Posting-related attributes driving differential engagement behaviors in online travel communities

Telematics and Informatics xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Telematics and Informatics journal homepage: www.elsevier.com/locate/tele

Posting-related attributes driving differential engagement behaviors in online travel communities ⁎

Jiaming Fanga, , Juan Lia, Victor R. Prybutokb a

School of Management and Economics, University of Electronic Science and Technology of China, No. 2006, Xiyuan Ave, West Hi-Tech Zone, Chengdu, Sichuan 611731, China College of Business, University of North Texas, 1155 Union Cir, Denton, TX 76203, USA

b

AR TI CLE I NF O

AB S T R A CT

Keywords: Virtual community participation Online travel community Posting-related attributes Engagement behavior Stimulus-organism-response framework

Users’ engagement behaviors such as likings, sharing and social interactions in online communities are critically important to the viability and the ultimate success of these communities. However, empirical research investigating which posting-related attributes driving these participation behaviors still lags. The purpose of this study is to understand what and how postingrelated attributes drive the engagement behaviors. Using travelogue data from a large traveling knowledge-sharing community, we used econometric models to investigate and compare the antecedents leading to three different kinds of engagement behavior that users exhibit in online communities (i.e., consuming, contributing, and creating). The results reveal that five attributes are associated with these engagement behaviors. These attributes demonstrate the differential effectiveness on the engagement behaviors with different intensities. Our empirical findings provide both theoretical and practical implications for online community operators to build a vibrant and successful online community.

1. Introduction User generated content (UGC) has become a major source for tourists who have a demand for knowledge of their prospective destinations. Online travel communities (OTCs) thus become increasingly prevalent as a credible information network as it provides the tourists with trustworthy reviews and recommendations (Chung & Buhali, 2008). It is believed that the success of an OTC is dependent on the ability to attract OTC users to contribute and keep them engaged (Cheng et al., 2014). User engagement is defined as the intensity of an individual's participation and connection with the organization's offerings initiated by either the user or the organization (Vivek et al., 2012). The extant literature has associated user engagement behavior with enhanced satisfaction, loyalty, trust, positive word of mouth, emotional bonding, commitment, and sales growth (Agag and EI-Masry, 2016; Brodie et al., 2013; Hu et al., 2016). For example, Casaló et al. (2010) demonstrate that users’ intention to participate in an OTC can significantly influence their recommendation intentions. Agag and EI-Masry (2016) report that active participation plays a positive role in users’ intention to purchase and positive WOM toward an OTC. Harrigan et al. (2017) observe that user participation and engagement are positively associated with loyalty in tourism social media. Lee et al. (2014) also claim that the lack of participation in OTCs makes the majority of such communities fail at growing beyond mere existence into powerful forms of social media. Given the importance of the active participation in online communities (OCs), researchers have shown great interest in investigating factors driving users’ engagement behaviors in OCs (e.g., Gharib et al., 2017; Ul Islam & Rahman, 2017; Huang et al.,

⁎

Corresponding author. E-mail addresses: [email protected] (J. Fang), [email protected] (J. Li), [email protected] (V.R. Prybutok).

https://doi.org/10.1016/j.tele.2018.02.008 Received 28 October 2017; Received in revised form 27 February 2018; Accepted 28 February 2018 0736-5853/ © 2018 Elsevier Ltd. All rights reserved.

Please cite this article as: Fang, J., Telematics and Informatics (2018), https://doi.org/10.1016/j.tele.2018.02.008

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

2014). One notable common feature of this stream of studies is that the majority have focused heavily on delineating the effects of psychological factors on engagement behaviors, and little existing research has investigated the antecedents of engagement behaviors in the OC context from the posting-related attributes' perspective. However, the existing studies in the fields of information systems and e-commerce overwhelmingly support that the attributes of an object (such as websites, mobile apps, etc.) with which users interact are prominent stimuli that drive users' engagement behaviors (Fang et al., 2017; Parboteeah et al., 2009; Peters et al., 2016). Drawing on these findings, posting-related attributes also might have a salient impact on users' engagement behaviors in the OTC context. Nevertheless, this issue has been largely ignored by the studies of online engagement. Clearly, there is a strong necessity to extend the extant literature by developing and empirically examining research models to better understand the effect of postingrelated attributes on engagement behaviors. In two recent studies, Luarn et al. (2015) and Schultz (2017) investigate brand post attributes in Facebook brand pages affecting user engagement behavior. However, these studies only consider the contributing engagement behavior (i.e., liking, commenting, and sharing) and focus on the attributes of post itself such as content vividness and interactivity, content categories, and publication timing. There are many different types of engagement behavior in the social media environment. In the case of OCs, liking, sharing, and social interaction are all the manifestations of behavioral engagement. No research to our knowledge has investigated and compared the differences of influencing mechanisms driving the different forms and intensity of engagement behavior within OCs, which impedes the academic and practical gaining a comprehensive understanding on the different types of engagement behaviors that users exhibit in OCs. In addition, the existing studies examine engagement behavior in the context of social media based communities, transactional online communities and social Q&A communities, relatively few studies examine the engagement behavior in the context of OTCs. Different from social media based communities, transactional online communities and social Q&A communities, OTC, as a typical form of knowledge-sharing OCs, is generally based on loose social relationships among members and the members take their own initiative to facilitate free knowledge contribution and dissemination. Moreover, the sharing posts in OTCs are in general much longer and more vivid than other OCs. Considering that user engagement behavior differs in various online settings due to context-dependent nature of this phenomenon (Dovaliene et al., 2015), it is of necessity to examine engagement behaviors across various forms of OCs (Brodie et al., 2011; Brodie et al., 2013; Dovaliene et al., 2015). This paper also responds to Dolan et al.'s (2016) call for further empirical investigations exploring different types of engagement behaviors that users exhibit in OCs and understanding the mechanisms of how the content attributes that trigger users to engage at different levels of intensity. In essence, this study seeks to find the answer to the following questions: (1) What and how do posting-related attributes stimulate the engagement behaviors? (2) Is there any effect difference of these posting-related attributes in affecting different forms and intensities of engagement behavior? To answer these questions, we develop econometric models to explain engagement behaviors by drawing upon recent literature on OC engagement behavior. Specifically, we identify five attributes from three distinct levels (i.e., information features, contributor features and receiver features) as critical external stimuli leading to engagement behaviors. Further, we also investigate and compare the differential effectiveness of these stimuli in affecting low intensity levels of engagement behavior (i.e., viewing), medium levels of engagement behavior (i.e., liking, sharing, and commenting), and intense levels of engagement behavior (i.e., social interaction). We validate the proposed model by using the real-world data collected from a major OTC. The main contributions of this study are threefold. First, our study is the first to empirically identify posting-related attributes that affect users’ engagement behaviors in the OTC context. Second, this study serves as a novel attempt to explore the relative effectiveness of these attributes on the different engagement behaviors. Third, in contrast to prior research of OTC participation behavior that gathers research data through self-reporting questionnaires, our study is one of the rare studies gathering data from the web pages of an OTC. Based on the findings, we propose practical implications that can be used by OTC managers for making better operation strategies and the informed design of OTCs. 2. Theoretical background and hypotheses 2.1. OTC engagement behaviors Engagement behaviors refer to behavioral manifestations of engagement. Users exhibit behavioral manifestations of engagement at different intensities and with different valence within social media platforms (Brodie et al., 2016; Muntinga et al., 2011). The existing research has characterized engagement behaviors on a continuum of low to high activity (Dolan et al., 2016; Muntinga et al., 2011; Wong, 2007). In particular, Brodie et al. (2016) propose that there are three types of positively valenced social media engagement behaviors at different intensities: creating, contributing, and consuming. The majority of OTCs nowadays have integrated social media characteristics into their platforms. OTC users not only can like, comment, share and create posts but also can be able to follow and be followed by other users. As such, OTCs can be considered as a type of social media platforms. The current study focuses on investigating these positively valenced active engagement behavior in OTCs. “Socializing” is epitomized as creating engagement behavior, reflecting a highly active level of social media engagement behavior. Social interaction in OCs can be described as the back-and-forth interactions among OC members. The back-and-forth interactions complement and extend the knowledge of focal sharing posts. As such, creating the new knowledge for OC members. As for contributing engagement behavior, users distribute the existing content and as such they are instrumental in conveying and 2

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

popularizing the social media content. As such, “liking”, “sharing” and “commenting” behaviors in OCs can be regarded as typical contributing engagement behaviors. In addition, consuming engagement behavior refers to the passive consumption of the social media content, in which case, users do not create or contribute to content actively (Brodie et al., 2016). Viewing or post reading can be considered as consuming engagement behavior, the minimum level of engagement behavior (Gummerus et al., 2012). Collectively, the viewing behavior, the liking, sharing, commenting behavior, as well as the social interaction behavior represent three different intensity levels of positively valenced engagement behaviors in OCs. Some studies have investigated the antecedents to engagement behaviors in OTCs. Yoo and Gretzel (2011) observe that altruistic and hedonic benefits are important factors facilitating creating engagement behavior for US travelers. Casaló et al. (2013) show that perceived similarity and reciprocity can affect OTC engagement behavior. Munar and Jaccobsen (2014) report that personal and community-related benefits as well as social capital can be significantly related to members’ engagement in OTCs. Yuan et al. (2016) claim that user innovativeness, subjective knowledge, perceived ease of use and usefulness are significantly associated with content contributions in OTCs. Harrigan et al. (2017) also demonstrate that involvement is an important antecedent leading to engagement behaviors. Taken together, the existing research suggests that perceived benefits such as utilitarian and hedonic value, satisfaction and positive attitude are important factors relating to engagement behavior. 2.2. The S-O-R framework The need of understanding how posting-related attributes affecting engagement behaviors leads us to the S-O-R framework. S-O-R framework claims that different stimuli of the environment influence an individual’s cognitive or affective experiences (organism), which in turn result in his responses (Hu et al., 2016; Benlian, 2015). As a robust and parsimonious framework, the S-O-R theory has been successfully applied to diverse online contexts (e.g., Cui & Lai, 2013; Fang et al., 2017; Ul Islam & Rahman, 2017). The stimulus element in the S-O-R framework is conceptualized as an influence that arouses an organism or individual. Previous study demonstrates that the three building blocks (information, source and receiver) influence an individual's information assessment (Hovland et al., 1953). We thus in this study propose that posting-related attributes representing the characteristics of source, information, and receiver levels might affect users' information assessment. Specifically, this study considers five attributes (length, vividness, source credibility, source degree centrality, and prior comments length) of a sharing post to be the stimuli. The current study posits that these attributes (stimuli) trigger the users' information assessment leading to users' internal cognitive and emotional reactions (i.e., perceived usefulness and perceived enjoyment) (organism), which finally result in the positively valenced engagement behaviors such as liking, commenting and sharing. Consistent with prior research (e.g., Schultz, 2017), we in this study use a black box approach to investigate how posting-related attributes as stimuli affect engagement behaviors as responses (the S-R linkage). We take the users’ cognitive and emotional reactions as the black box that is not explored further. 2.3. Research model and hypotheses 2.3.1. The influence of post content features Prior research shows that the content related factors (i.e., vividness, information) in social networking sites are significantly associated with users’ engagement behaviors (Cvijikj & Michahelles, 2013; De Vries et al., 2012). A travelogue in OTCs can be regarded as a location-oriented or scene-based document consisting of textual and pictorial information, which can create a comprehensive and representative knowledge about destinations (Pang et al., 2011). Therefore, this study focuses on two attributes of the information feature of a posting: the length and the information vividness of a travelogue. In this study, information vividness reflects the richness of visual experience in a travelogue. Coyle and Thorson (2001) show that the increased vividness can positively influence users’ attitude towards websites and behavior consistency. The vivid presentation can also improve the appeal of information, which becomes more emotionally attractive (Nisbett & Ross, 1980). Information vividness have ability of delivering more information and attracting users to engage based on its direct impact on various sense (Xu et al., 2009; Coyle and Thorson, 2001; Luarn et al., 2015). Weathers et al. (2007) also report that information with pictures plays a positive role in lowering the performance uncertainty in online retailer fields. The increased vividness in product presentations also influences users’ perceived enjoyment significantly (Jiang & Benbasat, 2007). In sum, information vividness can drive utilitarian value, hedonic value and positive attitude in OCs. According to Schultz (2017), information vividness can affect engagement behaviors by improving users' innate utilitarian and hedonic value (Yang & Lin, 2014). In the field of OTCs, a travelogue with higher levels of vividness can be perceived to be more attractive and high quality, thus, it will be more likely to appear at the top of the page by the website to attract users to click. In this case, the travelogue will receive relatively high viewing numbers. Furthermore, previous studies observe that information vividness can positively influence the clickthrough rate (Luarn et al., 2015; De Vries et al., 2012), indicating high page view volumes. Moreover, according to reciprocity theory, if a travelogue is considered as useful, interesting and valuable, users may have positive attitude and give positive feedbacks such as the liking and sharing behaviors (Leung and Tanford, 2015; Yang et al., 2017). Hong et al. (2017) report that users click the “like” button on social media when they have favorable attitudes towards the messages. Prior studies also show that information vividness can positively influence the contributing behaviors such as liking, commenting and sharing behavior (Schultz, 2017; Luarn et al., 2015; De Vries et al., 2012). Information vividness may also influence creating engagement behavior by improving the social presence and involvement. Hess et al. (2009) propose that vividness can be considered as a media capability which is associated with social presence. Fortin and Dholakia (2005) suggest that information vividness can be regarded as characteristics of the communication settings and can be 3

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

positively associated with social presence and involvement, which form positive attitudes and behavioral intention. Coyle and Thorson (2001) claim that the increased level of vivid features can positively drive enduring attitude, enhanced feelings of presence and behavior consistency. On the other hand, prior research suggests that social presence is required to enhance and foster online participation and social interaction (Kear et al., 2014; Tu & McIssac, 2002). In order to enhance the level of online social interaction, the degree of social presence must also be increased (Cobb, 2009). In addition, Kujur and Singh (2017) report that vividness can stimulate users to have active interactions with the brand in social networking websites. Watkins (2017) also suggests that a positive attitude can be an indication of likelihood to drive social interactions on social media. Considering all above mentioned, information vividness of a travelogue might positively influence users’ perceived benefits and positive attitudes, which in turn lead to the consuming, contributing and creating engagement behavior. As such, we propose: H1a. There is a positive relationship between information vividness and consuming engagement behavior. H1b. There is a positive relationship between information vividness and contributing engagement behavior. H1c. There is a positive relationship between information vividness and creating engagement behavior. Apart from information vividness, travelogue length is also related to engagement behavior. A travelogue’s textual content, similar to an online product review, can significantly improve the perceived diagnosticity and reduce perceived travelling risk (Xu et al., 2017; Ladhari & Michaud, 2015; Berezina et al., 2016). Previous studies suggest that textual content of online reviews plays an important role in the assessment of products and service on the Internet, and affects users’ attitude towards the certain product and service significantly (Willemsen et al., 2011). Filieri (2015) explores that the depth of information can be represented by the length of a sharing post. Prior research suggests that longer review can provide more information, and the review length can be a measurement of the review’s quality (Lee et al., 2017), which has a significant influence on other users’ perceived enjoyment (Ahn et al., 2007; Shin, 2009; Yan & Huang, 2014). A longer review posting generally includes rich diversity information details about the certain product and has a positive influence on the helpfulness of the review (Mudambi & Schuff, 2010; Filieri, 2015). Ma et al. (2013) propose that the length of reviews reflects the depth and integrity of the information, which can influence the extent to which the others adopt. Moreover, helpfulness is usually regarded as a measure of perceived value in the literature (Pavlou et al., 2007; Pavlou & Fygenson, 2006; Mudambi & Schuff, 2010). In sum, travelogue length can drive utilitarian value, hedonic value, and positive attitude in OCs. In OTCs, travelogues with high quality or perceived to be more helpful will be more likely to be positioned at the top of the page. Thus, the lengthy travelogues are more likely to be clicked and generate high viewings. Moreover, perceived usefulness of information can significantly influence users’ attitude (Watkins, 2017), which in turn inspires users' contributing engagement decision (Hong et al., 2017). Besides, Kulkarni et al. (2013) suggest that length of the blog has an influence on user engagement behaviors, including liking, sharing, commenting and voting. According to previous studies, the pursuit of information is a crucial reason for individuals to like, share or comment postings (Luarn et al., 2015; Lin & Lu, 2011; De Vries et al., 2012). Yan and Huang (2014) also observe that microblogging messages with high information quality is positively associated with users’ sharing intention by increasing users’ perceptions of utilitarian and hedonic benefits. In addition, longer travelogues are usually perceived to be of high quality and full of enjoyment, and can induce positive attitude which can motivate readers to interact with the contributors to explore more knowledge and entertainment (Kang & Shin, 2016; Watkins, 2017). Collectively, the length of a travelogue might positively influence users’ perceived benefits, positive attitudes, which in turn may affect the consuming, contributing and creating engagement behavior. Taken together, the following hypotheses are proposed: H2a. There is a positive relationship between the travelogue’s length and consuming engagement behavior. H2b. There is a positive relationship between the travelogue’s length and contributing engagement behavior. H2c. There is a positive relationship between the travelogue’s length and creating engagement behavior. 2.3.2. The influence of source features Previous studies show that source credibility has a positive influence on engagement behavior in social network settings (Imlawi et al., 2015). In addition, prior studies report that source's centrality is significantly associated with receivers' perceived information helpfulness and their subsequent knowledge contribution behavior (Moghavvemi et al., 2017). Collectively, we consider the expertise level and degree centrality of the source attribute. Source credibility refers to the degree to which the receivers trust the source of information (Ohanian, 1990). The source's expertise is considered as one of the important fundaments of source credibility (Hovland et al., 1953; Hussain et al., 2017). Chang and Wu (2014) suggest that information from experts is regarded as credible. Expertise is an individual’s capability forming within certain fields by the accumulated knowledge to complete the tasks (Mayer et al., 1995). Expertise can motivate online trust strongly (Hu et al., 2016), as experts are often regarded as having sufficient knowledge and experience, having abilities to solve questions and can provide accurate evaluations as well as recommendations (Senecal & Nantel, 2004; Amos et al., 2008). Besides, Willemsen et al. (2012) report that individuals are more likely to identify with the experts and adjust their original attitudes. Thus, information provided from experts can enhance the confidence of an individual to believe the authenticity, reliability and usefulness of a posting, and motivate users’ perceived utilitarian benefits. In the context of OTCs, travelogues can be considered as an important information source about the destination, and can provide a comprehensive and representative knowledge (Pang et al., 2011). Users are more likely to trust travelogues provided by experts to 4

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

avoid information overload and uncertainty. Therefore, users have high likelihood to click links when the posters are experts, and the page views are generated. Similarly, Yan and Huang (2014) indicate that source credibility of microblogging messages can significantly influence users’ reposting intention by the perceived utilitarian and hedonic benefits of the information. Furthermore, individuals who are experts are usually popular in the online communities, users are willing to “like”, comment or share the travelogues posted by them to express their preference and build connections. In addition, user perceived benefits are significantly associated with user interactions in OCs (Kang & Shin, 2016). Individuals who are experts are usually of great popularity in the OCs, and the other users in general enjoy the interactions with them. Hence, it is reasonable to propose the following hypotheses: H3a. There is a positive relationship between the expertise of a contributor and consuming engagement behavior. H3b. There is a positive relationship between the expertise of a contributor and contributing engagement behavior. H3c. There is a positive relationship between the expertise of a contributor and creating engagement behavior. Degree centrality refers to the number of links related to a node, including in-degree centrality and out-degree centrality. Indegree centrality describes the number of links point to the node, while the out-degree centrality is regard as the number of ties point to the others (Freeman, 1978). Degree centrality is important in identifying the most important people at the central position of a certain social media or those that are connected well in social network analysis (Freeman, 1978). According to Wasko and Faraj (2005), the centrality of network reflects structural capital (the linkage between actors), and network centrality is also associated with individual’s degree of involvement (Mossholder et al., 2005). Individuals with high out-degree centrality in a social network increase their opportunities of social learning (Liu et al., 2016) about how to write eye-catching and popular posts. As such, the posts by contributors with high out-degree centrality are more likely to be positioned at the top of pages and regarded as effectiveness in delivering utilitarian and/or hedonic value. Based on reciprocity theory, individuals who perceive benefits from others may provide positive feedbacks including likings, commentings and sharings to reciprocate the benefits they receive. On the other hand, in-degree centrality reflects greater popularity within the network and higher chance to receive more feedbacks from followers (Liu et al., 2016). Thus, posts written by individuals with high out-degree centrality and in-degree centrality are likely to drive the consuming and contributing engagement behaviors (Schultz, 2017; Yang & Lin, 2014). Sabate et al. (2014) report that brand pages with more fans are more likely to be associated with user engagement, including the liking, commenting and sharing behavior. Similarly, user perceived benefits are associated with his interactions in OCs (Kang & Shin, 2016). Moreover, in-degree centrality reflects greater popularity within the network (Liu et al., 2016), and OC members might enjoy discussing with them. Thus, it is reasonable to expect that the degree centrality of a contributor is also related with the creating engagement behavior. Collectively, we propose: H4a. There is a positive relationship between the degree centrality of a contributor and consuming engagement behavior. H4b. There is a positive relationship between the degree centrality of a contributor and contributing engagement behavior. H4c. There is a positive relationship between the degree centrality of a contributor and creating engagement behavior.

2.3.3. The influence of receiver features Previous research suggests that receivers' demographic characteristics and interaction behaviors may significantly influence individual behaviors in OCs (Cheung et al., 2014; Zhu & Zhang, 2010). However, receivers' characteristics such as experiences, knowledge and preference cannot be easily obtained. Thus, this study only considers prior comment length at the receiver level. We propose that prior users' comment length as an interactivity feature might has an impact on an individual's subsequent engagement decisions. Comment length reflects a reviewer's enthusiasm and efforts (Chevealier & Mayzlin, 2006; Chen & Huang, 2013). Comments can be regarded as a knowledge supplement to the travelogue, and longer comments can help to give a more comprehensive understanding of the destinations. In the context of OTCs, travelogues with longer comments in general reflect reviewers’ interests and communication enthusiasm (Chevealier & Mayzlin, 2006; Chen & Huang, 2013). Thus, travelogues with longer comments have higher likelihood of being interesting and of uniqueness, which are more likely to be positioned at the top of the pages, attracting more viewings. Moreover, the longer reviews often provide more product/service details which are positively associated with the perceived helpfulness of the review (Mudambi & Schuff, 2010; Willemsen et al., 2011). The comments especially those high quality comments complement and extend the knowledge of travelogues provided by the contributor, which enhance OTC members’ perceived functional and/or hedonic benefits. Increased perceived value are likely to stimulate consuming, contributing and creating engagement behaviors because of the norm of reciprocity. Further, comments improve the perceived social presence. The comment length to some extent reflects the levels of interactivity among OTC members. The social influence model (Dholakia et al., 2004) claims that a member’s participation decision relies on others’ participation. In other words, OTC members may decide to participate only if they observe high levels of interaction among other members. As such, prior comment length should have an impact on subsequent creating engagement behaviors. Taken together, the following hypotheses are proposed: H5a. There is a positive relationship between the average prior comment length and consuming engagement behavior. H5b. There is a positive relationship between the average prior comment length and contributing engagement. H5c. There is a positive relationship between the average prior comment length and creating engagement behavior. 5

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

2.3.4. Control variables Elapsed time, gender of contributors, heterogeneity of destination and location were included as control variables in our research model. Elapsed time is defined as the interval time from the post time to the date we collected the data. Elapsed time affects the amount of viewing, liking, commenting and sharing. Previous studies also revealed gender differences in online behavior and reported that females are more likely to participate in engagement activities, and enjoy making suggestions than males (Li, 2006). Heterogeneity of destination is mainly reflected in the types of the destinations including natural, heritage and urban type destinations. According to a survey conducted by World Tourism Cities Federation and Ipsos (2016), natural and heritage destinations are more popular than urban destinations for Chinese tourists. Finally, overseas destinations often incur a higher consumption cost than domestic destinations, and the high cost probably decreases the prospective tourists and participation enthusiasm. 3. Research method 3.1. Data collection To test the hypotheses, we collected data from the largest travel social media platform in China, Mafengwo.cn (http://www. mafengwo.cn/), which covers more than 200 countries and regions. We developed a crawler to automatically download web pages of travelogues and users' engagement behaviors from Mafengwo and developed another system to parse web pages into our database. The destinations were randomly chosen from the three different categories, i.e., natural destinations (e.g., Hawaii), heritage destinations (e.g., Angkor Wat), and urban destinations (e.g., Paris). The classification of a specific destination is based on a panel discussion. Members of the independent expert panel are made up of three carefully selected travel experts: Two experts come from a public university and another is from a large travel agency. We retrieved all available information of the each selected destination since the website inception in 2006. In total, we collected 33,801 travelogues for 30 worldwide destinations, including 15 destinations in China, and 15 destinations outside China. 3.2. Operationalization of variables Drawing on the work by Muntinga et al. (2011) and Brodie et al. (2016), this study operationalizes three levels of engagement according to the intensity levels of participation in OTCs. Specifically, the consuming engagement behavior is measured by the number of viewings of a travelogue which represents a minimum level of engagement behavior. Contributing engagement behavior is operationalized by the number of likes, shares and comments on a travelogue, representing a moderate level of positively valenced engagement behavior. Creating engagement behavior is measured by the number of social interactions which indicates a highly active level of social media engagement behavior. As for independent variables, the travelogue’s length (TravelogueLength) is measured by the word count of a travelogue, which represents the content quality of the travelogue. Information vividness (Pictures) is measured by the number of photos in a travelogue. Expertise (Expertise) is treated as a dummy variable, and the top 20 percent of the OC members in terms of ranking scores are considered as experts. The degree centrality of a contributor (DegreeCentrality) is measured by the sum of the number of followers and the number of followees. The average prior comment length (AveCommentLength) is measured by the grand average word count of all comments prior to a specific comment. Specifically, we first computed the average length of all comments prior to each piece of comment of a travelogue, and then calculated the mean of these means to get the grand mean. Table 1 provides the description statistics of the key variables. Table 2 shows the correlation matrix of these variables. Endogeneity concerns are a question that cannot be neglected in econometrics. In this study, some scholars may be concerned about that “expertise” could be an endogenous variable, as it can be influenced by OC members’ engagement behaviors (the reverse causality problem). This study measures “expertise” by the users’ ranking scores in the community. The ranking score of a contributor Table 1 Description statistics. Variable

Mean

SD

Min

Max

PageViews Likes + Shares+Comments Interactions Pictures TravelogueLength Expertise DegreeCentrality AveCommentLength ElapsedTime Male Overseas Natural Heritage Urban

2043.54 70.04 14.79 53.50 4445.47 0.20 340.19 26.12 379.64 0.33 0.35 0.33 0.16 0.51

9384.83 519.35 62.57 76.62 6508.36 0.40 8655.34 19.31 346.46 0.47 0.48 0.47 0.36 0.50

3 0 0 0 0 0 0 0 0 0 0 0 0 0

479,023 33,580 2587 2589 138,955 1 1,554,186 382 2784 1 1 1 1 1

6

7

1 0.809* 0.640* 0.109* 0.291* 0.275* 0.347* 0.252* 0.391* −0.042* 0.280* 0.001 −0.173*

Log(PageViews) Log(Likes + Shares+Comments) Log(Interactions) Log(Pictures) Log(TravelogueLength) Expertise Log(DegreeCentrality) Log(AveCommentLength) Log(ElapsedTime) Male Overseas Heritage Urban

Notes: *p < 0.05.

1

Variables

Table 2 Correlations.

1 0.745* 0.200* 0.280* 0.304* 0.361* 0.245* 0.162* −0.028* 0.193* 0.043* −0.136*

2

1 0.229* 0.293* 0.242* 0.375* 0.313* 0.064* −0.047* 0.112* 0.007 −0.084*

3

1 0.143* 0.067* 0.078* 0.043* −0.088* 0.008 −0.056* 0.045* 0.002

4

1 −0.030* −0.049* 0.206* −0.112* −0.158* 0.171* 0.060* −0.090*

5

1 0.606* −0.011 0.237* 0.153* −0.048* 0.059* −0.017**

6

1 0.045* 0.301* 0.168* −0.035* 0.060* −0.005

7

1 0.055* −0.059* 0.062* −0.009 −0.017*

8

1 0.051* −0.082* −0.037* 0.048*

9

1 −0.072* 0.040* −0.035*

10

1 0.265* −0.423*

11

1 −0.439*

12

1

13

J. Fang et al.

Telematics and Informatics xxx (xxxx) xxx–xxx

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

in Mafengwo.com is in general determined by the quality and quantity of postings she or he shared. The quality of a posting is recognized by the authorities. Travelogue is likely to be regarded as high quality if it has extraordinary and detailed tour guides and suggestions, and can satisfy the utilitarian and hedonic needs of the users. Other users’ engagement behavior (e.g. travelogue viewing, liking, sharing, commenting and interactions) will not affect the ranking score of a contributor. Thus, contributors with high ranking scores in this website are likely to be travel expects who have visited several places, have sufficient domain knowledge and experience, as well as have abilities to solve questions and provide recommendations to others. According to prior studies (e.g., Mayer et al., 1995), experts are considered to have accumulated knowledge in certain field and have the ability to provide accurate evaluations as well as solve questions. Therefore, the ranking level of a contributor can be considered as a manifestation of his/her expertise. Because reverse causality problem does not exist in this case, endogeneity of the expertise variable is not a concern. 4. Data analysis and results 4.1. Model specification The models are specified in Eqs. (1)–(3). We specify the dependent variable and some of independent variables in logarithmic form and we add one to the variables to avoid logarithms of zeroes.

log(PageViews ) = α + β1 • log(Pictures ) + β2 • log(TravelogueLength) + β3 •(Expertise ) + β4 • log(DegreeCentrality ) + β5 • log(AveCommentLength) + γ1 • log(ElapseTime ) + γ2 •(Male) + γ3 •(Overseas ) + γ4 •(Heritage) + γ5 •(Urban) + ε

(1)

log(Likes + Shares + Comments ) = α + β1 • log(Pictures ) + β2 • log(TravelogueLength) + β3 •(Expertise ) + β4 • log(DegreeCentrality ) + β5 • log(AveCommentLength) + γ1 • log(ElapseTime ) + γ2 •(Male ) + γ3 •(Overseas ) + γ4 •(Heritage ) + γ5 •(Urban) + ε

(2)

log(Interactions ) = α + β1 • log(Pictures ) + β2 • log(TravelogueLength) + β3 •(Expertise ) + β4 • log(DegreeCentrality ) + β5 • log(AveCommentLength) + γ1 • log(ElapseTime ) + γ2 •(Male) + γ3 •(Overseas ) + γ4 •(Heritage) + γ5 •(Urban) + ε

(3)

4.2. Hypotheses testing Models were fitted using the seemingly unrelated regression (SUR) estimator. SUR model allows correlated errors across equations, which can improve the efficiency of estimation. Control variables were entered into the models firstly, then all the postingrelated attributes were entered into the models. The results are listed in Table 3. The results reported in Table 3, column (1), (3), and (5) suggest that the set of control variables has good explanatory power. We calculated the variance inflation factor (VIF) measure for all the variables. Because all of the VIF values are well below 10 (the largest VIF value is 5.4), it is reasonable to infer that multicollinearity is not a concern. The overall R-squared value of the full-model in Table 3, Column (2) is 0.45, which indicates that posting-related attributes play an important role in affecting the consuming engagement behavior in the OTC context. The results show that information vividness, the word count of a travelogue, source credibility, degree centrality of a contributor, and average length of prior comments are positively related to the consuming engagement behavior, supporting H1a, H2a, H3a, H4a, and H5a. The control model of the contributing engagement behavior has an Rsquared value of 0.07 (Table 3, column 3), and the R-squared value of full model is 0.32 (Table 3, column 4). The results indicate that the posting-related attributes exert their significant impacts on the contributing engagement behavior, supporting H1b, H2b, H3b, H4b, and H5b. As for creating engagement behavior, the R-squared value of the control model is 0.02 (Table 3, column 5), while the R-squared value of the full model is 0.34 (Table 3, column 6), which suggests that the posting-related attributes are able to significantly affect creating engagement behavior. Thus, H1c, H2c, H3c, H4c, and H5c are supposed. To identify the relative importance of the independent variables, we conducted a dominance analysis that is based on the contribution to the overall model fit statistic (Grömping, 2007). The results are reported in Table 4. As shown in Table 4, degree of centrality is the most important factor for all the three engagement behaviors. The results also suggest that travelogue length is more important than picture numbers for all the consuming, contributing and creating engagement behaviors. Finally, average prior comment length is especially important for the creating engagement behavior. 4.3. Coefficient difference testing In order to examine the differences of the posting-related attributes in affecting engagement behaviors at differential intensities, we conducted a series of regression coefficient comparison tests, and the results are reported in Table 5. The results show that posting-related attributes have differential effects on stimulating the three engagement behaviors. Postingrelated attributes exhibit the maximum influence on the contributing engagement behaviors, the moderate effect on creating 8

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

Table 3 Model estimation results. Variable

Consuming Engagement Behavior (1)

Consuming Engagement Behavior (2)

0.40*** (0.00) −0.04*** (0.00) 0.29*** (0.01) −0.15*** (0.01) −0.11*** (0.01) 1.91*** (0.01) 0.27

0.08*** (0.00) 0.12*** (0.00) 0.11*** (0.01) 0.14*** (0.00) 0.18*** (0.01) 0.35*** (0.00) −0.04*** (0.00) 0.26*** (0.00) −0.18*** (0.01) −0.11*** (0.00) 1.10*** (0.01) 0.45

Log(Pictures) Log(TravelogueLength) Expertise Log(DegreeCentrality) Log(AveCommentLength) Log(ElapsedTime) Male Overseas Heritage Urban Constant R-squared Notes: *p < 0.05,

**

p < 0.01,

Contributing Engagement Behavior (3)

Contributing Engagement Behavior (4)

0.51*** (0.01) −0.08 (0.02) 0.52*** (0.02) −0.13*** (0.02) −0.23*** (0.02) 1.54*** (0.04) 0.07

0.36*** (0.01) 0.30*** (0.01) 0.49*** (0.02) 0.50*** (0.01) 0.60*** (0.02) 0.27*** (0.01) −0.13*** (0.01) 0.47*** (0.01) −0.28*** (0.02) −0.23*** (0.01) −0.99*** (0.05) 0.32

Creating Engagement Behavior (5)

Creating Engagement Beha vior (6)

0.08*** (0.01) −0.05*** (0.01) 0.11*** (0.01) −0.07*** (0.01) −0.07*** (0.01) 0.54*** (0.61) 0.02

0.15*** (0.00) 0.12*** (0.00) 0.06*** (0.01) 0.28*** (0.00) 0.34*** (0.02) −0.02*** (0.01) −0.07*** (0.01) 0.09*** (0.01) −0.13*** (0.01) −0.08*** (0.01) −0.66*** (0.02) 0.34

***

p < 0.001. Cluster-robust standard errors in parentheses.

Table 4 Dominance analysis results. Variable

Consuming Engagement Behavior

Contributing Engagement Behavior

Creating Engagement Behavior

Log(Pictures) Log(TravelogueLength) Expertise Log(DegreeCentrality) Log(AveCommentLength) Log(ElapsedTime) Gender DestinationLocation DestinationType

0.02 0.15 0.08 0.15 0.09 0.29 0.00 0.15 0.06

0.09 0.18 0.15 0.26 0.13 0.05 0.01 0.09 0.04

0.11 0.19 0.09 0.33 0.22 0.01 0.01 0.02 0.02

Table 5 The difference test of regression coefficients. Level

Consuming Engagement Behavior (1)

Contributing Engagement Behavior (2)

Creating Engagement Behavior (3)

Difference (1)−(3)

Difference (2)−(3)

Difference (1)−(2)

Information

0.20***

0.66***

0.27***

Source

0.25***

0.99***

0.34***

Receiver

0.18***

0.60***

0.34***

−0.07*** (0.00) −0.09*** (0.01) −0.16*** (0.01)

0.39*** (0.01) 0.65*** (0.01) 0.26*** (0.01)

−0.46*** (0.01) −0.74*** (0.01) −0.42*** (0.01)

Notes: *p < 0.05,

**

p < 0.01,

***

p < 0.001. Standard errors in parentheses.

engagement behaviors, and the minimum impact on consuming engagement behaviors. The source-level attributes consisting of expertise and the degree centrality of a contributor consistently exhibit greatest influence on the three engagement behaviors. 4.4. Robustness test We further corroborated our findings by conducting three robustness tests. First, one may be concerned about the dummy variable specification of the expertise variable. We thus used the continuous ranking scores to measure the expertise, and the results are 9

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

Table 6 Robustness tests. Variable

Consuming Engagement Behavior (1)

Contributing Engagement Behavior (2)

Creating Engagement Behavior (3)

Consuming Engagement Behavior (4)

Contributing Engagement Behavior (5)

Creating Engagement Behavior (6)

Log(Pictures)

0.07*** (0.00) 0.12*** (0.00) 0.01*** (0.00) 0.12*** (0.00) 0.19*** (0.01) 0.34*** (0.00) −0.05*** (0.00) 0.26*** (0.01) −0.19*** (0.01) −0.11*** (0.00) 1.09*** (0.01) 0.45

0.34*** (0.01) 0.30*** (0.00) 0.04*** (0.00) 0.40*** (0.01) 0.61*** (0.02) 0.24*** (0.01) −0.15*** (0.01) 0.47*** (0.01) −0.31*** (0.02) −0.23*** (0.01) −1.03*** (0.05) 0.33

0.14*** (0.00) 0.12*** (0.00) 0.01*** (0.00) 0.23*** (0.00) 0.35*** (0.01) −0.37*** (0.01) −0.08*** (0.01) 0.08*** (0.01) −0.13*** (0.01) −0.07*** (0.01) −0.65*** (0.02) 0.34

0.08*** (0.00) 0.09*** (0.00) 0.10*** (0.01) 0.11*** (0.00) 0.16*** (0.00) 0.37*** (0.00) −0.03*** (0.00) 0.25*** (0.01) −0.16*** (0.01) −0.09*** (0.00) 1.20*** (0.01) 0.44

0.38*** (0.00) 0.24*** (0.01) 0.48*** (0.02) 0.46*** (0.01) 0.60*** (0.01) 0.23*** (0.01) −0.11*** (0.01) 0.42*** (0.02) −0.22*** (0.02) −0.19*** (0.01) −0.68*** (0.04) 0.29

0.14*** (0.00) 0.08*** (0.00) 0.05*** (0.01) 0.25*** (0.00) 0.33*** (0.00) −0.02*** (0.00) −0.05*** (0.00) 0.07*** (0.01) −0.09*** (0.01) −0.05*** (0.01) −0.53*** (0.02) 0.32

Log(TravelogueLength) Expertise Log(DegreeofCentrality) Log(AveCommentLength) Log(ElapsedTime) Male Overseas Heritage Urban Constant R-squared Notes: *p < 0.05,

**

p < 0.01,

***

p < 0.001. Cluster-robust standard errors in parentheses.

reported in Table 6, column (1), (2) and (3). The results are consistent with those in Table 3. Second, instead of using OLS, we re-estimated the model using feasible generalized least squares (FGLS) method. FGLS is preferred over OLS in obtaining more efficient point estimates under heteroskedasticity. As shown in Table 6, column (4), (5), and (6), the model estimates are generally consistent with those in Table 3.

Table 7 Robustness tests using travelogue subsamples. Variable

Consuming Engagement Behavior (1)

Contributing Engagement Behavior (2)

Creating Engagement Behavior (3)

Consuming Engagement Behavior (4)

Contributing Engagement Behavior (5)

Creating Engagement Behavior (6)

Log(Pictures)

0.18*** (0.00) 0.08*** (0.00) 0.13*** (0.01) 0.10*** (0.00) 0.14*** (0.01) 0.36*** (0.01) −0.04*** (0.00)

0.63*** (0.02) 0.20*** (0.01) 0.55*** (0.02) 0.38*** (0.01) 0.50*** (0.02) 0.33*** (0.02) −0.12*** (0.02)

0.22*** (0.01) 0.09*** (0.00) 0.06*** (0.01) 0.26*** (0.00) 0.29*** (0.01) −0.00*** (0.01) −0.06*** (0.01)

0.02*** (0.01) −0.04*** (0.01) 1.07*** (0.02) 0.39

−0.00*** (0.03) −0.07*** (0.02) −1.03*** (0.06) 0.27

−0.02*** (0.01) 0.01*** (0.01) −0.73*** (0.02) 0.32

0.08*** (0.00) 0.11*** (0.00) 0.11*** (0.01) 0.13*** (0.01) 0.19*** (0.01) 0.34*** (0.01) −0.04*** (0.01) 0.26*** (0.01) −0.20*** (0.01) −0.12*** (0.01) 1.12*** (0.02) 0.44

0.37*** (0.02) 0.28*** (0.01) 0.50*** (0.03) 0.51*** (0.02) 0.63*** (0.03) 0.25*** (0.02) −0.13*** (0.02) 0.47*** (0.02) −0.29*** (0.03) −0.25*** (0.02) 0.94*** (0.07) 0.32

0.15*** (0.01) 0.11*** (0.00) 0.06*** (0.01) 0.28*** (0.01) 0.34*** (0.01) −0.04*** (0.01) −0.08*** (0.01) 0.08*** (0.01) −0.13*** (0.01) −0.08*** (0.01) −0.63*** (0.03) 0.33

Log(TravelogueLength) Expertise Log(DegreeofCentrality) Log(AveCommentLength) Log(ElapsedTime) Male Overseas Heritage Urban Constant R-squared Notes: *p < 0.05,

**

p < 0.01,

***

p < 0.001. Cluster-robust standard errors in parentheses.

10

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

Finally, we examined whether our finding is robust across the different sample sizes. We examined the results using travelogues associated with destinations in China, and the results are reported in Table 7, column (1), (2) and (3). Further, we checked the results using 15,000 travelogues selected at random from the complete dataset, and the estimation results are reported in Table 7, column (4), (5) and (6). Both the results remain consistent with the coefficients reported in Table 3. 5. Discussion and implication Users' engagement behaviors greatly determine the survival and success of an OTC. Although product attributes have long been recognized in information systems research as one of the most relevant determinants of user engagement, in prior studies investigating engagement behavior in the OTC context, posting-related attributes have been largely overlooked. The existing research on online user engagement behaviors in general focuses on investigating the influences of user psychological factors. The literature still lacks insights on which and how posting-related attributes affect engagement behaviors. Further, users in OCs actually exhibit different types of engagement behavior, and there is no published research that identify and compare the antecedents driving these engagement behaviors. Taking into account the above mentioned reasons, we in this study seek to understand what and how the posting-related attributes affect users' differential engagement behaviors. Drawing upon the S-O-R framework, we identified five posting-related attributes that serve as environmental prompts (stimulus) which might affect users' cognitive and affective experiences (organism) that in turn influence OTC users' engagement behaviors (responses). We validated the proposed models by using publicly available information about users in combination with the sharing posts they wrote in a large OTC. The empirical results of this study lead to some important and interesting findings that had not been revealed by previous studies. Our results reveal that posting-related attributes from information-, contributor-, and receiver-level are significantly associated with engagement behaviors and exhibit differential effectiveness on the different engagement behaviors. More precisely, the results show that information-level attributes, including pictures and travelogue length, contributor-level attributes, consisting of expertise and degree centrality, as well as receiver-level attribute (i.e., average prior comment length) are important factors driving engagement behaviors in OTCs. The results consistently show that degree centrality has the largest impacts on the three different engagement behaviors which stresses the importance of the social relationships in OTCs. The results also suggest that travelogue length and prior comment length are important catalysts driving engagement behaviors. Besides, picture numbers in travelogues and the contributor's expertise are both positively associated with the engagement behaviors. Moreover, the results report differential effectiveness of the posting-related attributes on the three different engagement behaviors. Generally speaking, compared with consuming and creating engagement behaviors, posting-related attributes have relatively larger influences on contributing engagement behavior. For the consuming and contributing engagement behaviors, source-level attributes have the largest impact, followed by information- and receiver-level attributes. While for the creating engagement behavior, source-level attributes have the largest impact, followed by receiver- and information-level attributes. Specifically, degree centrality and travelogue length equally contribute to the consuming engagement behavior, followed by average prior comment length. By contrast, for contributing engagement behavior, degree centrality has the largest impact, followed by travelogue length and expertise. While, for creating engagement behavior, average prior comment length has a relatively larger influence than travelogue length and expertise. 5.1. Theoretical implications This study contributes to the existing OC engagement and tourism research literature by providing empirical evidence to support the impact of the posting-related attributes on OTC members' engagement behaviors with different intensity levels. The study makes some unique and important contributions to the extant literature. First, this study contributes to the OC engagement literature by serving as the first attempt to empirically explore the effects of posting-related attributes on users’ engagement behaviors in the OTC context. Our research is among the first to identify and examine the impact of posting-related attributes on engagement behaviors. The results suggest that attributes from the information-, source-, and receiver-level significantly influence engagement behaviors. In particular, source-level attributes including expertise and degree centrality consistently exert the largest impact across the three engagement behaviors. These findings enable OC operators to gain a better understanding of users' engagement behaviors and hence to manage user-generated content (UGC) more efficiently. Although the factors we provide in this study may be not comprehensive, compared to previous studies, this study still adds useful knowledge to the understanding of the engagement formation mechanism in the OTC context. This study provides a stimulus and guidance for further deep inquiries into the role of posting-related attributes in OTCs. Second, this study also enriches our knowledge about user engagement behavior with different intensity levels in OTCs and relative effectiveness of the posting-related attributes on these engagement behaviors. Specifically, we identify three engagement behaviors in OTCs, including consuming, contributing and creating engagement behavior. Our empirical results show that postingrelated attributes differentially affect the different intensity of engagement behaviors. It is the first study that comprehensively captures positively valenced engagement behavior within OTC contexts and investigates and compares the posting-related attributes driving the different engagement behaviors, which helps us to gain a better understanding on the formation of engagement behaviors in the OC context. Finally, this research contributes to the extant online engagement literature by extending the applicability of the S-O-R framework to the OC engagement context. The S-O-R framework provides its theoretical justification for the inclusion of the distinct postingrelated attributes as environmental stimuli. The S-O-R model provides a parsimonious and structured guidance to explain how 11

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

posting-related attributes contribute to users’ engagement behaviors. 5.2. Practical implications With the rapid development of social media, traditional marketing strategies are not efficient enough in current fierce competitive environment, in this situation, engaged users in OCs become an asset which attracts a large quantity of organizations to pay attention to (Chan et al., 2014). Specifically, this study provides practical value to OTC managers who are trying to find effective strategies in driving user engagement in the OC environment. This study also takes engagement behaviors differences into consideration, which can provide more target and differentiated strategies for practitioners. Our results suggest that the consuming engagement behavior, contributing engagement behavior and creating engagement behavior can be addressed through various posting-related attributes. First, the results of this study suggest that photo numbers and travelogue length are significantly related to the three engagement behaviors. Photo numbers and the length of a travelogue can be a reflection of information richness, which is related to individuals’ perceived utilitarian and hedonic benefits. Thus, in order to encourage the contributors to post high quality travelogues, OTC managers can consider providing some travelogue writing paradigms for travelers to refer to and offering some monetary and non-monetary incentives to encourage users to write high quality travelogues. Second, this study also highlights the importance of source credibility and receiver attributes. The result suggests that average prior comment length and contributors who are perceived to be expert or have high degree centrality can positively influence the consuming engagement behavior, contributing engagement behavior and creating engagement behavior. Thus, some measures by leveraging these attributes could be taken to facilitate users' engagement behaviors. For example, OTC operators can provide incentives to encourage high-ranking users to write travelogues and high quality replies. Community operators can also enact some rules and regulations to increase contributors’ motivation to interact back and forth with other OTC members. Moreover, operators should also refine the website design to increase the convenience for the members to reply the sharing posts. Finally, the findings indicate that posting-related features differentially affect the three engagement behaviors. Therefore, considering different engagement objectives, OTC operators can utilize different posting-related attributes and attach different importance to them so as to address differentiated user needs. For example, when the operators want to increase the liking, commenting and sharing behavior, they should especially focus on gaining cooperation and support from the contributors with high degree centrality and increasing the average comment length. OTC operators thus should provide incentives to encourage high degree centrality members to engage in high quality content creation and sharing and inspire social interactions between contributors and other OTC members. 6. Limitations and future research Despite the significant theoretical and practical implications, this study has several limitations that need to be addressed in future research. Firstly, this study adopts cross-sectional data in investigating the factors affecting the engagement behaviors, which prevents the study from assessing causality. Secondly, we only used a single receiver attribute (i.e., average prior comment length) when we assess the impact of posting-related attributes on engagement behaviors. However, average length might not be able to truly epitomize the nature and capture the influence of receivers. Future research could overcome this issue by collecting more data from the receiver's level. Finally, our research is based on archival data and is data-driven in nature, thus it cannot provide convincing supports for the underlying mechanisms presented in the hypotheses. Further survey research is needed to provide additional insight into the underlying psychological mechanisms accounting for our findings. Acknowledgements The work described in this paper was partially supported by the grant from the National Natural Science Foundation of China (No.71571029) and Chengdu Foundation for Philosophy and Social Science (No. 2017P25). References Agag, G., EI-Masry, A., 2016. Understanding consumer intention to participate in online travel community and effects on consumer intention to purchase travel online and WOM: an integration of innovation diffusion theory and TAM with trust. Comput. Hum. Behav. 60, 97–111. Ahn, T., Ryu, S., Lngoo, H., 2007. The impact of web quality and playfulness on user acceptance of online retailing. Inf. Manage. 44 (3), 263–275. Amos, C., Holmes, G., Strutton, D., 2008. Exploring the relationship between celebrity endorser effects and advertising effectiveness. Int. J. Advert. 27 (2), 209–234. Benlian, A., 2015. Web personalization cues and their differential effects on user assessments of website value. J. Manage. Inform. Syst. 32 (1), 225–260. Berezina, K., Bilgihan, A., Cobanoglu, C., Okumus, F., 2016. Understanding satisfied and dissatisfied hotel customers: text mining of online hotel reviews. J. Hosp. Market. Manage. 25 (1), 1–24. Brodie, R.J., Hollebeek, L., Jurié, B., Ilié, A., 2011. Customer engagement conceptual domain, fundamental propositions, and implications for research. J. Serv. Res. 14 (3), 252–271. Brodie, R.J., Ilic, A., Juric, B., Hollebeek, L., 2013. Consumer engagement in a virtual brand community: an exploratory analysis. J. Bus. Res. 66 (1), 105–114. Brodie, R.J., Hollebeek, L., Conduit, J., 2016. Customer Engagement: Contemporary Issues and Challenges. Routledge, London. Casaló, L., Flavián, C., Guinalíu, M., 2010. Determinants of the intention to participate in firm-hosted online travel communities and effects on consumer behavioral intentions. Tour. Manage. 31 (6), 898–911. Casaló, L.V., Flavián, C., Guinalíu, M., 2013. New members' integration: key factor of success in online travel communities. J. Bus. Res. 66 (6), 706–710. Chan, T.K.H., Zheng, X., Cheung, C.M.K., Lee, M.K.O., Lee, Z.W.Y., 2014. Antecedents and consequences of customer engagement in online brand communities. J.

12

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

Market. Anal. 2 (2), 81–97. Chang, H.H., Wu, L.H., 2014. An examination of negative e-WOM adoption: brand commitment as a moderator. Decis. Support Syst. 59 (1), 206–218. Chen, H.N., Huang, C.Y., 2013. An investigation into online reviewers’ behavior. Eur. J. Mark. 47 (10), 1758–1773. Cheng, J., Danescu-Niculescu-Mizil, C., Leskovec, J., 2014. ‘How community feedback shapes user behavior’, available at: http://cs.stanford.edu/people/jure/pubs/ disqus-icwsm14.pdf. Cheung, C.M.K., Xiao, B.S., Liu, I.L.B., 2014. Do actions speak louder than voices? the signaling role of social information cues in influencing consumer purchase decisions. Decis. Support Syst. 65 (1), 50–58. Chevalier, J.A., Mayzlin, D., 2006. The effect of word of mouth on sales: Online book reviews. J. Market. Res. 43 (3), 345–354. Chung, J.Y., Buhalis, D., 2008. Information needs in online social networks. Inform. Technol. Tourism. 10 (4), 267–281. Cobb, S.C., 2009. Social presence and online learning: a current view from a research perspective. J. Interact. Online Lear. 8 (3), 241–254. Coyle, J.R., Thorson, E., 2001. The effects of progressive levels of interactivity and vividness in web marketing sites. J. Advertising. 30 (3), 65–77. Cui, X., Lai, V., 2013. E-loyalty to online auction websites: a stimulus-organism-response model. 2013 Pacific Asia Conference on Information Systems (PACIS), Jeju Island, Korea. Cvijikj, I.P., Michahelles, F., 2013. Online engagement factors on Facebook brand pages. Soc. Network Anal. Mining. 3 (4), 843–861. De Vries, L., Gensler, S., Leeflang, P.S., 2012. Popularity of brand posts on brand fan pages: an investigation of the effects of social media marketing. J. Interact. Mark. 26 (2), 83–91. Dholakia, U.M., Bagozzi, R.P., Pearo, L.K., 2004. A social influence model of consumer participation in network- and small-group-based virtual communities. Int. J. Res. Mark. 21 (3), 241–263. Dolan, R., Conduit, J., Faby, J., 2016. Social media engagement: A construct of positively and negatively valenced engagement behaviors. In: Brodie, R.J., Hollebeek, L., Conduit, J. (Eds.), Customer Engagement: Contemporary Issues and Challenges. Routledge, London, pp. 102–123. Dovaliene, A., Masiulyte, A., Piligrimiene, Z., 2015. The relations between customer engagement, perceived value and satisfaction: the case of mobile applications. Proc. – Soc.Behav. Sci. 213, 659–664. Fang, J., Zhao, Z., Wen, C., Wang, R., 2017. Design and performance attributes driving mobile travel application engagement. Int. J. Inf. Manage. 37 (4), 269–283. Filieri, R., 2015. What makes online reviews helpful? A diagnosticity-adoption framework to explain informational and normative influences in e-WOM. J. Bus. Res. 68 (6), 1261–1270. Fortin, D.R., Dholakia, R.R., 2005. Interactivity and vividness effects on social presence and involvement with a web-based advertisement. J. Bus. Res. 58 (3), 387–396. Freeman, C.L., 1978. Centrality in social networks conceptual clarification. Soc. Networks. 1 (3), 215–239. Gharib, R.K., Philpott, E., Duan, Y., 2017. Factors affecting active participation in B2B online communities: an empirical investigation. Inf. Manage. 54 (4), 516–530. Grömping, U., 2007. Estimators of relative importance in linear regression based on variance decomposition. Am. Stat. 61 (2), 139–147. Gummerus, J., Liljander, V., Weman, E., Pihlström, M., 2012. Customer engagement in a Facebook brand community. Manage. Res. Rev. 35 (9), 857–877. Harrigan, P., Evers, U., Miles, M., Daly, T., 2017. Customer engagement with tourism social media brands. Tour. Manage. 59, 597–609. Hess, T.J., Fuller, M.A., Campbell, D.E., 2009. Designing interfaces with social presence: using vividness and extraversion to create social recommendation agents. J. Assoc Inf. Syst. 10 (12), 889–919. Hong, C., Chen, Z., Li, C., 2017. “Liking” and being “liked”: how are personality traits and demographics associated with giving and receiving “likes” on Facebook? Comput. Hum. Behav. 68, 292–299. Hovland, I., Janis, L., Kelley, H., 1953. Communications and Persuasion: Psychological Studies of Opinion Change. Yale University Press, New Haven. Hu, M., Zhang, M.L., Luo, N., 2016. Understanding participation on video sharing communities: the role of self-construal and community interactivity. Comput. Hum. Behav. 62, 105–115. Huang, Q., Davison, R., Liu, H.F., 2014. An exploratory study of buyers’ participation intentions in reputation systems: the relationship quality perspective. Inf. Manage. 51 (8), 952–963. Hussain, S., Ahmed, W., Jafar, R.M.S., Rabnawaz, A., Yang, J., 2017. eWOM source credibility, perceived risk and food product customer's information adoption. Comput. Hum. Behav. 66, 96–102. Imlawi, J., Gregg, D., Karimi, J., 2015. Student engagement in course-based social networks: the impact of instructor credibility and use of communication. Comput. Educ. 88, 84–96. Jiang, Z., Benbasat, I., 2007. Investigating the influence of the functional mechanisms of online product presentations. Inf. Sys. Res. 18 (4), 454–472. Kang, M., Shin, D.H., 2016. The effect of customers’ perceived benefits on virtual brand community loyalty. Online Inf. Rev. 40 (3), 298–315. Kear, K., Chetwynd, F., Jefferis, H., 2014. Social presence in online learning communities: the role of personal profiles. Res. Lear. Technol. 22. http://dx.doi.org/10. 3403/rlt.v22.19710. Kujur, F., Singh, S., 2017. Engaging customers through online participation in social networking sites. Asia Pacific Manage. Rev. 22 (1), 16–24. Kulkarni, A.V., Joseph, S., Raman, R., Bharathi, V., Goswami, A., 2013. Blog content and user engagement – an insight using statistical analysis. Int. J. Engine. Technol. 5 (3), 2719–2733. Ladhari, R., Michaud, M., 2015. eWOM effects on hotel booking intentions, attitudes, trust, and website perceptions. Int. J. Hospitality Manage. 46 (3), 36–45. Lee, H., Reid, E., Kim, W., 2014. Understanding knowledge sharing in online travel communities antecedents and the moderating effects of interaction modes. J. Hospitality Tour. Res. 38 (2), 222–242. Lee, S.G., Trimi, S., Yang, C.G., 2017. Perceived usefulness factors of online reviews: a study of Amazon.com. J. Comput. Inf. Syst. 3, 1–9. Leung, X.Y., Tanford, S., 2015. What drives Facebook fans to ‘like’ hotel pages: a comparison of three competing models. J. Hospitality Market. Manage. 25 (3), 314–345. Li, Q., 2006. Computer-mediated communication: a meta-analysis of male and female attitudes and behaviors. Int. J. E-Learn. 5, 525–570. Lin, K.-Y., Lu, H.-P., 2011. Intention to continue using Facebook fan pages from the perspective of social capital theory. Cyberpsychol. Behav. Soc. Network 14 (10), 565–570. Liu, L., Cheung, C.M.K., Lee, M.K.O., 2016. An empirical investigation of information sharing behavior on social commerce sites. Int. J. Inf. Manage. 36 (5), 686–699. Luarn, P., Lin, Y.F., Chiu, Y.P., 2015. Influence of Facebook brand-page posts on online engagement. Online Inf. Rev. 39 (4), 505–519. Ma, X., Khansa, L., Yun, D., Kim, S.S., 2013. Impact of prior reviews on the subsequent review process in reputation system. J. Manage. Inf. Syst. 30 (3), 279–310. Mayer, R.C., Davis, J.H., Schoorman, F.D., 1995. An integrative model of organizational trust. Acad. Manage. Rev. 20 (3), 709–734. Mossholder, K.W., Settoon, R.P., Henagan, S.C., 2005. A relational perspective on turnover: examining structural, attitudinal, and behavioral predictors. Acad. Manage. J. 48 (4), 607–618. Mudambi, S.M., Schuff, D., 2010. What makes a helpful online review? A study of customer reviews on Amazon.com. MIS Quart. 34 (1), 185–200. Munar, A.M., Jacobsen, J.K.S., 2014. Motivations for sharing tourism experiences through social media. Tour. Manage. 43 (8), 46–54. Muntinga, D.G., Moorman, M., Smit, E.G., 2011. Introducing cobras. Int. J. Advertising. 30 (1), 13–46. Moghavvemi, S., Sharabati, M., Paramanathan, T., Rahin, N., 2017. The impact of perceived enjoyment, perceived reciprocal benefits and knowledge power on students ' knowledge sharing through Facebook. Int. J. Manage. Educ. 15 (1), 1–12. Nisbett, R., Ross, L., 1980. Assigning weights to data: The “vividness criterion”. In: Nisbett, R., Ross, L. (Eds.), Human Inference: Strategies and Shortcomings of Social Judgment. Prentice-Hall Inc, Englewood Cliffs, pp. 43–61. Ohanian, R., 1990. Construction and validation of a scale to measure celebrity endorsers' perceived expertise, trustworthiness, and attractiveness. J. Advertising. 19 (3), 39–52. Pang, Y., Hao, Q., Yuan, Y., Hu, T., Cai, R., Zhang, L., 2011. Summarizing tourist destinations by mining user-generated travelogues and photos. Comput. Vis. Image Und. 115 (3), 352–363. Parboteeah, D.V., Valacich, J.S., Wells, J.D., 2009. The influence of website characteristics on a consumer’s urge to buy impulsively. Inf. Syst. Res. 20 (1), 60–78. Pavlou, P., Fygenson, M., 2006. Understanding and predicting electronic commerce adoption: an extension of the theory of planned behavior. MIS Quart. 30 (1),

13

Telematics and Informatics xxx (xxxx) xxx–xxx

J. Fang et al.

115–143. Pavlou, P., Liang, H., Xue, Y., 2007. Uncertainty and mitigating uncertainty in online exchange relationships: a principal– agent perspective. MIS Quart. 31 (1), 105–131. Peters, T., Işık, Öykü, Tona, O., Popovič, A., 2016. How system quality influences mobile bi use: the mediating role of engagement. Int. J. Inf. Manage. 36 (5), 773–783. Sabate, F., Berbegal-Mirabent, J., Cañabate, A., Lebherz, P.R., 2014. Factors influencing popularity of branded content in Facebook fan pages. Eur. Manage. J. 32 (6), 1001–1011. Schultz, C.D., 2017. Proposing to your fans: which brand post characteristics drive consumer engagement activities on social media brand pages? Electron. Commer. R. A. 26, 23–34. Senecal, S., Nantel, J., 2004. The influence of online product recommendations on consumers' online choices. J. Retail. 80 (2), 159–169. Shin, D.H., 2009. An empirical investigation of a modified technology acceptance model of IPTV. Behav. Inform. Technol. 28 (4), 361–372. Tu, C.H., McIsaac, M., 2002. The relationship of social presence and interaction in online classes. Am. J. Distance Educ. 16 (3), 131–150. Ul Islam, J., Rahma, Z., 2017. The impact of online brand community characteristics on customer engagement: an application of stimulus-organism-response paradigm. Telematics Inform. 34 (4), 96–109. Vivek, S.D., Beatty, S.E., Morgan, R.M., 2012. Customer engagement: exploring customer relationships beyond purchase. J. Market. Theory Pract. 20 (2), 122–146. Watkins, B., 2017. Experimenting with dialogue on Twitter: an examination of the influence of the dialogic principles on engagement, interaction, and attitude. Public Relat. Rev. 43 (1), 163–171. Wasko, M.M., Faraj, S., 2005. Why should I share? Examining social capital and knowledge contribution in electronic networks of practice. MIS Quart. 29 (1), 35–57. Weathers, D., Sharma, S., Wood, S.L., 2007. Effects of online communication practices on consumer perceptions of performance uncertainty for search and experience goods. J. Retail. 83 (4), 393–401. Willemsen, L.M., Neijens, P.C., Bronner, F., Ridder, J.A.D., 2011. “Highly recommended!” the content characteristics and perceived usefulness of online consumer reviews. J. Comput-mediat. Commun. 17 (1), 19–38. Willemsen, L., Neijens, P., Bronner, F., 2012. The ironic effect of source identification on the perceived credibility of online product reviewers. J. Comput-mediat. Comm. 18 (1), 16–31. Wong, A., 2007. ‘Social media and marketing: Evolution or revolution?’, available at: http://www.slideshare.net/wah17/social-media-35304. World Tourism Cities Federation and Ipsos. 2016. Market research report on Chinese outbound tourist (city) consumption (2015-2016). Xu, H., Oh, L.B., Teo, H.H., 2009. Perceived effectiveness of text vs. multimedia location-based advertising messaging. Int. J. Mob. Commun. 7 (2), 154–177. Xu, X., Wang, X.Q., Li, Y.B., Haghighi, M., 2017. Business intelligence in online customer textual reviews: Understanding consumer perceptions and influential factors. Int. J. Inf. Manage. 37 (6), 673–683. Yan, W., Huang, J., 2014. Microblogging reposting mechanism: an information adoption perspective. J. Tsinghua Uni. (Sci. Techn) 19 (5), 531–542. Yang, H.L., Lin, C.L., 2014. Why do people stick to Facebook web site? a value theory-based view. Inf. Technol. People. 27 (1), 21–37. Yang, X., Li, G., Huang, S.S., 2017. Perceived online community support, member relations, and commitment: differences between posters and lurkers. Inf. Manage. 54 (2), 154–165. Yoo, K.H., Gretzel, U., 2011. Influence of personality on travel-related consumer-generated media creation. Comput. Hum. Behav. 27 (2), 609–621. Yuan, D., Lin, Z., Zhuo, R., 2016. What drives consumer knowledge sharing in online travel communities? Personal attributes or e-service factors? Comput. Hum. Behav. 63, 68–74. Zhu, F., Zhang, X., 2010. Impact of online consumer reviews on sales: the moderating role of product and consumer characteristics. J. Market. 74 (2), 133–148.

14