourism Planning and Research, Sun Yat-sen University, Building 329, 135 Xingangxi Road, Guangzhou 510275, PR China

h i g h l i g h t s
We viewed eWOM communication on SNSs as a network based on social relationships.
We examined social ties and network structure with social network analysis.
Travel-related eWOM communication relies on strong, middling, or weak social ties.
The communication is structured, loose-knit, flat, and of high centrality.
Travel-related eWOM on SNSs tends to be dominated by travel interests.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 26 February 2013 Accepted 7 July 2014 Available online 26 July 2014

Social networking sites (SNSs), which are platforms based on user interactions, currently play increasingly important roles in sharing electronic word-of-mouth (eWOM) among tourists. Viewing eWOM communication on SNSs as a network based on the users' social relationships, this study applied social network analysis to examine the communication characteristics of travel-related eWOM on SNSs from the perspective of both ego and whole networks. Results show that travel-related eWOM communication via SNSs relied on existing social relationships, ties of which can be categorized as strong, of middling strength, or weak. Furthermore, the effect of transmitted information was stronger than that of influential decision-making. The communication network studied was found to be structured, loose-knit, flat, and of high centrality. These results enrich current research on the effects of eWOM and provide a dynamic perspective for understanding how eWOM disseminates and influences users through interactions. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Travel-related electronic word-of-mouth Communication characteristics Social networking sites Social network analysis Ego network Whole network

1. Introduction A significant symbol of Web 2.0, the boom in social networking sites (SNSs) has also aroused a worldwide upsurge in tourism destination marketing. With SNSs, a great deal of tourists post and share real-time feelings (Gretzel, 2006; Pan, MacLaurin, & Crotts, 2007), as well as travel reviews, opinions, and personal experiences while traveling (Xiang & Gretzel, 2010). In particular,

* Corresponding author. School of Tourism Management, Sun Yat-sen University, Building 329, 135 Xingangxi Road, Guangzhou 510275, PR China. Tel.: þ86 20 84112735/13450357112. E-mail addresses: [email protected] (Q. Luo), [email protected] (D. Zhong). 1 Tel.: þ86 18011718710. http://dx.doi.org/10.1016/j.tourman.2014.07.007 0261-5177/© 2014 Elsevier Ltd. All rights reserved.

individuals younger than 35 years old with at least a college degree chiefly participate in sharing travel experiences and photos on SNSs (Lo, McKercher, Lo, Cheung, & Law, 2011). Given the general popularity of sharing photos on SNSs, photos depicting travel have especially become a way of self-expression and self-image construction among younger generations (Lo et al., 2011). As mobile Internet capabilities progress, users more often share travel information whenever and wherever possible, which makes sharing via SNSs increasingly prevalent. In fact, travel information provided by SNSs has quickly become commonplace in the day-to-day lives of SNS users. SNSs such as Facebook, Twitter, Myspace, and Microblog are platforms with dynamic, multimodal features by which users can post, share, and discuss interests with other interested users (Jansen, Zhang, Sobel, & Chowdury, 2009). These features of SNSs

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

expand users' social circles as well as increase the frequency of interpersonal contact. Unlike traffic on other websites, users more often form close-knit relationships with each other (Ding & Wang, 2010). Given the strength of these ties, SNSs have transformed traditional information dissemination that relies on central mass media (e.g., newspaper and television). With the popularity of SNSs and general Internet use, a dual-core dumbbell structure of online information dissemination has emerged that includes both mainstream forums and microblogs as well as mainstream portals, which as the two core sources of influence have transformed how peer-to-peer influence works (Li, 2011). In terms of consumption, consumers are no longer passive recipients of information; instead, they actively engage in peer-to-peer product recommendations and electronic word-of-mouth (eWOM) (Chu & Kim, 2011). eWOM offers a useful perspective from which to study information dissemination and its influence on users and followers. Since the development of Web 2.0, traditional word-of-mouth (WOM) has had to accommodate eWOM (Chatterjee, 2001; Hennig-Thurau, Gwinner, Walsh, & Gremler, 2004), which by comparison is more influential due to its speed, convenience, broadcast appeal, and lack of the pressures of face-to-face interaction (Sun, Youn, Wu, & Kuntaraporn, 2006). Another aspect of such influence is that any communication and contact between communicators and receivers might alter the recipient's attitude, especially regarding purchase decisions (Cheung, Lee, & Thadani, 2009; Kiecker & Cowles, 2002; Park & Kim, 2008; Park & Lee, 2008). Likewise, travel-related eWOM on SNSs may significantly affect the cognition and behavior of potential tourists. Tourism is an experiential good; consumers cannot perceive the quality of tourism products in advance. Therefore, interpersonal communications have become an important technique to reducing the risks of travel (Murray, 1991). Litvin, Goldsmith, and Pan (2008) point out that interpersonal influence and WOM were ranked the most important sources of information for purchase decisions. Partly as a result, Chu and Kim (2011) suggest that product-focused eWOM on SNSs is a unique phenomenon with important social implications. Therefore, the characteristics of communication via eWOM on SNSs requires more sustained attention, particularly from the perspective of network structure and social relationships, which allows a more thorough examination of how interpersonal influence can spread among users and followers (Chu & Kim, 2011). From this perspective, studying the communication characteristics of travelrelated eWOM on SNSs can expand the present understanding of eWOM's influence, especially as it pertains to tourists and the tourism industry. Currently, SNS regarding tourism has received scant scholarly attention. Most research exploring the function of SNSs for locating tourism information, as well as users' motivations and behavior, has neglected to investigate communication among users. By contrast, eWOM communication and how it affects consumers' purchase decisions has gradually attracted the attention of researchers (Jansen et al., 2009; Lee & Youn, 2009; Riegner, 2007). Current research is conducted from three perspectivesdnamely, those of the communicator, the receiver, and the communication process. Although studies on the communication process are well outnumbered by those on communicators and receivers, recent research has begun to study the social characteristics of eWOM communication. Nevertheless, most studies thus far have considered consumers to be independent individuals and have thus emphasized the effects of eWOM on online purchase decisionmaking, while research on eWOM via SNSs remains in its infancy. In the meantime, eWOM communication in those studies is static, for few have conducted their research from a dynamic perspective and considered communication as a dynamic dissemination process. Therefore, this study focuses on the communication of travel-

275

related eWOM on SNSs to underscore its practical and academic significance. To these ends, this study performed social network analysis (SNA) to examine the communication characteristics of travelrelated eWOM on SNSs from the perspective of social ties and network structure. Its results not only enrich the existing theoretical research, but also provide further inspiration for conducting effective word-of-mouth marketing on SNSs in the tourism industry. 2. Literature review 2.1. SNS research in tourism Most research on SNSs has been published since 2008 and primarily emphasized user motives and behaviors. Among SNS research, the few travel-related studies can be grouped into two categories. On the one hand, most studies have considered SNSs to be one kind of social media in terms of their use for travel-related information searches. Using Google as a search engine, Xiang and Gretzel (2010) investigated the role of social media in online searches for travel-related information. The results showed that SNSs were not yet the main sources for users seeking travel-related information. Meanwhile, other research has suggested that user trust of travel websites varies significantly; the three types considered most trustworthy were official websites of tourism bureaus, websites of travel agencies, and third-party websites (Burgess, Sellitto, Cox, & Buultjens, 2011; Yoo, Lee, & Gretzel, 2009). Though trust of SNSs was lower than expected and SNSs are far from the most popular way to gather travel-related information, the reasons for both conditions have gone unaddressed in these studies. Furthermore, rapid changes that occur as mobile Internet become popularized may have altered the conditions in recent years. On the other hand, tourism studies have also focused on the use of SNSs in terms of user characteristics and motivations for sharing. Current studies in this category remain in the descriptive stage. Lo et al. (2011) found that most people sharing travel photos were young and well-educated, as well as had substantial incomes, rich travel experiences, and a willingness to involve themselves in the destination. Huang, Basu, and Hsu (2010) identified three functional motives for sharing travel-related information via SNSsdnamely, obtaining travel information, disseminating information, and documenting personal experiencesdand that of these motives, obtaining travel information was the most important. Both studies described nevertheless failed to present the characteristics of the social networkdnamely, the effect of social features on tourists. In sum, research of SNSs in tourism remains in its infancy. Though earlier studies explored the function of SNSs for locating travel information, most neglected to investigate the communication process, for few conducted their research from a dynamic perspective. If substantial characteristics of SNSs have been overlooked, such oversight precludes further understanding of the acquisition and impact of travel information. At the same time, since few studies viewed online travel-related information as eWOM, we have viewed travel-related information as such and, moreover, sought to provide a dynamic perspective for understanding how eWOM disseminates information and influences users. 2.2. Communication research on eWOM Current research on eWOM is conducted from three perspectivesdnamely, those of the communicator, the receiver, and the communication process.

276

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

Research conducted from the perspective of the communicator emphasizes the motives for eWOM communication. In this strain of the literature, Hennig-Thurau et al. (2004) identified eight motives for sharing product reviews, including social benefits and the need for advice. These results also suggest that social relationships among users cannot be ignored in eWOM studies. The perspective of the receiver has drawn the most scholarly attention. In this strain, relevant research has primarily discussed the effect of eWOM on the receiver in two aspects: the receiver's attitude and the receiver's willingness to purchase. Regarding the effect of eWOM on the receiver's attitude, various factors influence receivers' adoption of eWOM, including both the features of eWOM and of receivers. The features of eWOM include quantity of eWOM (Doh & Hwang, 2009), source characteristics (e.g., its reliability, objectivity, and expertise), information characteristics (e.g., its general allure, completeness, accuracy, and timeliness) (Chen & Zhang, 2008), relevance, and completeness (Cheung, Lee, & Rabjohn, 2008). In particular, the usefulness of information is a mediator that influences eWOM adoption (Cheung et al., 2008). At the same time, the features of the receiver include involvement and prior knowledge, both of which variously affect the recipient's attitude (Doh & Hwang, 2009). By comparison, studies on the effect of eWOM in purchase decisions are more diverse, for scholars have conducted research on a variety of influential factors. Park and Kim (2008) concluded that benefit-centric eWOM has a greater influence on the willingness to purchase for consumers who lack expertise, while attribute-centric eWOM exerts a greater influence on consumers with professional knowledge of the product. Poyry, Parvinen, Salo, and Blakaj (2012) showed that, compared to utilitarian information searches, hedonic information searches significantly improve the consumers' perception of eWOM's usefulness and shortens the decision-topurchase time. Though it is clear that SNSs prefer hedonic information searches, whether there is any perceptible influence on tourists' decision-making requires further exploration. De Bruyn and Lilien (2008) developed a multi-stage model to identify the role of eWOM plays during each stage of recipients' decisionmaking process. Compared to the perspectives of the communicator and receiver, the perspective of the communication process has received scant scholarly attention, though recent research has begun to study the social characteristics of eWOM communication. On one hand, social relationships between consumers come into notice. Among studies of eWOM, Chu and Choi (2011) have evaluated the effects of social relationships between consumers' purchase decisions on SNSs. Their results suggest that Chinese users communicate most and most trustfully with users with whom they have strong social relationships, thus the social capital of a preexisting social relationship plays a significant role in Chinese users' eWOM communication. By contrast, Americans interact more with extended social circles or with other users with whom they have no social relationship. Chu and Kim (2011) developed and tested a conceptual framework that identifies tie strength, homophily, trust, normative and informational interpersonal influence as an important antecedent to eWOM behavior in SNSs, and tie strength is positively associated with eWOM behavior. As might be expected, several researchers have suggested that WOM communication has relied on social relationships and that consumers were inclined to trust acquaintances and people with whom they maintained strong social ties (Brown & Reingen, 1987), family members, and friends (Jansen et al., 2009). In the era of Web 2.0, social interaction on SNSs determined by social relationships continues to merit in-depth investigation. On the other hand, some researchers studying the features of eWOM communication networks have produced results indicating

that eWOM communication networks are structured instead of random. Among these researchers, Vilpponen, Winter, and Sundqvist (2006) conducted a case study of communication on personal websites that used a downloadable banner to show resistance to a proposed copyright law in Finland. Using SNA, Vilpponen et al. (2006) concluded that eWOM communication can be characterized as a loose-knit network of high centralization and cliques. In another study, by modeling an eWOM communication network on multi-agent simulation, Jiang (2009) found that the structure of any eWOM communication network influences both the scale and efficiency of communication. Altogether, current research on eWOM regarding the communication process remains insufficient. On the one hand, since researchers viewed users as independent individuals, most research failed to consider the relationships among users and the pathways of communication most taken by users seeking to share and exchange information. On the other hand, few studies address eWOM communication on SNSs. Certain features of SNSsdstrong interactivity and timeliness, to name twodare likely to distinguish SNSs from general websites. Our study has thus aimed to provide indepth research on travel-related eWOM communication via SNSs from the aspect of user interaction. 3. Research design In all social communication processes, at least two individuals are needed to form an information-sharing relationship in order to share information symbols (Schramm & Porter, 2010). In this sense, the process of information exchange should not be viewed as a specific behavior (i.e., A acts upon B) but as information sharing that leads to a common understanding (Schramm & Porter, 2010). SNA views the social structure as an interpersonal network that emphasizes interpersonal relationships, the content of the relationships, and the interpretation of social phenomena within the structure of a social network (Luo, 2010). A social network is a collection of social actors and the relationships among them (Liu, 2009). Consequently, each node in the network represents one actor, which can be a social unit or entity, and each link represents the relationship between the actors. SNA emphasizes three network levels: ego, partial, and whole. During the past 30 years, SNA has been applied to many studies in sociology, organizational behavior, and social relationships. More recently, SNA has been increasingly applied to social media-based communication research. This study applied SNA to answer research questions from the perspective of each single relationship and then extended that perspective to the whole network. In short, this study concerns two aspects, one is the features of social ties of each communication pathway, and the other is the structural features of a travel-related eWOM communication network. To do so, we applied ego-network analysis to examine social ties as well as whole-network analysis to measure the structure of travel-related eWOM communication network (Fig. 1). The interaction between users on SNSs can be silent (i.e., not directly observable) or visible (Pempek, Yermolayeva, & Calvert, 2009). Since silent contact is difficult to assess, the present analysis was conducted based on visible contact. Visible interactive behavior between users, including their comments and forwarding comments, represents the completion of information dissemination. 3.1. Ego-network analysis From the perspective of social ties, ego-network analysis was used to analyze the strength of social ties between the

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

277

Fig. 1. Research framework.

communicator and the receiver in travel-related eWOM communication. An ego network refers to any network consisting of an individual and other users directly connected to him or her; it is used to interpret the relationship features between the communicator and receiver in travel-related eWOM. In this study, data were collected by questionnaire. A nomination method was used to identify five users who had made travelrelated eWOM communication with respondents during a period of six months (October 2011eApril 2012). After the ego network was constructed, a nomination interpretation method was used to describe and measure each communication relationship. The questionnaire design was based on the standard ego-network questionnaire exhibiting high reliability used in general social surveys in the U.S. formulated by Burt (1984), which uses three constructs to interpret questionnaire data: contact duration, contact frequency, and intimacy (Granovetter, 1973). The questionnaire used in this study was revised based on local research conducted by Luo and Xie (2008), added a construct (i.e., “relationship between close circles of friends”), and adjusted the range of years of each item for the construct of contact duration, as well as the measures and contents of the items in the construct of intimacy. Additionally, we added travel-related questions to explore respondents' tourism preferences, common travel experiences, and reciprocal behaviors in travel-related eWOM. Five undergraduate students from different universities were recruited to take the pilot test, after which the questionnaire was adjusted accordingly. The questionnaire was distributed from April 1e8, 2012 to highfrequency users of SNSs in Guangdong, China. Users were mostly office workers and college studentsdsome foreign exchange studentsdselected based on three considerations. First, the sample originated from the primary group of SNS users in China, which is representative of China. Second, the questionnaire design was new and informative and thus required respondents with adept comprehension skills and patience. Third, active SNS users (i.e., those who log in more than two or three times per week) may have various travel-related eWOM communication behaviors. To ensure a high reliability of questionnaire results, we distributed the questionnaires to each participant one at a time. In this study, an ego network consisted of each respondent and up to five of his or her nominated contacts of travel-related eWOM. Each network was a sample set, in which the respondent served as the core, while each communication relationship directly connected to the respondent formed an independent sample. In total, 64 questionnaires were collected; those of participants whose SNS use frequency was less than two or three times per week or who could not provide a complete dataset of at least one nominated contact were excluded. Altogether, 61 questionnaires (95.3%) were deemed valid. A total of 303 (97.7%) independent samples was

obtained, of which 289 were deemed valid; 14 samples with missing values were excluded. The profile of respondents is shown in Table 1. The sample sets generally represented the typical SNS user. 3.2. Whole-network analysis Whole-network analysis examines the network structure of eWOM communication. For this study, a representative microblog was selected as a sample. This study only focused on the network characteristics of travel-related eWOM in a single circle of microblogging relationships instead of multiple circles. With wholenetwork analysis, the study aimed to develop a directional adjacency matrix to analyze travel-related eWOM communication. The whole network refers to all relationships among all group members

Table 1 Profile of respondents (n ¼ 61). Demographic characteristic Gender Male Female Age <18 years 19e30 years 31e40 years 41e50 years >50 years Highest level of education achieved Junior high school High school or technical secondary school University or college Postgraduate Travel frequency within a year More than twice Once None Occupation Governmental agencies or institutions Corporations or enterprises in the service industry Individual industrialists and businessman Researchers and teachers Retired Housewives Students Other SNS use frequency Everyday Two or three times weekly Weekly Monthly Rarely

n

Percent

26 35

42.6 57.4

0 59 2 0 0

0.0 96.7 3.3 0.0 0.0

0 4 55 2

0.0 6.6 90.2 3.3

47 12 2

77.0 19.7 3.3

4 6 0 0 0 0 51 0

6.6 9.8 0.0 0.0 0.0 0.0 83.6 0.0

53 8 0 0 0

86.9 13.1 0.0 0.0 0.0

278

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

(Liu, 2009). Network density, graph centralization, centrality, and subgroup analysis were four important measurements used to explore the network cohesion, integration, role and position, and composition and structure of the travel-related eWOM communication network. The sample microblog originated on Sina Microblog, the most popular microblogging platform in China. Users on Sina Microblog can be classified into two types: authenticated and ordinary users. An authenticated user must be a well-known figure in a particular field with an authentication icon highlighted. Since there are no identity constraints on ordinary users, they form the major group and account for a larger percentage of users. To understand the communication structure of ordinary users was therefore more remarkable for destination marketing regarding SNSs. In this study ordinary bloggers were used as subjects for whole-network analysis; the microblog account of user Gaoli_Ivy provided data. Factors such as the quantity of travel-related microblogs and the degree of interaction were considered during the sample selection to ensure a sufficient amount of data. The selected blogger loved traveling and on average traveled threeefive times per year, including one or two long-distance journeys. After registering as a user in September 2010, from September to December 2011 she posted a total of 92 travel-related messages, each of which addressed topics such as travel experiences, air travel, accommodation deals, and recommendations for travel destinations. The blogger followed a total of 120 blogging friends and was followed by 227 users (March 26th, 2012). Data from September to December 2011 were collected, which consisted of all travel-related microblogging contacts followed by the blogger and her followers. The research dates included two public holidays in Chinadnamely, the Mid-Autumn Festival and National Daydto ensure data adequacy. The blogger was asked to review her use history in order to organize her travel microblogs and microblog accounts she had commented on from September to December 2011. During the same period, we read travel-related microblogs in order to construct a travel-related eWOM communication pathway for other users. A case-by-case adjacency matrix was constructed for wholenetwork analysis, for which each node represented an independent microblog user. The communication relationships between the nodes indicating travel-related eWOM communication behaviors, which included forwarding or commenting on travel-related microblogs, were represented by directional links. The data were screened, and inactive users with fewer than 50 followers were excluded. A 155  155 adjacency matrix was eventually constructed for data analysis. UCINET 6 statistical analysis software was used for wholenetwork analysis. 4. Results 4.1. Ego networks 4.1.1. Results of tourist behavior Regarding travel as a hobby, 75.6% of contacts enjoyed travel, 21.9% neither particularly enjoyed nor disliked travel, and 2.5% did not enjoy travel. In terms of common travel experience, 39.4% of contacts had traveled with respondents in the previous year. Because the frequency of travel was less than the frequency of daily entertainment activities, 39.4% can be considered a large percentage. Finally, regarding the sharing of travel information, 53.0% of contacts shared travel information with respondents. Citing other users or sending private messages were two ways they had shared travel-related eWOM and can thus be viewed as individual-toindividual eWOM communication. Information sharing was a

form of reciprocal behavior between the respondents and contacts. These results suggest that users who enjoyed travel or had common travel experience with communicators were more likely to have visible contact and more inclined to travel-related eWOM communication (Table 2). 4.1.2. Results of social ties Three constructsdnamely, contact duration, contact frequency, and intimacydwere used to measure the strength of travel-related eWOM communication ties on SNSs. To distinguish communication relationships by strength of social ties, a k-means analysis was conducted for 289 pairs of contact relationships. We conducted two-, three-, and four-category clustering analyses to better interpret and categorize the samples and finally selected a threecategory analysis since it best explained the differences. Variance analysis of the cluster results showed that all indicators from the three clusters were significantly different, which was consistent with the required statistical significance. Results of cluster analysis are shown in Table 3. The cluster characteristics of travel-related eWOM communication relationships included the following: Category I: Strong social ties. A total of 90 relationships (31.1%) fell into this category, the most significant characteristic of which was a high average of the five indicators. Average contact frequency in these relationships occurred more than two or three times weekly, while the average contact duration was threeeten years. The topics and behaviors of communication were intimate and diverse, and on average, a small group of familiar and common friends was shared among the respondents and their contacts. Of the contacts, 77.8% shared travel-related eWOM by citing other users and sending private messages to and from the respondents with strong reciprocity. Category II: Social ties of middling strength. Compared to Category I, slightly fewer relationships (n ¼ 78, 27.0%) were considered to have ties of middling strength. Average contact frequency for this category was once or twice per week; contact duration was slightly shorter than that of the contacts with strong social ties; the topics and behaviors of communication were more general; and the degree of overlap between the circles of close friends among contacts was slightly lower than of that of contacts with strong social ties. In this category, 51.3% of contacts were reciprocal subjects of travel-related eWOM from the respondents. Based on the indicators, members of this category of communication relationship were determined to have social ties of middling strength. Category III: Weak social ties. Given their low scores for each item, most relationships (n ¼ 121, 41.9%) belonged to the category encompassing contacts with weak social ties. Contact frequency within the sample was as little as oneethree times monthly, while the average contact duration was from one to three years and the intimacy of topics and behaviors was extremely low. Furthermore, the degree of overlap between

Table 2 Analytical results of travel behaviors of travel-related eWOM receivers. Analysis of travel behavior Contact enjoys travel

Contact traveled with respondent in the previous year Would share online travel information with respondent

Yes Neutral No Yes No Yes No

n

Percent

211 61 7 113 174 151 134

75.6 21.9 2.5 39.4 60.6 53.0 47.0

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

279

Table 3 Clustering analysis results.

Contact frequency Contact duration Intimacy of communication Intimacy of behavior Relationship between close circles of friends Mutual confiding Percentage

Category I: Strong social ties

Category II: Social ties of middling strength

Category III: Weak social ties

M

SD

M

SD

M

4.01 3.98 14.42 26.93 3.21 77.8% 31.1%

1.13 0.80 5.16 3.03 0.63

3.38 3.86 11.10 15.08 3.06 51.3% 27.0%

1.30 0.80 4.35 4.48 0.67

2.63 3.11 3.23 2.63 2.61 35.3% 41.9%

SD 1.34 1.02 3.22 2.84 1.06

Note. Score for indicators ranged from 1 to 5 points for contact frequency, 0 to 5 points for contact duration, 1 to 22 points for intimacy of communication, 1 to 31 points for intimacy of behavior, and 1 to 4 points for a close circle of friends; SD ¼ standard deviation.

close circles of friends was low; most individuals in these relationships did not have contacts in common. Above all, since social contact in this category was sporadic, these relationships were determined to have weak social ties. The social tie strength was comparable to that of general colleagues and classmates, neither of whom share much contact. In an empirical study, Marsden and Campbell (1984) showed that the degree of intimacy is the best indicator to measure the strength of social relationships. In this study, the degree of intimacy (i.e., the intimacy of topics and behavior indicators) was the primary index for distinguishing social tie strength. More recently,  (2010) categorized virtual social rePetroczi, Nepusz, and Bazso lationships, which guided the definition of the strong social ties category in this paper, which exhibited intimate friendship since contacts were familiar with and supported each another (e.g., by offering suggestions, sympathizing, and giving spiritual support). By contrast, the categories of middling and weak social ties were characterized the relationships of acquaintance, in which contacts only knew one another in passing or were friends. The contacts with middling and weak social ties might share interests, hobbies, and exchanged messages in private. 4.2. The whole network The 155  155 directional multi-value matrix constructed from the data of 155 users was screened and coded. From the subsequent analysis we excluded 100 users who had no visible communication of travel-related content with any other users in their social circles from September to December 2011 in order to examine the characteristics of travel-related eWOM communication without the interference of the data of users who had not participated in relevant communication. A 55  55 matrix of travel-related eWOM communication was constructed, in which each node was coded in sequential order (AA, AB, …; BA, BB, …; CA, CB, …). Fig. 2 shows the general pathways and active members in a travel-related eWOM communication process. The contacts between users represented by the following nodes in the red circle were more frequent and dense: AJ, BH, BM, CD, CF, CM, CO, CQ, CR, CW, and DJ, as well as pairs AA and AI, DJ and BI, and CQ and CO. 4.2.1. Network density Network density is a measure of network cohesion (Webster & Morrison, 2004). In this sense, density signifies the ratio of the actual number of links versus the maximum number of links possible in the network (0e1). To measure network density, the multi-value matrix was converted into a two-value matrix using UCINET 6. Nodes with information dissemination behavior between them were assigned values of one whether the information dissemination went both

ways or not. By processing the 55  55 directional two-value matrix in UCINET 6, sample network had a total of 105 ties of eWOM communication, which meant that network density was 0.0354. When density was measured among the groups with frequent contacts (represented by the red circle in Fig. 2), density increased to 0.1703, which was nevertheless quite low. The results above show that in most instances, the travel-related eWOM contact network on an SNS was loose-knit instead of densely connected. This result was significantly affected by the fact that not all contacts in eWOM communication had a connection with one another. 4.2.2. Graph centralization Graph centralization measures the overall cohesion or integration of a network and describes the extent to which such cohesion was organized around particular nodes (Scott, 2007). Regarding the degree of centrality of graph centralization, the outdegree centralization of the sample network was 73.73%, while the indegree was 20.92%. The betweenness centrality of the graph centralization was low (24.41%). A high degree of centrality of outward communication indicates that the level of information integration was high. Any node with high centrality, which indicates that the person has more travelrelated interaction with others, had a large effect in the network. By contrast, as the degree centrality of the eWOM-receiving network diminished, the receiving pathways became more diverse. Low betweenness centrality of the graph centralization suggests a low level of distortion, which indicates fast and effective eWOM communication. 4.2.3. Centrality analysis Centrality is an indicator of an individual's structural position that assesses the importance of the individual in the network (Luo,

Fig. 2. The diagram of the sample networks.

280

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

Table 4 The output of the centrality.

Table 5 The output of betweeness centrality.

N

SOC

SIC

N

SOC

SIC

N

SOC

SIC

N

SOC

SIC

N

BC

SBC

N

BC

SBC

N

BC

SBC

N

BC

SBC

AA AJ CQ CP DJ DL BI BP DO CR DM DE CB AK

75.93 24.07 18.52 12.96 9.26 7.41 5.56 5.56 5.56 3.70 3.70 3.70 1.85 1.85

24.07 5.56 9.26 1.85 5.56 1.85 5.56 1.85 1.85 1.85 1.85 0 1.85 1.85

CF BK CW CM CV AC DP AI BH AT BN AN AH AB

1.85 1.85 1.85 1.85 1.85 1.85 1.85 1.85 0 0 0 0 0 0

9.26 0 1.85 1.85 3.70 3.70 0 5.56 3.70 5.56 1.85 1.85 1.85 3.70

BX AF AE CD BZ CG BM CK AL CN AG AO AR AS

0 0 0 0 0 0 0 0 0 0 0 0 0 0

3.70 1.85 5.56 7.41 1.85 3.70 3.70 1.85 1.85 1.85 5.56 1.85 1.85 1.85

CU AV BA DB DD CH DF DI BO CO BQ BV BW

0 0 0 0 0 0 0 0 0 0 0 0 0

3.70 1.85 1.85 1.85 1.85 5.56 1.85 1.85 1.85 9.26 1.85 3.70 3.70

AA CQ AJ

704.17 125.83 77.67

24.60 4.40 2.71

DJ BP AI

38.00 17.00 16.50

1.33 0.60 0.58

BI AC CF

9.00 8.33 5

0.31 0.29 0.18

DO CP DM

1.50 0.50 0.50

0.05 0.02 0.02

Note. N ¼ nodes; SOC ¼ standard outdegree centrality; SIC ¼ standard indegree centrality.

2010). This indicator is used to reflect the core-margin position of actors in the diagram by focusing on each node in the network. There are three centrality indexes: degree, closeness, and betweenness (Scott, 2007). The higher the degree centrality index, the more actors the user has contact within the network, thus the more unofficial power and greater effect the individual exerts in the network. By contrast, users with high betweenness centrality occupy the central position of contact between two members in the network. The more opportunities this user has to guide resources, the more critical the position he or she occupies in the flow of resources. Regarding outdegree centrality, AA was the main core in the network, with a standard outdegree centrality of 75.93, followed by AJ (24.07) and CQ (18.52). The standard outdegree centrality of more than 33 actors was zero. The results suggest that the sample network had a structure dominated by one core surrounded by several secondary cores. Because the sample network was the extended microblog network of AA, the attributes of AA did not have a reference value. Among the 10 individuals with the highest outdegree centrality (excluding AA), AJ, CQ, DJ, BI, and CR were all travel lovers,2 whereas DL, DO, and DM were the official microblogs of AirAsia, Asiago, and Qyer.com, respectively. CP and BP traveled frequently on business (i.e., approximately twice every six months). Regarding indegree centrality, the standard indegree centrality of three nodes was high; AA was the highest (24.07). The standard indegree centrality of CQ, CF, and CO was 9.26 and then <8 for the remaining nodes. The receivers of eWOM were relatively evenly distributed and not centralized, which suggests that travel-related eWOM attracted attention from a relatively large number of users, not only a few actors. The sample network formed a network structure of dispersed receiving (Table 4). The betweenness centrality of only 12 nodes in the network was greater than zero. AA occupied the central position, with a standard betweenness centrality of 24.60, followed by CQ and AJ. AA occupied the center of the network and had a great effect on the thoughts of the other actors in the online social circles. Except for these three nodes, the betweenness centrality of the remaining tourists was low, and many tourists were marginal (Table 5). 4.2.4. Subgroup analysis Subgroup analysis examines the group characteristics of cohesion in the network by analyzing the substructures of the whole network. Generally, a subgroup refers to a coalition of many

2 According to an interview with blogger AA. Authors listed users who had contacts with blogger AA, and asked AA to select travel lovers.

Note. N ¼ nodes; BC ¼ betweeness centrality; SBC ¼ standard betweeness centrality.

contacts who share a goal and have many stable contacts with each another. From the perspective of social psychology, an individual is an actor in the group and subjected to the concepts, influences, norms, and values of the group (Liu, 2009). Therefore, it was important to investigate whether there was a subgroup in the travel-related eWOM communication network in order to understand how eWOM further affects receivers. A component analysis was conducted on the 55  55 directional two-value matrix and revealed a strong component composed of 14 nodes (AA, AI, AJ, BI, CB, CM, CP, CQ, CR, CW, DJ, DL, DM, and DO). A strong component refers to a component in which the connection direction is considered. In the subgroup of 14 nodes, AI, AJ, BI, CM, CQ, CR, CW, and DJ were travel lovers, while DL, DM, and DO were the official microblogs of three travel websites. The subgroup members shared an interest in travel and were therefore more likely to form a close-knit subgroup in the travel-related eWOM communication network. A k-core collapse sequence analysis was also conducted to analyze whether the sample network on SNSs was structured. The index analyzes the similarities in relationships and structure between the component and other nodes of the sample. Results show that the core collapse sequence was 0, 0, 0.44, 0.71, and 0.89. The core collapse sequence is thus gradual as k increases from zero, which suggests that the communication and contact in the travelrelated eWOM network was not random but structured (Table 6). 5. Discussion and conclusions As an early empirical attempt to understand the characteristics of travel-related eWOM communication on SNSs, this study examined the social ties and social network structure with SNA. It thus offered a new perspective for better understanding how eWOM disseminates and influences within user interactions. In ego-network analysis, we examined the social relationship variables among users in travel-related eWOM communication on SNSs. We specifically examined tie strength as a potential predictor of interpersonal influence in eWOM communication. In wholenetwork analysis, we also examined the network structure of travel-related eWOM communication. Our results first show that travel-related eWOM communication via SNSs relied on existing social relationships, which can be categorized into three groups: having strong social ties, social ties of middling strength, or weak social ties. Only 0.7% relationships were newly established between respondents and contacts. 1.7% relationships had been established for six monthseone year. SNSs stood apart from other social media in encouraging their users with existing social relationship to interact online, which underscores its academic significance. Relationships with weak social ties formed Table 6 The output of the k-core collapse sequence analysis. k

k-remainder

k-remainder percentage

0 1 2 3 4

0 0 24 39 49

0 0 0.44 0.71 0.89

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

the largest category, while the other two categories were similar in size. Contacts in relationships with strong, middling, or weak social ties played different roles in communications with information receivers (Granovetter, 1973); effects on the promotion and optimization of social learning therefore also differed. This conclusion confirms the results of Chu and Choi (2011), who reported that Chinese users were more likely to contact familiar users on SNSs instead of expanding their existing social relationships. In the present study, we further segmented tie strength to better understand the characteristics of eWOM communication and its effects. Secondly, concerning its effect of communication, results reveal that eWOM can transmit information and influence decisionmaking, though the effect of the former was stronger than that of the latter. Travel-related eWOM on SNSs could overcome spatiotemporal limitations and spread to all corners of the social network. Furthermore, eWOM also affected the attitudes and decisionmaking of contacts with strong social ties, since strong social ties were conducive to influencing others and building trust, whereas weak social ties were conducive to transferring knowledge and information. The fact that relationships with strong social ties occupied the smallest category suggests that travel-related eWOM plays a more important role in knowledge and information dissemination. Moreover, network density was low for travelrelated eWOM communication; the travel microblog did not inspire frequent contact among tourists, which weakened eWOM's influence. Characteristics of social relationships have been indexed to reflect how eWOM works among tourists. In this study, the effects of eWOM were divided into two kinds: eWOM transmission and influence. Compared to studies of trust of eWOM on SNSs (Burgess et al., 2011; Yoo et al., 2009), the present study offers a new perspective for examining the extent of eWOM's influence, though this perspective nevertheless requires further study. Thirdly, we found the communication of travel-related eWOM on SNSs to be dominated by travel interests, while information and influence were evenly disseminated among active travel-interested users. Individuals who loved to travel and had common travel experience were more likely to follow travel-related content in relationship circles on SNSs, make visible contact, be reciprocated, and have subsequent contact concerning travel-related eWOM. According to centrality analysis, the effect of these individuals cannot be ignored, given their significant centrality and impact. In a travel-related eWOM communication network, these people would be more likely to take important positions and from there influence other users in the network. Fourthly, we discovered that the communication network structure of travel-related eWOM on SNSs bears three characteristics. One, the communication of travel-related eWOM was not random but structured. The communication network of travel-related eWOM could be divided into subgroups. In the present study, the sample was dominated by a single subgroup with a close-knit network. The structure of the overall network and components was consistent, and there were no conditions in which dense areas were surrounded by marginalized nodes. Two, the communication of travel-related eWOM on SNSs was loose-knit based on social relations. The strength of most social ties was middling or weak and its density low. Three, the degree of centrality was high, while the degree of betweenness centrality was low in the sample network. Network structure therefore exhibits high centrality. To communicate travel-related eWOM, actors in the network would bypass redundant relationships, which implies that eWOM at important nodes (i.e., hubs) would influence other nodes and thereby flatten communication. The above results thus also essentially confirmed the conclusions of Jiang (2009) and Vilpponen et al. (2006). Similar to information dissemination via personal websites, travel-related eWOM communication on SNSs was loose-knit and occurred among small groups. The degree of connections

281

between nodes was not evenly distributed; a few nodes had many connections in the network, whereas the majority of nodes had only a few. Fifth and above all, this study's findings illuminate the process of travel-related eWOM communication on SNSs. To begin, the network structure of travel-related eWOM outflows was dominated by a core and closed by several secondary cores. Travelrelated eWOM revealed a high degree of centrality and integration. It was disseminated from the nodes of high centrality to those of low centrality with a high degree of information integration. Users who loved to travel were more likely to take an important role in travel-related eWOM communication. Moreover, the network structure of inflows was dispersed. The degree centrality of eWOM inflow was low; the pathways were diverse and exhibited a dispersed reception network structure. The receiving nodes of travel-related eWOM were relatively even with low centralization; travel-related eWOM could therefore attract the attention of many users, and accordingly, there were no instances in which travelrelated eWOM was concentrated to a few actors. By comparison, Jiang (2009) and Vilpponen et al. (2006) investigated the communication network structure on a macro level, while Chu and Kim (2011) and Litvin et al. (2008) both provided conceptual frameworks of eWOM dissemination that judged interpersonal influence to be an important variable. This empirical study examined the communication process from the perspective of social network, from where it was viewed as a dynamic, interactive process. Our findings gave the inspirations on how eWOM disseminating its influence via social relationships on SNSs. Perhaps above all, this study was based on social interaction, which is the core feature of SNSs. Altogether, our findings emphasize the importance of social relationships and social networks upon eWOM communication on SNSs by making the following contributions. First, travel information on SNSs was considered eWOM, which provided a new angle for studying emergent media respecting tourism. Second, this study focused on social interactions, whereas previous eWOM research primarily emphasized the perspective of individuals and viewed the communicator and receiver as independent individuals with no connections. However, any SNS is an interactive platform for users to establish contacts in social circles. The features of communication, such as social ties and communication network structure, likely influence the effect of eWOM significantly; therefore, neglecting to examine communication contacts and relationships between communicators and receivers will yield a distorted understanding of eWOM's effects. Furthermore, since travel-related eWOM communication was viewed as a network based on the user's social relationships, current eWOM studies have been enriched by this study's implication that eWOM dissemination and influence can be better understood given user interactions from a dynamic perspective. Third, though both Chu and Kim (2011) and Litvin et al. (2008) had proposed a conceptual framework of eWOM communication that considered interpersonal influence, empirical studies remain insufficient. By contrast, this study was conducted by using SNA, which constitutes an empirical study of travel-related eWOM on SNSs. To a great extent, this study therefore provides a practical way to study eWOM communication: focusing on interpersonal influence. SNSs act as amplifiers of travel information. Tourists intuitively comment on destinations and tourism experiences both while traveling and in retrospect. By reaching a wide audience in social circles and by depending on the strength of social ties in their relationships, users' perceptions of destinations and tourism products are greatly affected by eWOM on SNSs. The advantages and disadvantages of destinations and the travel experiences as evident in comments can be strengthened and amplified when provided by people close to the potential tourists. These tourists' contributions

282

Q. Luo, D. Zhong / Tourism Management 46 (2015) 274e282

constitute an important force given the communication's influence bolstered by strong social ties and the communication's power with weak social ties. Since SNSs have become important marketing tools in tourism, the phenomenon of travel-related eWOM on SNSs promises to become a topic increasingly visited by scholars and industry players alike. For these reasons, eWOM on SNSs requires further research, especially regarding its dynamic and interactive communication processes. Acknowledgments The research contained in the paper has been financially supported by a grant from National Natural Science Foundation of China (to Luo Qiuju) (No.40971041). The authors express their gratitude to Miss Gao, all respondents and proofreading editors who have offered help in this study. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi:10.1016/j.tourman.2014.07.007. References Brown, J. J., & Reingen, P. H. (1987). Social ties and word-of-mouth referral behavior. Journal of Consumer Research, 14(3), 350e362. Burgess, S., Sellitto, C., Cox, C., & Buultjens, J. (2011). Trust perceptions of online travel information by different content creators: some social and legal implications. Information System Front, 13, 221e235. Burt, R. S. (1984). Network items and the general social survey. Social Networks, 6(4), 293e339. Chatterjee, P. (2001). Online reviews: Do consumers use them? Advances in Consumer Research, 28, 129e133. Chen, M., & Zhang, J. (2008). An experimental research on the determinants of rediffusion intention of online word-of-mouth. Journal of Zhejiang University (Humanities and Social Sciences), 38(5), 127e135. Cheung, C. M. K., Lee, M. K. O., & Thadani, D. R. (September 16e18 2009). The impact of positive electronic word-of-mouth on consumer online purchasing decision. In Proceedings of the 2nd world summit on the knowledge society: Visioning and engineering the knowledge society. A web science perspective. Chania, Crete, Greece. Cheung, C. M. K., Lee, M. K. O., & Rabjohn, N. (2008). The impact of electronic wordof-mouth: the adoption of online opinions in online customer communities. Internet Research, 18(3), 229e247. Chu, S. C., & Choi, S. M. (2011). Electronic word-of-mouth in social networking sites: a cross-cultural study of the United States and China. Journal of Global Marketing, 24(3), 263e281. Chu, S. C., & Kim, Y. (2011). Determinants of consumer engagement in electronic word-of-mouth (eWOM) in social networking sites. International Journal of Advertising, 30(1), 47e75. De Bruyn, A., & Lilien, G. L. (2008). A multi-stage model of word of mouth influence through viral marketing. International Journal of Research in Marketing, 25(3), 151e163. Ding, H., & Wang, Y. (2010). Analyzing ‘opinion leader’ attributes in SNS cyberspace: an investigation of Douban.com. Journalism & Communication, 3, 82e91. Doh, S. J., & Hwang, J. S. (2009). How consumers evaluate eWOM (electronic wordof-mouth) messages. Cyberpsychology & Behavior, 12(2), 193e197. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360e1380. Gretzel, U. (2006). Consumer generated contentetrends and implications for branding. E-Review of Tourism Research, 4(3), 9e11. Hennig-Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004). Electronic wordof-mouth via consumer-opinion platforms: what motivates consumers to articulate themselves on the Internet? Journal of Interactive Marketing, 18(1), 38e52. Huang, Y., Basu, C., & Hsu, M. K. (2010). Exploring motivations of travel knowledge sharing on social network sites: an empirical investigation of U.S. college students. Journal of Hospitality Marketing & Management, 19, 717e734. Jansen, B. J., Zhang, M., Sobel, K., & Chowdury, A. (2009). Twitter power: tweets as electronic word of mouth. Journal of the American Society for Information Science and Technology, 60(11), 2169e2188. Jiang, S. (2009). Study on formation mechanism of online word-of-mouth network based on multi-agent simulation (master's thesis). Retrieved from http://www.cnki.net/. Kiecker, P., & Cowles, D. (2002). Interpersonal communication and personal influence on the Internet: a framework for examining online word-of-mouth. Journal of Euromarketing, 11(2), 310e325. Lee, M., & Youn, S. (2009). Electronic word of mouth (eWOM): how eWOM platforms influence consumer product judgement. International Journal of Advertising, 28(3), 473.

Li, B. (2011). Study on spatial communication structure of network events: an analysis of 40 network events in recent years. Journalism & Communication, 3, 90e99. Litvin, S. W., Goldsmith, R. E., & Pan, B. (2008). Electronic word-of-mouth in hospitality and tourism management. Tourism Management, 29(3), 458e468. Liu, J. (2009). Lectures on whole network approacheA practical guide to UCINET. Shanghai: Truth & Wisdom Press, Shanghai People's Publishing House. Lo, I. S., McKercher, B., Lo, A., Cheung, C., & Law, R. (2011). Tourism and online photography. Tourism Management, 32, 725e732. Luo, J. (2010). Lectures on social network analysis (2nd ed.). Beijing: Social Sciences Academic Press (China). Luo, J., & Xie, Z. (2008). Local research on social ties. In J. Luo (Ed.), Lectures on social network analysis (2nd ed.) (pp. 102e119). Beijing: Social Sciences Academic Press (China). Marsden, P. V., & Campbell, K. E. (1984). Measuring tie strength. Social Forces, 63(2), 482e501. Murray, K. B. (1991). A test of services marketing theory: consumer information acquisition activities. Journal of Marketing, 55, 10e25. Pan, B., MacLaurin, T., & Crotts, J. C. (2007). Travel blogs and the implications for destination marketing. Journal of Travel Research, 46, 35e45. Park, D. H., & Kim, S. (2008). The effects of consumer knowledge on message processing of electronic word-of-mouth via online consumer reviews. Electronic Commerce Research and Applications, 7(4), 399e410. Park, D. H., & Lee, J. (2008). eWOM overload and its effect on consumer behavioral intention depending on consumer involvement. Electronic Commerce Research and Applications, 7(4), 386e398. Pempek, T. A., Yermolayeva, Y. A., & Calvert, S. L. (2009). College students' social networking experiences on Facebook. Journal of Applied Developmental Psychology, 30(3), 227e238. , F. (2010). Measuring tie-strength in virtual social Petroczi, A., Nepusz, T., & Bazso networks. Connections, 27(2), 39e52. Poyry, E., Parvinen, P., Salo, J., & Blakaj, H. (2012). Hedonic and utilitarian search for electronic word-of-mouth. In 2012 45th Hawaii international conference on system sciences (pp. 1797e1806). Riegner, C. (2007). Word of mouth on the web: the impact of Web 2.0 on consumer purchase decisions. Journal of Advertising Research, 47(4), 436e447. Schramm, L. W., & Porter, W. E. (2010). Men, women, messages, and media: Understanding human communication. He, D., Trans. (2nd ed.). Beijing: China Renmin University Press Scott, J. (2007). Social network analysis: A handbook. Liu, J., Trans. (2nd ed.). Chongqing: Chongqing University Press Sun, T., Youn, S., Wu, G., & Kuntaraporn, M. (2006). Online word-of-mouth (or mouse): an exploration of its antecedents and consequences. Journal of Computer-Mediated Communication, 11(4), 1104e1127. Vilpponen, A., Winter, S., & Sundqvist, S. (2006). Electronic word-of-mouth in online environments: exploring referral network structure and adoption behavior. Journal of Interactive Advertising, 6(2), 63e77. Webster, C. M., & Morrison, P. D. (2004). Network analysis in marketing. Australasian Marketing Journal (AMJ), 12(2), 8e18. Xiang, Z., & Gretzel, U. (2010). Role of social media in online travel information search. Tourism Management, 31, 179e188. Yoo, K. H., Lee, Y., Gretzel, U., & Fesenmaier, D. R. (2009). Trust in travel-related consumer generated media. In W. Hopken, U. Gretzel, & R. Law (Eds.), Information and communication technologies in tourism 2009 (pp. 49e60). New York: Springer.

Qiuju Luo (1968), Ph.D., female, professor, deputy dean of School of Tourism Management, Sun Yat-sen University ( G u a n g z h o u , G u a n g d o n g , C h i n a 510 27 5 ; e - m a i l : [email protected]). Prof. Luo, researching in the area of exhibition, convention, mega event and event tourism, has published more than 40 academic articles in major journal home and abroad. With a series of influential researches, she has earned a good reputation in event industry and tourism industry in China. As the popularization of new media in tourism industry, Prof. Luo devotes herself to conducting innovative research on social media and its influence in tourism domain.

Dixi Zhong (1989), female, master student of School of Tourism Management, Sun Yat-sen University (Guangzhou, Guangdong, China 510275; e-mail: dreamy_cecilia@ foxmail.com). As mobile Internet popularizes, Dixi Zhong develops her interests in e-tourism as well as social media. With an acute insight and rigorous academic attitude, she begins to conduct research on new tourism phenomena. The electronic referrals on social networking sites are one of the topics she's probing into.