edin

y, T

Information & Management 51 (2014) 833844

ent

atio

wee

erlo

effe

f co

mu

Contents lists available at ScienceDirect

Information &

w.1. Introduction

The Internet and the World Wide Web have become a majorsource of information and knowledge [1,58,60,62]. These resourcesrepresent a powerful disseminative tool for users to acquire andshare information efciently and easily. The microblog is a newform of communication that has been enabled by a variety ofonline social networking tools (e.g., Facebook, Twitter, Jaiku, andPownce) [51]. The term microblog refers to a set of activities inwhich users broadcast brief text updates about small events intheir daily lives and work, such as what they are reading, thinkingand experiencing. Because microblog entries are typically postedvia mobile phone text messages and are currently restricted to alimited number of characters per entry, the messages are brief,sharp and to the point [12]. Microblogs are used for a wide varietyof social purposes and have rapidly become a popular option foronline social networking. Twitter is the best known microblog [10].Approximately 65 million tweets were posted each day in themonth of June 2010an astonishing 750 tweets sent per secondamong nearly 174 million users worldwide (Twitter, 2010).

The problem of information overload has become widelyrecognized [31]. People in todays information society arebombarded with information regardless of whether they actively

seek such information. The particular characteristics of microblogsallow users to receive a variety of messages from different sourcesvia a wide range of communication channels presented in aninnite number of ways. The exponentially increasing amount ofinformation provided by microblogs may result in what hasbecome known as information overload [32]. Informationoverload is the state in which an individual has been presentedwith an overabundance of information [50,64,86] to the point thathe or she is no longer able to process messages [83]. This problemcauses the process of discovering and retrieving information frommicroblogs to become an inefcient, time-consuming activity.Furthermore, users experiencing information overload may easilybecome frustrated and stressed [57]. All individuals are affected bythe ever-increasing number of information sources. The rate ofincrease in the quantity of available information is both stunningand unpredictable [87]. While too little information may render avirtual community useless, too much information can cause usersto feel overwhelmed and frustrated because they cannot nd whatthey seek. Microblog service providers need ways to ameliorate thechallenges caused by users limited processing capacity. Thus, thisstudy addresses the inverted U-shaped relationship between thestickiness of a microblog and users perceptions of the accessibilityof its information.

The literature on communication theory [69,75,76] notes that ausers information seeking motivation and ability are the mainfactors constituting information processing capacity. Experiencedonline users have been shown to be more capable of processing

Perceived information accessibility

Stickiness

Sense of community

stickiness.

2014 Elsevier B.V. All rights reserved.

* Corresponding author.

E-mail address: alanhsu8399@gmail.com (C.-L. Hsu).

http://dx.doi.org/10.1016/j.im.2014.08.005

0378-7206/ 2014 Elsevier B.V. All rights reserved.Exploring the linkages between perceivand microblog stickiness: The moderatcommunity

Chien-Lung Hsu a,*, Yi-Chuan Liao b

aDepartment of Marketing Management, Takming University of Science and TechnologbDepartment of Business Administration, National Taipei University, Taipei, Taiwan

A R T I C L E I N F O

Article history:

Received 4 June 2012

Received in revised form 29 July 2013

Accepted 7 August 2014

Available online 19 August 2014

Keywords:

A B S T R A C T

In the context of the expon

between perceived inform

U-shaped relationship bet

avoid the information ov

examines the moderating

linear as levels of sense o

increase the sense of com

jo u rn al h om ep ag e: ww information accessibilityg role of a sense of

aipei, Taiwan

ial increase of information in society, this study examines the relationship

n accessibility and microblog stickiness. The results indicate an inverted

n perceived information accessibility and stickiness. To determine how to

ad that can compromise the stickiness of the microblog site, this study

cts of sense of community. The inverted U-shaped relationship becomes

mmunity increase. The results suggest microblog service providers can

nity in order to reduce the negative impact of information overload on

Management

els evier .c o m/lo c ate / im

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844834information efciently and effectively with less perceived infor-mation load than are novice online consumers [24]. A typicalmicroblog provides a dynamic message platform that requires liveinteractions between members to maintain the communityinfrastructure and generate new and updated information. Whenpeople are highly involved with the topic being communicated andare able to process the arguments, they are more likely to bepersuaded. Therefore, the degree to which a user is involved andhas the capacity to process information may inuence both theusers degree of loading perception [90,99] and the extent to whichhe or she can provide social or emotional support to othermembers [77,80]. Thus, this study discusses the logic behindeffective communication messages that can increase the stickinessof microblogs without inducing information overload.

Because a microblog is not only an information site but also asocial networking service, this study asserts that social interactionis the microblog factor most worthy of investigation. However,surprisingly few studies have examined how social interactionmoderates the relationship between the perceived informationaccessibility and stickiness of a microblog. Therefore, this studyadopted the sense of community (consisting of two subfactors:membership and shared emotional connection) as a moderatingfactor.

This study addresses these issues in the literature by examining(1) the inverted U-shaped curve in the relationship between theperceived information accessibility and stickiness of a microblogand (2) the social interaction factor (sense of community) thatmoderates this curvilinear linkage between perceived informationaccessibility and stickiness.

The remainder of this paper is structured as follows. In the nextsection, we review some of the literature regarding microblogs,perceived information accessibility, sense of community andstickiness. We then present our research model and hypotheses.The following section describes the methodology used to verify thehypotheses. The next section includes the analysis and discussionof the results, followed by the theoretical contributions of thisstudy and a discussion of the managerial implications. Afterdescribing the limitations of this study and making suggestions forfurther research, we present our conclusion.

2. Literature review

2.1. Microblog

A microblog is a form of communication in which users candescribe their current status in short posts distributed in the formof instant messages (e.g., G-talk, Yahoo!), mobile phones (SMS),RSS, email or websites [12]. A microblog combines the features ofsocial networking and mobility to share thoughts and opinions, toinvite discussion, and to tell others about ones life. This form ofcommunication provides an opportunity for individuals to sharetheir activities, interact with others, build a reputation and releasepent-up emotionsall in real time. Users are becomingproducers and consumers rather than simply being the audience[19]. While a traditional blog allows entries of unlimited length,microblog entries are typically posted via mobile phone textmessages, in which the number of characters is limited. Messageson microblogs are brief, sharp and to the point. Twitter, which waslaunched in 2006, may be the best-known example of a microblog.Other examples, both active and defunct, include Plurk. Twitter issimilar to an instant messenger that allows users to respond to thequestion what are you doing now? Twitter users post shortinformation updates without a headline or additional informationon their page. Users can follow (i.e., receive feeds from) anyoneelse using the service, from friends and family to public gures andnews outlets. Once users choose whom they want to follow,Twitter provides them with a live news feed of tweets from thosesources. Thus, users receive a constant, real-time stream of news inshort text bites from almost anyone of interest. Users can also sendtweets, which involves instantly delivering messages to their ownfollowers. All of the messages from a members social network arelisted together in chronological order on the members start page.

The microblog fullls the need for a mode of communicationthat is even faster than a blog. By encouraging shorter posts,microblogs reduce the amount of time and thought that users mustinvest to generate contentone of the key factors that distin-guishes microblogs from blogs in general [51]. The secondimportant distinction is the update frequency. On average, aprolic blogger updates his or her blog once every few days. Bycontrast, a microblogger may post several updates within a singleday. The main reasons for microblogging are maintaining contactwith others, raising the visibility of interesting things, gatheringuseful information, seeking help and opinions, and releasingemotional stress [100]. Using Twitter as an example, researchershave dened four types of user intentions: daily chat, conversation,information sharing, and news reporting news.

2.1.1. Daily chat

Daily chatting is the most common intention for Twitter users.The original idea of Twitter was to ask people What are youdoing?; therefore, most Twitter posts pertain to users dailyroutines or current activities.

2.1.2. Conversation

The microblog is not a unidirectional information dissemina-tion website; users can interact in two-way conversations on amicroblog. Approximately 12.5% of all posts in our collected datacontain a conversation, and almost 21% of the users in thecollection have used this form of communication.

2.1.3. Sharing information

Users share information with their friends or followers.Approximately 13% of all posts in the collected data contain alink that was shared.

2.1.4. Reporting news

Twitter emphasized its news-and-information network strate-gy in November 2009 by changing its question to users from Whatare you doing? to Whats happening? Since then, reporting thenews as it occurs has become an important microblogging activity.Many people now use Twitter to report the latest news or tocomment on current events. Some users or agents automaticallypost updated information, such as weather reports and new storiesfrom RSS feeds [51].

Scholars have categorized microblogs into three types based ontheir purpose: information sharing, information seeking, andfriend-wise relationship building. News is posted from informa-tion sources that tend to have a large base of followers; these sourcesmay be individuals or automated services. Users who rarely post butwho follow others statuses regularly are called informationseekers. Friend-wise relationship is a broad term that includesrelationships with many other users, such as family members, co-workers and strangers. Typically, users are motivated to use amicroblog that has a community with which they have developed astrong association [51], that helps keep them informed of activities[73] and that graties their need for contact with others [67].

2.2. Perceived information accessibility

Perceived information accessibility has been a critical factoraffecting information system usage and success [27]. Scholars havedened perceived information accessibility as a combination of the

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844 835type and quantity of information as well as how that information isorganized to be accessible to participants [91]. The communicationchannel is also a major component of perceived informationaccessibility. Nearly every virtual community has certain entrybarriers that prevent some participants from accessing the fullrange of services or information available on the website [94]. Inparticular, microblogs provide a public place in which users cancommunicate and post messages via e-mails, text messages,instant messages or websites.

However, caution is advised, as microblogs sometimes providetoo much or too little information. Users are surrounded by a datasmog, which is an expression for the muck and druck of theInformation Age [87]. Scrutinizing the plethora of messagesreceived daily is both irrational and impossible [72]. Whereastoo much information can make it difcult for users to nd whatthey want, too little information renders the virtual communityuseless. Hence, we consider the optimal quantity of microbloginformation to be an important issue for discussion.

Each person engages his or her cognitive system to processinformation [68]. When people are exposed to a certain amountof external information, their cognitive process can inuencetheir decision regarding which portion of the information will beperceived and comprehended. Specically, a perception ofinformation overload results when the information load appearsto exceed the limits of an individuals cognitive process [42].Accordingly, researchers have identied the information over-load phenomenon, which is an inverted U-shaped relationshipbetween the amount of information and the number of subjectswho make the correct choices [50]. In the study conducted byJacoby et al. [50], the authors acknowledged that they hadomitted individual differences in the ability to process informa-tion and had presumed a universally critical informationoverload point. The inverted U-shaped feature of the model isfound to be the result of information processing limitations onthe part of individuals and is consistent with various otherinformation processing models [74]. Similarly, the criticalmass effect occurs when the information received becomes ahindrance rather than help, even if such information ispotentially useful [64].

A large amount of information or information that is receivedrapidly essentially becomes noise [55]. Under such situations, thereceivers of information cannot process it efciently. They becomeeasily distracted or stressed and are thus prone to making moredecision errors, possibly raising the costs in terms of time andeffort. Information overload is the point at which there is so muchinformation that using it effectively is no longer possible [36].Psychological symptoms associated with information overloadinclude feeling overwhelmed or lost [11], frustrated [45], stressed[21], anxious, confused or depressed [39].

The ability to process information differs from individual toindividual. People apply different information processing strate-gies to minimize the effect of information overload [42].Information processing is a dynamically complex procedure thatis heavily inuenced by a persons past experiences [93], whichmay act as a type of internal ltering mechanism. On a socialnetwork platform, users dynamically interact with others andaccumulate experiences, thereby increasing their informationprocessing capacity. Therefore, the current study proposes sense ofcommunity (consisting of two subfactors, membership and sharedemotional connection) as a moderating factor.

2.3. Sense of community

The use of the term community most often refers to a place-oriented concept dened as a group of socially interdependentpeople who participate together in discussions and decisionmaking and who share certain practices that both dene thecommunity and are nurtured by it [13]. According to the Dictionaryof Sociology [66], the most common elements that dene acommunity are the grouping of people within a singular socialstructure and a sense of belonging to that social structure.Currently, the meaning of the word is changing from its originalgeographically specic connotation to a more relationship-specicmeaning. A distinction can be drawn between geographiccommunities and relational communities. The former refers to aneighborhood, town, or region, and the latter involves humanrelationships without reference to location (e.g., communities ofinterest such as hobby clubs, religious groups or fan clubs). Avirtual community is a set of social relationships forged incyberspace through repeated contacts within a specied boundary[56]. A community is characterized primarily by the relationalinteractions or social ties that draw people together. Such acommunity does not form quickly. Throughout history, such acommunity has been regarded as a community of memory, denedin part by its past and its memory of the past.

Sense of community is dened as a result of interactions anddeliberations by people united by similar interests and commongoals [96]. Furthermore, a sense of community is dened as thefeeling that group members have: the feeling that members matterto one another and to the group as well as the shared faith thatmembers needs will be met through their commitment to betogether [70]. Since at least the 1960s, community researchershave been interested in exploring this sense of community becauseit is believed to have benecial outcomes that do not occur whenthis feeling is absent [17].

The current study proposes that when members have a strongsense of community, they may spend more cognitive effort dealingwith relevant information. Members may become deeply involvedwith their community, thereby increasing their informationprocessing abilities and allowing them to manage a larger numberof messages [90]. Members of online groups can experience a senseof virtual community, as the cyberspace equivalent of the sense ofcommunity experienced in traditional face-to-face encounters[16,92]. This sense of virtual community reects the feelings thatindividual members have when belonging to an online socialgroup.

As a construct, the sense of virtual community is complex andstill lacks an established conceptualization. Membership andshared emotional connection are the community factors mostcommonly mentioned in studies discussing the sense of virtualcommunity [16,17,34,56]. From one perspective, membershipincludes self-reinforcing aspect boundaries (knowledge of who iswithin the community), emotional safety (boundaries that provideprotection for intimacy), personal investment (to become avaluable member), the sense of belonging, identication withother group members, and a common symbol system [85].Membership includes the sense that one has invested part ofhimself/herself to become a member and therefore has the right tobelong [6].

From a different perspective, the shared emotional connectionis derived from a shared community history, shared events,positive interactions, and identication with the community. Themore frequently people interact, the more likely they are to formclose relationships, which subsequently leads to stronger bonds[70]. According to the theory of planned behavior and the theory ofreasoned action, if individuals believe that others who areimportant to them (e.g., part of their virtual community) wantthem to perform a given behavior, then they will be more likely toperform that behavior. Thus, the sense of virtual community can beregarded as a major source of social inuence that clearly affectusers behavioral intentions. Therefore, the current study proposesthat the sense of community (membership and shared emotional

connections) moderates the effects of the perceived informationaccessibility and stickiness of a microblog.

2.4. Stickiness

A sticky website is critical to success for a sites service provider[84]. Stickiness is the ability of a website to attract and retaincustomers [101]. The term is used to describe not only the ability ofa website to attract and retain visitors [62] but also the sites abilityto induce repeated visits [58]. A website has stickiness when a uservisits the site repeatedly, spends more time browsing than theaverage user does, and digs deeper into site than other users do.Stickiness has been recognized as one of the keys to protability

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844836because users who linger on a website will buy more goods orservices, view more advertisements or participate in other siteactivities [56]. Stickiness is typically used to indicate how well awebsite converts visitors to customers and retains existingcustomers. A users willingness to return is a strong indicator ofloyalty to a website [79]; thus, many recent studies focus onwebsite stickiness to understand user intentions to revisit aparticular website [23,60].

The intention to spend more time on a website is believed to beinuenced by users perceptions of the websites value; in thismanner, a website will attract its appreciators, inducing them tolinger on the site [61]. If individuals feel a website is notinteresting, then they will quickly leave and will not continue tosupport the site. By contrast, users will continue visiting thosewebsites that provide interesting information and satisfy theirneeds; this factor is important in maintaining individuals loyalty[20]. Three indicators are generally used to measure stickiness:duration, frequency and depth. A website is considered stickywhen its users spend an above-average amount of time browsingit, when they visit the site frequently, and when they browse thesite in depth. These indicators are determined primarily by webtrafc metrics that measure the average time spent at the site pervisit and the average number of pages viewed per visit [29]. Apreferred website will have increased stickiness when its usersbecome deeply committed to reusing the site consistently in thefuture. Moreover, stickiness to a community includes userdependency on certain areas and functions of the community ora cross-dependency on certain groups of people within thecommunity.

3. Research method and hypotheses

3.1. Effect of perceived information accessibility on stickiness

The purposes of this study are to examine the inverted U-shaped relationship between the perceived information accessi-bility and stickiness of microblogs and to explore the moderatingeffect of sense of community on this curvilinear linkage (see Fig. 1).

Fig. 1. Conceptual model.We posit that there is an inverted U-shaped relationshipbetween perceived information accessibility and stickiness. Inaccordance with [91], this study denes perceived informationaccessibility as the users perception of the type and amount ofinformation that is accessible. If the perceived informationaccessibility level is too high, then it may cause informationoverload. Information overload is dened as a decline in responserates that occurs when the input of information surpasses a usersprocessing capacity limits [89]. Decision scientists have describedhuman beings as information processing systems with limitedinformation processing capacity [24]. When a given informationload increases, a decision maker will also increase his or her effortto process this information. Once the input surpasses usersprocessing capacity, people are overloaded by the processing ofinformation, and the response rate declines [42,86], Therefore,because of their limited processing capacity, users can becomecognitively overloaded if they attempt to process too muchinformation in a limited amount of time, resulting in confusion,cognitive strain and other dysfunctional consequences. In the samevein, previous studies have shown that when too much informa-tion is readily available, users feel less capable of processing suchinformation, and they perceive that the informativeness of the siteis lower [75].

For virtual communities, communication channels constitute amajor component of perceived information accessibility [91].Users with low levels of perceived information accessibility willsense a lower level of access to such communication channels andwill thus experience a smaller quantity and less variety ofinformation. By contrast, a higher level of perceived informationaccessibility leads to easy access to a greater amount and variety ofinformation, which can lead to higher usage of that informationand stronger perceptions that a certain site is easy to use. Themicroblog platform provides many access channels, such as instantmessages (e.g., G-talk and Yahoo!), mobile phones (SMS), RSS,email or websites, and users can easily access their microblogs atany time in any place where connectivity is available. Thus,perceived information accessibility has been found to have apositive effect on perceived usefulness [78]. The rapid, almostinstantaneous character of microblogging allows users to quicklychoose what information they want to follow from a vast diversityof possible sources. Thus, this study argues that microblog usageallows users to gain a digestible amount of information via instantmessages while simultaneously facilitating interactions with otherparticipants. Therefore, the combination of information and aninstant community channel with friend or follower relationshipswill induce participants to spend more time on a site, thusincreasing their intention to return. This combination increases thestickiness of a site.

However, time and capability limitations cause informationoverload, in which people with too much information becomeconfused and make poor decisions [15,53,63]. The impracticabilityof too much information is caused by bounded rationality.However, it is importantboth theoretically and practicallytoconsider that people need sufcient information to be able toperform effectively at work (a desirable outcome), while consid-ering that excessive amounts of information can lead to stress anddistress (an undesirable outcome) [64]. Along similar lines, whenusers perceive information overload on social networking sites,they may become less attentive and signicantly decrease theiractivities on such networks. This situation can lead to emotionalstress and dissatisfaction for users [35]. The current study infersthat interactions in real-life situations and sustainable dynamicinteractions that occur through virtual publics (online sites wherethe public can gather) are similarly restricted by informationoverload. We argue that such a situation is the result of limitationsin an individuals ability to effectively process a variety of patterns

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844 837of virtual public interactions over sustained periods of time [52].When microblog users obtain all the information that they desireand believe that the sources are correct, they may increase theirefforts to process and respond to various types of information.However, because individuals continue to interact dynamicallywith different parts of a community and because incomingmessages constantly emerge, an individuals limited processingcapacity can be easily overwhelmed. An excessive amount ofcommunication (i.e., messages) thus results in informationoverload. Therefore, when too many messages are delivered,individuals who are unable to respond adequately may becomeeasily distracted or stressed and may reduce their time on themicroblog.

As a consequence, the current study argues that when perceivedinformation accessibility increases, the stickiness of microblogswill also increase. However, too much information could cause areduction in the level of stickiness.

H1. The relationship between perceived information accessibilityand stickiness is an inverted U-shaped curve.

3.2. Moderating role of sense of community

Membership is the rst subfactor of the sense of communityfactor postulated to moderate the curvilinear relationshipbetween perceived information accessibility and stickiness.Accordingly, this current study denes membership as one ofthe main aspects of individuals identication with a group in thesense that people come to view themselves as members of acommunity [30,70]. Membership is a feeling of belonging, of beinga part of something. This sense of belonging is crucial forparticipation in virtual communities because no involvement orparticipation would occur if this feeling were absent [59].Membership also has boundaries, which means that individualsform a certain self-awareness of the virtual community member-ship of the microblog, allowing them to distinguish people whobelong from those who do not [7]. These boundaries provide asense of emotional safety that encourages the development ofintimacy [70]. Furthermore, members use common language andsymbols in their virtual community to expand their ownboundaries. In an emotional sense, individuals who becomeinvolved and interact within a group will earn a place in the group,and as a result of this investment, membership will be morevaluable. Thus, membership fosters loyalty, trust, and citizenshipbehavior in group settings [14,71] and explains why some peopleseek to maintain committed relationships in the virtual commu-nity of a microblog.

The current study contends that as a message or piece ofinformation increases in personal relevance, it becomes moreimportant and applicable to the development of veridical opinions.Therefore, people are more motivated to devote the cognitive effortrequired to evaluate a member-relevant message or piece ofinformation when they have a strong sense of membership ratherthan a weaker sense of membership. An important part ofmaintaining a relationship with a group is to actively participatein online social interactions [49]. Stated differently, a strong senseof membership makes people feel like they belong; thus, they willspend more time in the virtual community and share messageswith the group. Therefore, stickiness induced by a strong sense ofmembership may mitigate the negative effect associated withhigher levels of perceived information accessibility. Conversely,people in an unfamiliar relationship are relatively poor atresponding to messages and sharing messages with others inthe group. Hence, perceived information accessibility relies on thelevel of membership in participating in a microblogging virtualcommunity.H2. Membership positively moderates the inverted U-shaped re-lationship between perceived information accessibility and sticki-ness.

Shared emotional connection (SEC) is the second sub-factor of thesense of community that is expected to moderate the nonlinearrelationship between perceived information accessibility andstickiness. An SEC is partially based on mutual history. Theinteractions of members in mutual and specic events maydevelop or restrict the strength of ones relationship in acommunity [70]. A shared emotional connection refers to thebonds developed over time through positive interactions withother community members. More interaction amongst themembers of a group increase the likelihood that members willbecome close [48]. For example, the community can discuss a topicon the microblog or hold a member activity on the site, such asvoting, competitions or games. If members experience such aprocess positively, then their relationships may become stronger,which facilitates cohesion. By contrast, if interactions areambiguous or argumentative, then group cohesiveness will berestricted [26,44]. Furthermore, the more important the mutualevent is to those involved, the greater the community bond is [70].

This study argues that when members have a mutual history,they are more able to expend the cognitive effort needed to processmessages or information. As previously noted, microbloggers sharethoughts and opinions and invite discussion, primarily based ontheir daily routines or current activities. The greater the informa-tion availability as a function of such variables as messagerepetition, the greater the ability of message recipients tothoughtfully consider issue-relevant information and becomemore deeply committed to the community. Therefore, mutualhistory is likely to provide members with a greater ability to absorbmessages or information that is repeated regularly. An importantpart of sustaining this relationship with a group is investment inonline interactions. A deeply shared emotional connectiongenerated by mutual history is likely to provide members withmore incentive and capacity to remain in the community.Therefore, stickiness from shared events may mitigate the negativeeffects associated with higher levels of perceived informationaccessibility. By contrast, people interacting in ambiguous orargumentative events may become increasingly reluctant torespond to messages or to share messages with others, and theymay even abandon a community for this reason.

H3. Shared emotional connection positively moderates theinverted U-shaped relationship between perceived informationaccessibility and stickiness.

4. Research methodology

4.1. Data collection and sampling plan

The data used in this research were collected via an onlinesampling survey hosted at my3q.com (http://www.my3q.com).The respondents were Taiwanese users who had followed orbecome fans of the microblog webpage of a corporation (e.g., Plurk,Twitter, Facebook). To encourage participation, the respondentswere offered the opportunity to win a lottery-based prize (7-Eleven icash). The respondents were asked to evaluate the items inthe questionnaire based on their usage behavior on the microblogthat they frequented most. Thus, the rst question asked therespondents to indicate which microblog they tend to use themost. All items were measured on a 7-point Likert scale anchoredby 7 = strongly agree and 1 = strongly disagree. A total of 562respondents completed the questionnaire. Of these responses, 26were considered invalid (too many answers were missing, or the

to free the parameter. A chi-squared test of the nested modelsassessed whether the chi-squares were signicantly lower for theunconstrained models. Most of the combinations resulted in ahigher critical value (Dx2(1) = 3.84 at the 5% signicance level),indicating acceptable discriminant validity for each scale. Second,we examined discriminant validity to determine whether thecondence interval (two standard errors) around the correlationestimate between the two factors included a value of 1.0 [5]. Theresult showed that none of the condence intervals for any of thefactor pairs had a value of 1.0. Thus, we concluded that our measureswere valid and reliable. Table 2 shows the standard deviations of themeans and the correlations of the constructs used in the followinganalyses.

We used a hierarchical regression analysis to test ourhypotheses because this technique allows the examination ofassociations for evidence of nonlinearity. To eliminate the biasresulting from multicollinearity, we followed a straightforwardprocedure suggested by Friedrich [40]. The criterion (perceived

Table 1The characteristics of the respondents.

Characteristic Sample Percentage (%)

Gender

Male 103 19

Female 433 81

Age

Under 19 15 3

1925 years old 373 69

2630 years old 101 25

3135 years old 13 3

More than 35 years old 0 0

Table 2Means, standard deviations, correlations.

Variable Perceived information

accessibility

Membership SEC Stickiness

1 1.00

2 0.68 1.00

3 0.59 0.84 1.00

4 0.35 0.48 0.43 1.00

Mean 5.08 4.89 4.75 4.57

SD 0.81 0.99 0.91 1.04

Note: SEC, shared emotional connection.

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844838respondent chose the same option too frequently). After discardingthe invalid responses, 536 questionnaires were used for ouranalysis. The valid response rate was 95.37%.

4.2. Measurements

To examine our hypotheses, we specied the denition andmeasurements of each variable. Perceived information accessibili-ty was dened as the type and amount of information organized insuch a way that was accessible to participants [91]. We measuredperceived information accessibility with ve items that, forexample, asked the respondents to evaluate the amount ofinformation and the available channels of the microblog. Thedenition of a sense of community is the feeling of belongingsensed by group members, along with a shared emotionalconnection with other members. We measured sense of commu-nity with eight items in total: three items for membership and vefor shared emotional connection (SEC). The current study denesmembership as one of the main aspects of individuals identica-tion with a group in the sense that people come to view themselvesas members of a community. This feeling of belongingof being apart of somethingand shared emotional connection refer to thebonds developed over time through positive interactions withother community members. As an example, one membership itemasked the respondents to evaluate their sense of belonging inrelation to the microblog. An example of a shared emotionalconnection item is a question asking the respondents to evaluatetheir connection with other members of the microblog. Wemeasured stickiness, which is the ability of a website to attract andretain visitors [62], using four items, including one item that askedthe respondents to evaluate the cost of switching to anothermicroblog.

4.3. Descriptive statistics

4.3.1. Descriptive statistics of the respondents

The descriptive information included participant character-istics, such as gender, age, educational background, and job. Thisinformation also included behaviors related to the use ofmicroblog website functions, such as the time spent on theInternet, the preferred microblog service provider, the time spentusing microblogging services, and the frequency of postingmessages and reading other bloggers messages. There were103 male respondents and 433 female respondents. Approxi-mately 94% of the respondents were 1930 years old. Approxi-mately 96% of the respondents had earned more than a bachelorsdegree. A large majority of the respondents (76%) used theInternet 79 h a day and 56% spent 79 h on microblog websites inparticular (Table 1).

4.3.2. Descriptive statistics of the questionnaire responses

Descriptive statistics regarding the completed questionnairesare shown in Appendix. There are ve items regarding perceivedinformation accessibility, eight items regarding sense of commu-nity, and ve items regarding stickiness.

4.3.3. Data analysis procedure and methods

After the online questionnaire, a conrmatory factor analysis(CFA) of four latent variables was used to assess the dimensionali-ty, reliability and validity of the scales. The t indices are presentedin Appendix. The CFA model showed a reasonable t to the data,with NFI, NNFI, CFI, and IFI all exceeding 0.90 (x2 = 341.07;d.f. = 129) [5]. Composite reliability (CR) was calculated for eachconstruct using the procedures suggested by Fornell and Larcker[38]. The CR for the four latent constructs ranged from 0.84 to 0.90,all exceeding the acceptable CR level (0.70) suggested by Bagozziand Yi [8]. All item loadings were signicant at the 1% level,indicating convergent validity [9].

The discriminant validity of all scales was assessed using twodifferent techniques. First, we analyzed all pairs of constructs intwo-factor CFA models [5]. Each model was run twice: once toconstrain the correlation between the constructs to unity and once

Education

Master 140 26

Bachelor 376 70

Senior high school 19 4

Junior high school 1 0

Elementary 0 0

Mean hours of microblog usage per day

Less than 1 h 42 8

13 h 64 12

46 h 102 19

79 h 281 56

1012 h 13 5

Mean hours of internet usage per day

Less than 1 h 8 1

13 h 12 2

46 h 92 18

79 h 386 76

1012 h 4 3

el 3

2 (6 (5

1 (0

8 (2

6 (2

8 (

1

0

8*

nal

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844 839information accessibility) and the regressors (membership andshared emotional connection) were rst standardized, and thecross-product terms and quadratic terms were then formed. Thevariance ination factors (VIFs) for all coefcient estimates werebelow the cutoff value of 5 [43], thus indicating that multi-collinearity did not contaminate the results. Entering theindependent variables and interaction variables in one blockand the squared variables in a second block enabled thedetermination of the signicance of the curvilinear relationshipsover and above any linear relationships. In regression models, acurvilinear relationship is evident if the addition of the nonlinearrelationship is evident and if the addition of the nonlinearpredictor results in signicant incremental variance after thelinear relationships have been considered [3]. A multiplehierarchical regression analysis is widely used to assess curvilinearrelationships in management and marketing research.

4.4. Regression analysis

Table 3 shows the results of the regression analysis. Eachconstruct is composed of a summated index of the items thatconstitute the construct. We entered the perceived informationaccessibility and moderator variables (membership and sharedemotional connection) in Model 1 and then added the interactionbetween perceived information accessibility and the moderatorvariables in Model 2. To test for a curvilinear relationship, weadded the quadratic terms to the regression equation in Model 3and added the interaction terms between the squared perceivedinformation accessibility variable and moderator variables inModels 4 and 5 [3]. Note that the signs of the regression coefcients

Table 3Results of regression analyses for stickiness (moderator: sense of community).

Variable Model 1 Model 2 Mod

PIA 0.02 (0.48) 0.01 (0.08) 0.0M 0.35 (5.02)*** 0.34 (5.07)*** 0.3

SEC 0.12 (1.91)* 0.02 (0.30) 0.0

PIA M 0.13 (2.07)* 0.1PIA SEC 0.15 (2.26)* 0.1PIA2 0.0PIA2 M PIA2 SEC R2 0.21 0.30 0.3

Adj2 0.21 0.30 0.3

F-change 46.58*** 30.45*** 5.1

Note: PIA, perceived information accessibility; M, membership; SEC, shared emotio* p < 0.05.** p < 0.01.*** p < 0.001.for the squared terms represent the direction of curvatureproduced by the effects of perceived information accessibility onstickiness. In other words, the regression-coefcient estimate ofthe quadratic terms for perceived information accessibilitydetermines whether the U-shaped relationship is normallyoriented or inverted. The coefcients of the interaction termsassociated with the coefcients of the squared terms depict howthe nonlinear relationships between perceived informationaccessibility and stickiness are moderated by the sense ofcommunity variables.

As shown in Model 3, the squared term was negative andsignicant (b = 0.08, p < 0.05). A one-tailed test was used for thehypotheses because directional predictions were offered. Thechange in R2 was also signicant; the addition of the perceivedinformation accessibility quadratic term (Model 3) to the maineffect model (Model 2) increased the R2 by approximately 1%, thussupporting the salience of the curvilinear effect. Consequently,Hypothesis 1 was supported. Fig. 2 shows the plot of therelationship between perceived information accessibility andstickiness.

Model 4 shows the moderating effect of membership on the linkbetween perceived information accessibility squared and sticki-ness. The addition of the interaction term (Model 4) to the squaredterms model (Model 3) increased the R2 by approximately 4%. TheF-value for the incremental R2 value achieved statistical signi-cance at the 5% level. A positive and signicant coefcient estimatewas observed (b = 0.12, p < 0.05), indicating that the curvilinearrelationship between the perceived information accessibilitysquared term and stickiness depends on membership. As predictedin Hypothesis 2, these results suggested that a strong sense ofmembership mitigates the inverted U-shaped relationship be-tween perceived information accessibility and stickiness.

To aid in interpretation, the nature of this interaction wasexamined using procedure tests for the signicance of regres-sion-coefcient estimates for the perceived information acces-sibility squared variable at a distance of one standard deviationaway from the means of the membership moderator [3]. Asignicant negative relationship between perceived informationaccessibility squared and stickiness was found (b = 0.23,p < 0.05) for a low level of membership. For a high level ofmembership, the relationship between perceived informationaccessibility squared and stickiness was positive but notsignicant (b = 0.01, p > 0.05) (see Fig. 3).

Model 5 displays the extent to which a shared emotionalconnection moderates the effects of the squared term for perceivedinformation accessibility on stickiness. After considering theshared emotional connection effect, we observed that thecoefcient estimate for the interaction term between the perceived

Model 4 Model 5 VIF

0.33) 0.13 (2.04)* 0.16 (3.15)*** 2.08.32)*** 0.24 (3.03)** 0.33 (5.09)*** 3.62

.20) 0.02 (2.59)* 0.11 (1.70)* 2.80

.74)** 0.25 (2.71)** 0.26 (4.00)*** 4.68

.35)** 0.06 (0.75) 0.01 (0.21) 4.602.28)* 0.11 (3.21)*** 0.13 (3.88)* 1.75

0.12 (3.72)*** 2.52

0.18 (6.69)*** 3.11

0.35 0.37

0.34 0.36

29.31*** 44.79***

connection.information accessibility squared term and its inuence ispositively signicant (b = 0.18, p < 0.05), supporting Hypothesis3. This result suggests that a high level of shared emotionalconnection mitigates the inverted U-shaped relationship betweenperceived information accessibility squared and stickiness. Thenature of this interaction was examined using Aiken and Wests [3]procedure. We plotted the interactions and conducted simpleslope tests. The simple slope test involved dividing the moderatorinto a high group (one standard deviation greater than the mean)and a low group (one standard deviation less than the mean) andre-estimating the relationship between perceived informationaccessibility and stickiness. At a high level of shared emotionalconnection, the relationship between perceived informationaccessibility squared and stickiness was positive but not signi-cant (b = 0.05, p > 0.05). At a low level of shared emotionalconnection, the relationship between perceived informationaccessibility squared and stickiness was negative and signicant(b = 0.31, p < 0.05) (see Fig. 4).

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 8338448405. Discussion and implications

This empirical survey-based study was conducted across avariety of different microblogs and found overall support for theproposed hypotheses. The purpose of this study was to verify theinverted U-shaped relationship between perceived informationaccessibility and the stickiness of microblogs. This study alsoproposed information processing strategies to mitigate the effectsof the information overload problem. That is, this study appliedsense of community as a moderating factor to weaken thecurvilinear relationship between perceived information accessi-bility and stickiness. As the results show, the hypotheses of thisstudy received support and raised several interesting issues thatare discussed below.

Fig. 2. The curvilinear relationship between

Fig. 3. Moderating effect of membership on the curvilinear rela6. Theoretical contributions

Academically, this study builds on a growing body of work thatexamines the inverted U-shaped relationship between perceivedinformation accessibility and microblog stickiness [50,64,86].Specically, this study suggests that useful information haspositive effects on individuals, inducing them to spend more timeon a website; however, beyond a certain level, even usefulinformation becomes detrimental to the stickiness of the site.Previous theories have suggested that excessive information maybe harmful (regardless of whether such messages are useful)because of the limited processing capacity of individual [35,64,65].This nding supports anecdotal reports that users often leave amicroblog when they perceive a decline in the amount of

information accessibility and stickiness.

tionship between information accessibility and stickiness.

tion

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844 841information. Conversely, if users feel that there is too muchinformation on a microblog, then they may not nd what theywant or may have insufcient time and ability to manage theoverload. Thus, microblog service providers should pay carefulattention to the amount of information that is provided: thisamount should be adequate but not overwhelming.

Fig. 4. Moderating effect of SEC on the curvilinear relaThis study adopts a sense of community factor consisting ofmembership and shared emotional connection in effectivelymoderating the curvilinear relationship between perceivedinformation accessibility and microblog stickiness. The analysisresults show that both subfactors have signicant effects inweakening the curvilinear relationship between perceived infor-mation accessibility and microblog stickiness. With regard tomembership, we nd a positive relationship between perceivedinformation accessibility and stickiness. Particularly, with a strongsense of membership, the relationship between perceivedinformation accessibility and stickiness becomes linear. The resultsindicate that participants who feel an intense sense of belongingare more motivated to engage in the cognitive effort required toevaluate member-relevant messages or information [49]. Further-more, members use shared language and symbols that enhancetheir capacity to process information without surpassing theirlimited cognitive ability. Therefore, users will spend more time ona site and even deepen their commitment to such communities orsites. By contrast, an inverted U-shaped relationship exists underconditions in which the sense of membership is weak, whichindicates that perceived information accessibility becomes detri-mental to stickiness beyond a certain level. If users experience aweak sense of membership, then they are less motivated toparticipate in online social interactions. Furthermore, under suchconditions, the lack of common language and symbols increasesthe difculty of absorbing the constantly emerging messages orinformation. Therefore, as messages or information increases,users can easily perceive information overload because of thebounded rationality, thus decreasing their willingness to attend[65]. Such individuals become less attentive and signicantlyreduce their activities on these sites [35].

The results show that a shared emotional connection (SEC) alsoplays a moderating role in the curvilinear relationship betweenperceived information accessibility and stickiness. With a strongSEC, the relationship between perceived information accessibility

ship between information accessibility and stickiness.and stickiness becomes linear. This nding indicates that userswho have a closer relationship and shared history with others inan online community will build the capability to absorb largeramounts of messages and information that are regularly repeated[25]. Microblogs provide a platform that allows individuals toshare opinions and exchange information and permits articula-tion to achieve effective communication. This capability results inshared understanding and cohesion, and participants are there-fore willing to expend cognitive effort to process messages. Themore messages or information that emerge on such sites, thehigher the likelihood that a participant will remain in thecommunity. Conversely, with a weaker SEC, a negative relation-ship between perceived information accessibility squared andstickiness is observed. That is, there are positive incentives forusers to process information, but once a certain threshold isexceeded, the accumulated noise becomes detrimental tostickiness. These results conrm our expectation that whenincreasing the quantity of messages or information in acommunity with a weak shared emotional connection, users willeventually deem such messages unnecessary, which may causethem to become stressed and distracted to the point that theyactually leave the community.

7. Managerial implications

Our empirical nding regarding the relationship of perceptionsof information accessibility to microblog stickiness has strategicimplications for both microblog service providers and companies.

Specically, our study suggests that once users sense that amicroblog contains insufcient information, they may intend toleave, whereas users who perceive an excess amount of informa-tion on a microblog may feel uncomfortable and decrease theirattachment to the microblog. Thus, microblog service providers

webpages of their microblogs. The spirit of microblogging is tocommunicate and interact; communication is no longer simply aone-way interaction. To conduct marketing through microblog-ging, it is necessary to make friends and build an atmosphere of one

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844842should make an effort to carefully adjust the amount of accessibleinformation.

For example, Plurk uses a Karma point system that averts theproblem of too few or too many messages. The Karma value is amechanism used to evaluate user involvement. According toPlurk.com, some tips on gaining more Karma include inviting realfriends, engaging in quality plurking each day, receivingresponses from other plurkers, and updating ones own prole.However, Plurk will lower the Karma value of a user who spamsother users, posts too much information at one time, is inactive fora long period of time, requests friendships and is rejected too often,and is unfollowed by friends. Therefore, a higher Karma valueindicates that a user has higher value to the community andgreater inuence with Plurk.

We also offer some clear implications for organizationsregarding the building of continuous conversation channels withusers. Currently, many companies are directing their effortstoward microblog marketing, aiming to provide crucial informa-tion to communicate with customers. Organizations shouldconsider that too little information might fail to give customersthe information that they want, while excessive information mightcause customers to experience cognition overload and inducethem to sever their relationship with the organization. Therefore,we recommend enhancing a company websites sense ofcommunity to reduce the negative effect of perceived informationaccessibility on stickiness.

To mitigate the curvilinear relationship between perceivedinformation accessibility and the stickiness of microblogs, wesuggest that microblog service providers and corporations adoptstrategies that increase users sense of community (including theirsense of membership and shared emotional connections). Forinstance, Plurk allows users to group their friends into cliques,which streamlines the process of sending plurks to a specic groupof members. Users in a clique are interested in the same topic orshare something in common [14,71]. They may build boundariesand distinguish between people who belong and those who do not.Under such situations, even when users sense an increasingamount of information, they may not feel as pressured and may bemore willing to manage what would otherwise appear to be anexcessive information load. Furthermore, taking Facebook as anexample, participants can share emotional connections by buildingtheir relationships with friends or fans. On a fan or group page,users can share histories and events, interact positively, andidentify with one another. Under such conditions, users may nd iteasier to handle the amount of information [50]. Moreover, themore the user interacts in such a microblog, the more likely theyare to continue using Facebook, as they perceive that they haveclose relationships there. The microblog has become an importantmarketing tool, a great platform on which companies can promotetheir products and services, broaden their business network andbuild their brand image. Businesses should not only postinformation but also foster more customer interactions on theirmicroblog pages to strength their connection with customers. Forexample, companies can propose certain topics for customers todiscuss or go directly online to improve interactions on thelarge family on such websites.

8. Limitations and future research

The results of this study have certain limitations and points thatrequire further research. First, this study applied the sense ofcommunity as the moderating factor to reduce the curvilinearrelationship between perceived information accessibility andstickiness. Future studies can use different treatment conditionsto identify different effects. In fact, because the moderators arecontinuous variables, we followed the procedures of Aiken andWest [3] and Cohen and Cohen [25] that suggest choosing onlyhigh and low values. The high level corresponds to one standarddeviation above the mean of the moderator, whereas the low levelcorresponds to one standard deviation below the mean of themoderator. However, consistent with Aiken and West [3], variousother theories, measurement considerations, or previous researchmay suggest interesting moderator values that should be exploredrather than an arbitrary statistical value.

Second, the sense of community is not the only factor thatmoderates the relationship. This study recommends furtherresearch to explore additional factors that trigger motivationand ability, such as various interests, professional elds, and userpersonality types.

Third, because the sampling was conducted on the Internet, thesample collection itself is a limitation of this study. The sample wasnot very general. The respondents were primarily studentsbetween the ages of 20 and 30. Future studies could thus conductsurveys in different elds. Moreover, the research subjects were allmicroblog users in Taiwan. Culture and lifestyles may differ fromcountry to country.

Finally, because microblogging has become prevalent andbecause consumers often create and share informationthusgenerating a signicant amount of collective powera microblogcan both help users obtain the latest information and act as a newtool for business marketing. The question of how positive value canbe provided to customers when corporations use a microblog as amarketing tool is worthy of future research and discussion.

9. Conclusion

This paper has presented seminal research on the nonlinearrelationship between perceived information accessibility and thestickiness of microblogs. This study further examined thecontingent effects of the sense of community on the inverted U-shaped relationship between perceived information accessibilityand microblog stickiness. The results showed a curvilinearrelationship that can be moderated by a sense of community.Furthermore, when a strong sense of membership or emotionalshared connection is felt, the relationship between informationaccessibility and stickiness becomes linear. However, when suchconditions are weaker, an inverted U-shaped relationship exists.The ndings of the current study offer an interesting contributionto the marketing and management literature, and the hypothe-sized relationship in this study is worthy of further examinationusing other constructs.

3, C

cro-b

his m

nt in

site.

site

to t

log

log

.

emb

ebsit

ebsit

to c

high

-blo

to a

site

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844 843References

[1] D. Abfalter, M.E. Zaglia, J. Mueller, Sense of virtual community: a follow up on itsmeasurement, Comput. Hum. Behav. 28 (2), 2012, pp. 400404.

[3] L.S. Aiken, S.G. West, Multiple Regression: Testing and Interpreting Interactions,Sage Publications, Newberry Park, CA, 1991.

[5] J.C. Anderson, D. Gerbing, Structural equation modeling in practice: a review andrecommended two-step approach, Psychol. Bull. 103 (3), 1988, pp. 411423.

[6] E. Aronson, J. Mills, The effect of severity of initiation on liking for a group, J.Abnorm. Soc. Psychol. 59 (2), 1959, pp. 177181.

[7] B.E. Ashforth, F. Mael, Social identity theory and the organization, Acad. Manage.Rev. 14 (1), 1989, pp. 2039.

[8] R.P. Bagozzi, Y. Yi, On the evaluation of structural equation models, J. Acad.Market. Sci. 16 (1), 1988, pp. 7494.

[9] R.P. Bagozzi, Y. Yi, L.W. Phillips, Assessing construct validity in organizationalresearch, Adm. Sci. Q. 36 (3), 1991, pp. 421458.

[10] S.J. Barnes, Modeling use continuance behavior in microblogging services: thecase of twitter, J. Comput. Inf. Syst. 51 (4), 2011, pp. 110.

[11] D. Bawden, C. Holtham, N. Courtney, Perspectives on information overload, AslibProc. 51 (8), 1999, pp. 249255.

[12] C. Beaumont, Tweet, tweet, here comes Twitter, Telegraph 12 (April), 2008, p. 21.http://re8ux2yp4z.scholar.serialssolutions.com/?sid=google&auinit=G&aulas-t=Akehurst&atitle=User+generated+content:+the+use+of+blogs+for+tourism+or-ganisations+and+tourism+consumers&id=doi:10.1007/s11628-008-0054-2&title=Service+business&volume=3&issue=1&date=2009&spage=51.

[13] R. Bellah, R. Madsen, W. Sullivan, A. Swidler, S. Tipton, Habits of the Heart:Individualism and Commitment in American Life, Harper & Row, New York, 1985.

[14] M. Bergami, R.P. Bagozzi, Self-categorization, affective commitment and group

Appendix A

Measure and source Operational measures of construct

Model t indices: NFI = 0.91, NNFI = 0.9

Information accessibility The information retrieved from this miCronbachs a = 0.84CR = 0.84 [91]

I believe the information provided by t

I can obtain complete information I wa

I can easily access this micro-blog web

I can easily browse this micro-blog web

Sense of community: membership I have a very strong sense of belongingCronbachs a = 0.90CR = 0.90 [70]

I consider the members of this micro-b

I like the members of this micro-blog.

Sense of community: SEC I believe the time spent on this micro-bCronbachs a = 0.89CR = 0.90 [70]

I can get what I want from this website

What I want is similar to what other m

The members of this micro-blogging w

The members of this micro-blogging w

Stickiness I think it takes a lot of time and effortsCronbachs a = 0.86CR = 0.87 [62]

The cost of time, money, and efforts is

I dont want to move to a similar micro

system of this website.

Its not worthy to take the risk moving

If I cannot use this micro-blogging web

a Standardized factor loading.self-esteem as distinct aspects of social identity in the organization, Br. J. Soc.Psychol. 39 (4), 2000, pp. 555577.

[15] J.R. Bettman, An Information Processing Theory of Consumer Choice, Addison-Wesley, MA, 1979.

[16] A.L. Blanchard, Testing a model of sense of virtual community, Comput. Hum.Behav. 24 (5), 2008, pp. 21072123.

[17] A.L. Blanchard, M.L. Markus, The experienced sense of a virtual community:characteristics and processes, Adv. Inf. Syst. 35 (1), 2004, pp. 6579.

[19] D.C. Brabham, Moving the crowd at iStockphoto: the composition of the crowdand motivations for participation in a crowd sourcing application, First Monday13 (6), 2008, pp. 122.

[20] S.J. Brock, Selling alliances: issues and insights, Ind. Market. Manage. 26 (2), 1997,pp. 146161.

[21] E.J. Burge, Learning in computer conferenced contexts: the learners, J. DistanceEduc. 9 (1), 1994, pp. 1943.

[23] M.K. Chang, W. Cheung, Determinants of the intention to use Internet/WWW atwork: a conrmatory study, Inf. Manage. 39 (1), 2001, pp. 114.

[24] C.Y. Chen, R.A. Shang, C.Y. Kao, The effects of information overload on consumerssubjective state towards buying decision in the Internet shopping environment,Electron. Commer. Res. Appl. 8 (1), 2009, pp. 4858.

[25] W.M. Cohen, D.A. Levinthal, Absorptive capacity: a new perspective on learningand innovation, Admin. Sci. Q. 35 (1), 1990, pp. 128152.

[26] S.W. Cook, Motives in conceptual analysis of attitude related behavior, in: W.J.Arnold, D. Levine (Eds.), in: Proceedings of the Nebraska Symposium on Motiva-tion, University of Nebraska Press, Lincoln, NE, 1970.

[27] M.J. Culnan, The dimensions of accessibility to online information: implicationsfor implementing ofce information systems, ACM Trans. Off. Inf. Syst. 2 (2), 1984,pp. 141150.[29] E. Demers, B. Lev, Rude awakening: Internet shakeout in 2000, Rev. Account. Stud.6 (23), 2001, pp. 331359.

[30] U.M. Dholakiaa, R.P. Bagozzia, L.K. Pearob, A social inuence model of consumerparticipation in network- and small-group-based virtual communities, Int. J. Res.Market. 21 (3), 2004, pp. 241263.

[31] A. Edmunds, A. Morris, The problem of information overload in business orga-nisations: a review of the literature, Int. J. Inf. Manage. 20 (1), 2000, pp. 1728.

[32] M. Efron, Information search and retrieval in microblogs, J. Am. Soc. Inf. Sci.Technol. 62 (6), 2011, pp. 9961008.

[34] H.K. Ellonen, M. Kosonen, K. Henttonen, The development of a sense of virtualcommunity, J. Web Based Commun. 3 (1), 2007, pp. 114130.

[35] M.J. Eppler, J. Mengis, The concept of Information overload: a review of literaturefrom organization science, accounting, marketing, MIS, and related disciplines,Inf. Soc. 20 (5), 2004, pp. 325344.

[36] J. Feather, In the Information Society: A Study of Continuity and Change, LibraryAssociation, London, 1998.

[38] A. Fornell, D.F. Larcker, Evaluating structural equation models with unobservablevariables and measurement error, J. Market. Res. 18 (1), 1981, pp. 3950.

[39] J.F. Fournier, Information overload and technology education, in: Proceedings ofthe Seventh International Conference of the Society for Information Technologyand Teacher Education, Phoenix, AZ, 1996.

[40] R.J. Friedrich, In defense of multiplicative terms in multiple regression equations,Am. J. Pol. Sci. 26 (4), 1982, pp. 797833.

[42] M.-L. Grise, R.B. Gallupe, Information overload: addressing the productivityparadox in face-to-face electronic meetings, JMIS 16 (3), 2000, pp. 157186.

[43] J.F. Hair Jr., W.C. Black, B.J. Babin, R. Anderson, Multivariate Data Analysis, PearsonPrentice Hall, Upper Saddle River, NJ, USA, 2009.

[44] R.L. Hamblin, Group integration during a crisis, Hum. Relat. 11 (1), 1958, pp. 6776.[45] L. Harasim, Teaching and learning on-line: issues in computer-mediated graduate

SFLa t-Value

FI = 0.94, IFI = 0.94

log is correct. 0.88 22.16

icro-blog. 0.88 24.11

this micro-blog. 0.64 15.40

0.58 13.71

. 0.55 12.92

his micro-blog. 0.85 23.33

my good friends. 0.90 25.44

0.87 24.12

is worthwhile. 0.83 22.27

0.84 22.62

ers of this website want. 0.80 20.97

e solve problems together. 0.77 20.07

e get alone very well. 0.75 19.19

reate a new account in other similar websites. 0.62 14.88

for me to change the micro-blogging website I am using. 0.77 14.91

gging website because I am already familiar with the 0.86 23.21

nother micro-blogging website. 0.83 22.07

anymore, itd be a big pity. 0.66 16.10courses, Can. J. Educ. Commun. 16 (2), 1987, pp. 117135.[48] F.M. Hsu, T.Y. Chen, S. Wang, The role of customer values in accepting information

technologies in the public information service sector, Serv. Ind. J. 30 (7), 2010, pp.10971111.

[49] W.M. Hur, J. Park, M. Kim, The role of commitment on the customer benetsloyalty relationship in mobile service industry, Serv. Ind. J. 30 (14), 2010, pp.22392309.

[50] J. Jacoby, D.E. Speller, C.A. Kohn, Brand choice behavior as a function of informa-tion load, J. Market. Res. 11 (1), 1974, pp. 6369.

[51] A. Java, X. Song, T. Finin, B. Tseng, Why we twitter: understanding microbloggingusage and communities, in: Proceedings of the 9th WebKDD and 1st SNA-KDD2007 Workshop on Web Mining and Social Network Analysis, San Jose, CA, 2007.

[52] Q. Jones, G. Ravid, S. Rafaeli, Information overload and the message dynamics ofonline interaction spaces: a theoretical model and empirical exploration, Inf. Syst.Res. 15 (2), 2004, pp. 194210.

[53] K.L. Keller, R. Staelin, Effects of quality and quantity of information on decisioneffectiveness, J. Consum. Res. 14 (2), 1987, pp. 200213.

[55] E. Klapp, Overload and Boredom: Essays on the Quality of Life in the InformationSociety, Greenwood Press, Connecticut, 1986.

[56] J. Koh, Y.G. Kim, Sense of virtual community: a conceptual framework andempirical validation, Int. J. Electron. Commerce 8 (2), 2003, pp. 7593.

[57] A.S. Krishen, R.L. Raschke, P. Kachroo, A feedback control approach to maintainconsumer information load in online shopping environments, Inf. Manage. 48 (8),2011, pp. 344352.

[58] D. Li, G.J. Browne, J.C. Wetherbe, Why do Internet users stick with a specic web site? Arelationship perspective Int. J. Electron. Commerce 10 (4), 2006, pp. 105141.

[59] H.F. Lin, Determinants of successful virtual communities: contributions fromsystem characteristics and social factors, Inf. Manage. 45 (5), 2008, pp. 522527.

[60] J.C.C. Lin, H. Lu, Towards an understanding of the behavioral intention to use aweb site, Int. J. Inf. Manage. 20 (3), 2000, pp. 197208.

[61] G. Lippert, R. Wilder, Add stickiness to your site with an intuitive and user-friendly interface, Inside the Internet, 2001, December.

[62] Z. Liu, J. Xu, Study of online stickiness: its antecedents and effect on repurchaseintention, 2010 International Conference on E-Business, E-Management and E-Learning, 2010116120.

[63] N.K. Malhotra, Information load and consumer decision making, J. Consum. Res. 8(4), 1982, pp. 419430.

[64] R.S. Mano, G.S. Mesch, E-mail characteristics, work performance and distress,Comput. Hum. Behav. 26 (1), 2010, pp. 6169.

[65] J.G. March, H.A. Simon, Organizations, John Wiley, New York, 1958.[66] G. Marshall, A Dictionary of Sociology, Oxford University Press, Oxford, 1998.[67] C. Masullo, Tweet this: a uses and gratications perspective on how active Twitter

use graties a need to connect with others, Comput. Hum. Behav. 27 (2), 2011, pp.755762.

[68] W.J. McGuire, Some internal psychological factors inuencing consumer choice, J.Consum. Res. 2 (3), 1976, pp. 302319.

[69] V. McKinney, K. Yoon, F.M. Zahedi, The measurement of web-customer satisfac-tion: an expectation and disconrmation approach, Inf. Syst. Res. 13 (3), 2002, pp.296315.

[70] D.W. McMillan, D.M. Chavis, Sense of community: a denition and theory, J.Community Psychol. 14 (1), 1986, pp. 623.

[71] J.P. Meyer, D.J. Stanley, L. Herscovitch, L. Topolnytsky, Affective, continuance, andnormative commitment to the organization: a meta-analysis of antecedents,correlates, and consequences, J. Vocat. Behav. 61 (1), 2002, pp. 2052.

[72] N. Miller, G. Maruyama, R.J. Beaber, K. Valone, Speed of speech and persuasion, J.Pers. Soc. Psychol. 34 (4), 1976, pp. 615624.

[73] E. Mischaud, Twitter: expressions of the whole self. An investigation into userappropriation of a web-based communications platform, Msc in Politics andCommunication, London School of Economics and Political Science, 2007.

[74] A. Newell, H.A. Simon, Human Problem Solving, Prentice-Hall, Englewood Cliffs,NJ, 1972.

[75] D.H. Park, S. Kim, The effects of consumer knowledge on message processing ofelectronic word-of-mouth via online consumer reviews, Electron. Commer. Res.Appl. 7 (4), 2008, pp. 399410.

[76] R. Petty, J. Cacioppo, Attitudes and Persuasion: Classic and ContemporaryApproaches, Brown, Dubuque, IA, 1981.

[77] J. Preece, Online Communities: Designing Usability, Supporting Sociability,Wiley, Chichester, 2000.

[78] S. Rafaeli, R.J. LaRose, Electronic bulletin boards and public-goods: explanationsof collaborative mass-media, Commun. Res. 28 (2), 1993, pp. 277297.

[79] F.F. Reichheld, P. Schefter, E-loyalty: your secret weapon on the web, Harv. Bus.Rev. 78, 2000, pp. 105113.

[80] Y. Ren, R. Kraut, S. Kiesler, Applying common identity and bond theory to design ofonline communities, Organ. Stud. 28 (3), 2007, pp. 377408.

[83] E.M. Rogers, R. Agarwala-Rogers, Organizational communication, in: G.L. Hanne-man, W.J. McEwen (Eds.), Communication Behaviour, Addision-Wesley, Reading,MA, 1975, pp. 218236.

[84] S. Rosen, Stick website is key to success, Commun. World 18 (3), 2001, pp. 3536.[85] M.S. Rosenbaum, A.L. Ostrom, R. Kuntze, Loyalty programs and a sense of

community, J. Serv. Market. 19 (4), 2005, pp. 222233.[86] H.M. Schroder, M.J. Driver, S. Steufert, Human Information Processing, Holt

Rinehart & Winston, New York, 1967.[87] D. Shenk, Data smog: surviving the information glut, Technol. Rev. 100 (4), 1997,

pp. 1826.[89] R. Soucek, K. Moser, Coping with information overload in email communication:

evaluation of a training intervention, Comput. Hum. Behav. 26 (6), 2010, pp.14581466.

[90] W.K. Tan, T.H. Chen, The usage of online tourist information sources in touristinformation search: an exploratory study, Serv. Ind. J. 32 (3), 2012, pp. 451476.

[91] H.H. Teo, H.H. Chan, K.K. Wei, Z. Zhang, Evaluating information accessibility andcommunity adaptively features for sustaining virtual learning communities, Int. J.Hum.-Comput. Stud. 59 (5), 2003, pp. 671697.

[92] L. Tonteri, M. Kosonen, H.K. Ellonen, A. Tarkiainen, Antecedents of an experiencedsense of virtual community, Comput. Hum. Behav. 27 (6), 2011, pp. 22152223.

[93] A. Tversky, D. Kahneman, Extension versus intuitive reasoning: the conjunctionfallacy in probability judgment, Psychol. Rev. 90 (4), 1983, pp. 293315.

[94] M. Valtersson, Virtual Communities, (Master thesis), Umed University, Sweden,1996.

[96] J. Westheimer, J. Kahne, Building school communities: an experience-basedmodel, Phi Delta Kappan 75 (4), 1993, pp. 324328.

[99] Y. Yi, T. Gong, An integrated model of customer social exchange relationship: themoderating role of customer experience, Serv. Ind. J. 29 (11), 2009, pp. 15131528.

[100] D. Zhao, M.B. Rosson, How and why people Twitter: the role that micro-bloggingplays in informal communication at work, GROUP 09: Proceedings of the ACM2009 International Conference on Supporting Group Work, New York, NY, USA,2009, pp. 243252.

[101] C. Zott, R. Amit, J. Donlevy, Strategies for value creation in e-commerce: bestpractice in Europe, Eur. Manage. J. 18 (5), 2000, pp. 463475.

C.-L. Hsu, Y.-C. Liao / Information & Management 51 (2014) 833844844

Exploring the linkages between perceived information accessibility and microblog stickiness: The moderating role of a sense of communityIntroductionLiterature reviewMicroblogDaily chatConversationSharing informationReporting news

Perceived information accessibilitySense of communityStickiness

Research method and hypothesesEffect of perceived information accessibility on stickinessModerating role of sense of community

Research methodologyData collection and sampling planMeasurementsDescriptive statisticsDescriptive statistics of the respondentsDescriptive statistics of the questionnaire responsesData analysis procedure and methods

Regression analysis

Discussion and implicationsTheoretical contributionsManagerial implicationsLimitations and future researchConclusionReferencesAppendix A