Popularity and Influence on Micro- Blogs: One and One is Two?

1

Popularity and Influence on

Micro-Blogs: One and One is Two?

By

Martin P. Bügel

Master Thesis

June 2013

University of Groningen

Faculty of Economics and Business

MSc Business & ICT

+31(0)6 522 510 36

m.p.bugel@student.rug.nl

2

A

BSTRACT

No clear view of influence and popularity of a person exists in current literature. Often only one of these views is considered and overlapping definitions have been given to both, sometimes considering them the same. This research aims to get a better image of popularity and influence of people by investigating them in a micro-blogging environment. It is hypothesized that popularity is often the result of a consideration of elements in the surroundings (e.g. societal position of the blogger), leading to heuristic information processing, whereas influence is often the result of the consideration of core elements (e.g. quality), leading to systemic information processing. No support is found for the relation between heuristic information processing and popularity, although it does marginally lead to influence. Strong support is found for the relation between systemic information processing and influence, but it also leads to popularity to a lesser extent. Conclusions and research implications are discussed.

3

P

REFACE

This thesis was written to graduate as a student Business Administration, specialisation Business & ICT. After months of hard work I am finally able to deliver the end result. Originally, I started this research in April 2012, but I had to shift it back to October 2012 down to the workload caused by board membership and teaching assistant activities throughout the period in between. As such, the research basically started from scratch in October.

It is safe to say this was the toughest challenge in my life as a student. It took more time than expected, but to a satisfying result and therefore I’m both proud and happy to have finished. The toughest part was model development which took a lot of time and discussion to get even close to the final research model. With hindsight, I could have done a few things different in terms of approach which would have saved me time (but still yielded the same results). I once again learnt a great deal working on it and I think that is very important.

I would like to thank a few people who helped me throughout the process. First of all, I’d like to thank my supervisor DongBack Seo for the time she put into this research. She was always open for a debate and I could reach her whenever it was needed. She provided direction, but also a great amount of freedom which I greatly appreciate. Second, I would like to thank my family and friends for their support during my work. I could always talk, discuss or vent frustration if needed. I specifically want to thank Ilse for her down to earth views and for helping me make some of the most difficult decisions throughout this thesis. Third, I would like to thank everybody who made an effort to gather survey respondents and spreading the survey to friends and or relatives. I struggled finding enough respondents initially and I could not have done this without your help. Finally, I want to thank you for taking time to read this thesis. I hope it will further contribute to your understanding on the topics as well as make a contribution towards already existing literature on it.

4

T

ABLE OF

C

ONTENTS

1.3 Contribution towards exiting research ... 7

1.4 Important definitions ... 8 1.5 Research Outline ... 8 2 Theory ... 9 2.1 Related research ... 9 2.1.1 (Micro-)Blogs ... 9 2.1.2 Literature on Influence ... 12 2.1.3 Literature on Popularity ... 17

2.1.4 Conclusion on Influence and Popularity research so far... 19

3 Research model ... 20

3.1 Model Variables ... 20

3.2 Model and Hypotheses overview ... 28

4 Methodology ... 29

5 Data Collection ... 32

6 Results ... 36

7 Discussion ... 42

7.1.1 Discussion and further analysis ... 42

8 Conclusion ... 46

8.1.1 Limitations and further research ... 47

References ... 48

5

1 I

NTRODUCTION

In recent years web 2.0 tools and social media grew massively in terms of users. Some of these are (social) networks, such as Facebook or LinkedIn. Others, such as blogs, allow users to share their views on popular subjects. One of the well-known blogs around the world is Twitter. Only founded it 2006, it can now be found in the top ten most visited websites on the World Wide Web (Top 500 sites on the web, 2013). Today, blogs are more used than ever before.

A specific form of blogging is Micro-blogging, which grew massively in popularity over recent years. It differs from Blogs because the updates or entries are much smaller in terms of number of words, often just one or two sentences, whereas rest of the functionality is similar to blogs. Readers can, for example, share, react and subscribe to the micro-blog. Micro-blogging systems such as Twitter and Tumblr are also known to be a valuable tool for marketing, internal communication, and online word of mouth (Burton & Soboleva, 2011).

On Twitter, each blogger has his own profile which consists of a photo or image and a short description on the user, often stating job or profession. The profile page displays all entries a user has made in reversed chronological order, including retweets by the blogger. A retweet is a message from someone else shared on the blogger’s own profile. Retweets are often the results of another blogger agreeing with a blogger therefore deciding to share the message. Each blogger can follow other bloggers (in which case they become a follower). When a blogger follows someone else, posts/entries from these other bloggers are automatically received and displayed. Messages from bloggers are accessible to all followers and most often also to non-followers (as it is possible to only display Tweets to people who follow).

6

1.1 I

1.1.1 Different views

A single view on the matter of popularity and influence of bloggers (or people) does not exist in academic literature and various definitions of both terms exist, often striking with one another. Almost all available research on the two topics is either just on popularity (very rare) or on influence. Literature suggests popularity and influence to be the same, or similar, sometimes explicitly, sometimes implicitly. Although the two are unquestionably at least partly related to each other it seems highly unlikely that they are exactly the same. Despite this research taking just blogs into account, the role of popularity and influence goes beyond that in for example product purchase decisions and websites (i.e. what makes a website popular or influential). An example that illustrates the possible differences between popularity and influence is a tourist visiting a restaurant in an unknown area. The tourist may enter a restaurant if it looks great. As such, visitors (or part of the restaurant’s popularity) are the results of the looks of the restaurant. The looks however, are not related to the quality of the food and rarely will a tourist recommend this restaurant to someone else because of its looks. The restaurant’s influence is a result of the quality rather than its looks. Visitors who like the quality are more likely to recommend it on the basis of the quality of the food than on the basis of the restaurant’s looks. With no clear views in academic literature, this example suggests that popularity and influence – although related – are different concepts.

This research aims to clarify the relation between popularity and influence and researches how popularity or influence of a blogger is established. The contribution towards existing research is further elaborated in section 3.

1.2 P

The impact of various variables on popularity and influence of online micro-blogs is investigated. As such, the research questions are formulated as follows:

1. What are unconventional factors determining a person’s online micro-blog to be influential?

7

1.3 C

The research aims to provide a greater understanding on popularity, influence and micro-blogs and therefore contributes towards existing literature in three different ways:

1. To increase understanding and insight in the relationship between influence and popularity.

2. To increase understanding in the variables that impact either or both in the area of (micro) blogs.

3. To increase knowledge in the area of micro-blogging, in which only limited research was done so far.

The first aim is to provide a greater insight in the relation between influence and popularity by researching their relationship in the context of a micro-blog. Available research on the relation between influence and popularity is limited at this moment, and the relationship as presented here has not been researched yet. Popularity has received little attention, if any, and was assumed to be the (almost) identical to influence. The results will indicate whether that point of view is justified, and – provided they are different – to what extent popularity is different and its position in terms of importance compared to influence.

Because this research uses mostly unconventional variables to identify the relation it will also provide insight in what leads to popularity and what leads to influence. If the outcome of this research were to be that both influence and popularity can be regarded as identical or almost identical outcomes, this research still contributes by having established a clear relation and having researched variables on which little or no research was done before.

Finally, this research contributes towards the available research in the area of blogging and micro-blogging. As mentioned before, the research in this area is limited which is partly down to the recent introduction of blogs (first available research in 2001) and micro-blogging where available research is even more limited. By researching the connection between various variables and their impact on popularity or influence of a person’s blog, this research will provide greater insight in the way blogs and micro-blogs work.

8

This research focuses on the micro-blogging system Twitter for a number of reasons. First, Twitter is one of the most used Web 2.0 tools nowadays in which people actively share their views on subjects (Garoufallou & Charitopoulou, 2011). It has over 200 million daily users and over 400 million blog entries on a daily basis. Second, although blogs are relatively new, micro-blogging is a growing trend of only the past few years. As such, limited research on micro-blogging is available. Third, because micro-blogging is much more used than normal blogs it could facilitate this research by ensuring enough respondents.

1.4 I

Blog Websites with information posted in reversed chronological entries, similar to a diary but then online, sometimes including multimedia.

Micro-blog Specific form of blogging with the same properties yet very small messages in terms of words or characters often just one or two sentences.

Twitter Largest micro-blogging service. Twitter user A person with an account on Twitter.

Tweet A Twitter message which can only contain 140 characters, but does allow for sharing of links, videos, and images.

Retweet Shared message originally posted by another Twitter user. Follower A person receiving updates on the posts made by the particular

blogger they follow.

1.5 R

O

9

2 T

HEORY

2.1 R

This section will present a literature review on research that has been done so far in the area of micro-blogs, popularity and influence. Research on normal blogs has also been reviewed because they are almost exactly the same, and the amount of research available in this area is bigger. Articles were accessed through the online database of EBSCOhost Business Source Premier and various search terms were used in order to gain the right articles. This section ends with a small conclusion, providing a short overview on the definition and relation of and between popularity and influence in academic literature. A more detailed description on the used methods is presented in chapter 4.

2.1.1 (Micro-)Blogs

Blogs can be used for various purposes such as conducting conversations, providing consumer advice, delivering news, spreading a message and raising important issues. As such blogs are not just used by individuals but businesses also use them and it can improve customer loyalty, satisfaction and increases sales (Hsu & Tsou, 2011). As such, blogs or bloggers are able to have a certain level of influence on its readers. This relationship is affected by blog credibility and the effect is stronger as blog involvement is higher. Jin and Fisher Liu define two types of blogs (Jin & Fisher Liu, 2010). Note that his distinction can also be recognized on Twitter which is used by both individuals and organizations.

1. Official organizational blogs: created, affiliated and run by organizations

2. External blogs: created and run by third parties (such as individuals or groups outside an organization)

The power of an official organizational blog has been illustrated by Dell, the company that became one of the front-runners in the (organizational) blogging area with their Direct2Dell blog, which is mainly used for customer interaction (Bernoff & Li, 2008). This blog allowed customers to directly contact Dell and what Dell did very well was that they actually admitted problems if there were any. Dell’s blog became highly used and still is as of today. Having originated in 1999, the number of blogs and blog readers has grown massively since then. Blogs can be used to support (group) decision making and as such, they play a role in customer purchase decisions (Saxena, 2011).

10

2010). From a business perspective, this is particularly interesting if such blogs are written by customers about a company, more specifically if these blogs are negative. There are several examples of people who were dissatisfied with the way they had been treated and because they could not sort the problems with the company directly, they decided to bring it into the open world by creating a website or a blog which included their story. Some of these became so popular in terms of visits that the company directly contacted the blogger to sort out the problem (Jin & Fisher Liu, 2010).

A research of Nielsen recently showed that consumer opinions online (blogs) were considered to be the second most reliable source of information placing them above newspaper articles, any form of advertisement and e-mails a person had signed up for (Nielsen, 2012). This finding is in line with earlier results from academic literature (Brown, Broderick, & Lee, 2007) (Jin & Fisher Liu, 2010). Readers are more likely to believe individual bloggers than official organizational blogs and therefore, firms should be wary about influential individual bloggers (Sepp, Liljander, & Gummerus, 2011). The reason why corporate blogs struggle is down to transparency and credibility. Furthermore, individual bloggers usually do not have commercial motivations for writing their blogs. Several strategies and factors that can aid companies in overcoming these issues in their corporate blogs. The strategies aim to increase credibility and transparency as that is the key reason for organizational blog struggles (Chua, Robertson, Parackal, & Deans, 2012). A different form of external blogs are the blogs written by employees about the company they work for. These blogs are not official company blogs as they are not posted on the company’s website and therefore may not necessarily share the company’s views. These blogs can include positive and negative views and companies often react differently towards negative posts. Negative blogs can still lead to positive outcomes because they attract a large number of viewers. Since employees write largely positive messages it is enough to overcome the negative posts. Frameworks have been developed to aid companies in selecting the appropriate strategy towards negative posts (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012). It is advised that negative posts should not be restrained because they can increase readership.

Blogging behaviour of a person can be related to network externalities, other persons within a blogger’s network. Higher usage of blogs within a person’s network is shown to increase own use of blogs. Positive reactions lead to increased usage and people whose manager is following their blog are more active (Wattal, Racherla, & Mandviwalla, 2010). As such, blogs and its environment cannot be treated separately.

11

influence, the need for a strategy or common principles in blogging practise is important (Yang & Lim, 2009). Some critical features in blogging are narrative structure, dialogical self, blogger credibility and interactivity. These features eventually lead to relational trust as a result of blog managed public relations. The aforementioned example of Dell’s corporate blog further illustrates this point.

12

2.1.2 Literature on Influence

Academic research has mainly focused on the role of influence rather than popularity. In other research, contradicting definitions are used and sometimes both terms are being used interchangeably. Research of Aggerwal et al. represents one of the rare cases both considering popularity and influence of a blog. The relation presented is that negative or positive posts (content) have an impact on blog readership. The measurement for influence used in this research is the sum of all weighted citations. A blog post with more citations is generally considered to be more influential, although weight is given depending on who is citing (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012). A random blogger is likely to be less influential than a big newspaper citing the blog in one of their articles. This definition (number of citations) has been applied to both influence and popularity in academic literature. Furthermore, the number of comments on a given blog post can also be used as a measurement to determine blogger influence where a post with more comments has more influence on its readers (Lin & Kao, 2010). A well-known example of influence is the ride of Paul Revere during the American Revolution in 1775 (Gladwell, 2000). At night, he rode through several villages to warn the people of an upcoming attack of the British, enabling the Americans to prepare for an attack. This allowed the entire region to mobilize for the upcoming fight. Clearly, this was the result him being able to transfer a message and influencing behaviour of others. Yet it was only started by a single person. At the same time, another rider took the same message to a different side of the country, and his ride did not alert a region.

Individual blogs are found to be more influential when it comes to purchase decisions than organizational blogs (Sepp, Liljander, & Gummerus, 2011). This can be explained by the commercial interest organizations have, in contrast to most individual bloggers. Findings were done based on 12 interviews. The blogger’s influence was determined by his or her position within the social network around the blog, where people in central positions become opinion leaders (this is addressed later in this section) and therefore have more influence on decisions of others. The position of the blogger within the blogging network was also shortly discussed in the previous section. Influential bloggers tend to increase traffic to their blog as well, making the blog interesting for marketers. Credibility of the blogger and level of involvement determines influence of bloggers (Jin & Fisher Liu, 2010). This is in line with other research on influential bloggers. Jin and Fisher Liu also consider influential bloggers to be opinion leaders.

13

experts or hubs in a blogger network are popular bloggers (Chau & Xu, 2012), which shows that no consistent view on influence and popularity exists in academic literature.

Users within social online communities usually have more than one hundred connections, yet their site usage behaviour is only influenced by a select few. Several models have been developed to identify these influential users within such networks because they can be extremely powerful (Trusov, Bodapati, & Bucklin, 2010). From a marketing perspective, these users are very interesting to target because they can influence others (and for example start a Word-of-Mouth campaign). Furthermore, they can also increase revenue and they are more likely to participate in customer referral programs for bringing in new customers (Walsh & Elsner, 2012). This was measured by identifying market mavens by a six item market maven scale, used for all survey respondents.

The general supported notion in literature is that people with more linkages are more influential (Goldenberg, Han, Lehman, & Hong, 2009). Other research found no correlation between the number of contacts and the influential power and even recorded the opposite effect: influential power reduced with the number of connections (Katona, Zubcsek, & Sarvary, 2011). These opposing outcomes again hint that influence should be treated carefully, but also requires more research to establish its actual position. This was also supported by research on influential users in internet communities which showed contradicting results: where it was commonly accepted that a few highly connected people generated influence, moderately connected people are just as willing to share marketing messages. As such, influence on the internet could work different from what traditional influence models have always assumed (Smith, Coyle, Lightfoot, & Scott, 2007).

14

impacts sales as positive product ratings lead to increased sales (Moe & Trusov, 2011). Sales impact was measured in order to establish this relation.

Further research on influential users found that adopters can help implement viral marketing strategies. This also means that finding influential users is very important in getting a message across. Google has recently patented an algorithm that is able to find influential users within an online network. An adopter is a user who already adopted a new product or service. A user connected to a lot of adopters is more likely to adopt himself as well. Furthermore, density within the group of already adopted users has a positive adoption effect of individuals connected to the group. The position within the network as well as some demographic options determine whether a person is likely to adopt or not. Furthermore, the same applies for influential people. Models like these can help in predicting the adoption process and as such, reveals target groups for marketing (Katona, Zubcsek, & Sarvary, 2011).

One of the most researched variables in the area of influence is trust and it has been researched in various areas varying from blogs to websites and web stores. Trust plays an important role in online web shops as trust causes an increased intention to purchase and also higher actual buying decisions (Kim & Benbasat, 2006) (Kim, Ferrin, & Rao, 2009). Because the internet does not represent a physical relation, trust plays an even bigger role on the internet, compared to traditional shops. Three aspects increase trusting beliefs: web experience, personal innovativeness and website quality.

15

the role of recommendations in blogs has not been researched at a similar level. As such, this concept will return in the research model of this paper.

Numerous research papers show that website characteristics is another important factor in the urge of consumer to buy a product. For example, impulsive buying behaviour is moderated by both website characteristics (e.g. ease of navigating) as well as visual characteristics (e.g. appeal) (Parboteeah, Valacich, & Wells, 2009) (Webster & Ahuja, 2006). The importance of visual appeal was further confirmed in other research as the initial emotional response of a user (e.g. evoked by visual appeal) is important in behaviour towards that website (Deng & Poole, 2010). This was measured experiments and how changes in websites affected user behaviour. Website design plays an important role in the perceived quality of an e-business. Increasing quality in return increases consumer’s willingness towards interaction with the business, illustrating the importance of website design (Gregg & Walczak, 2008). This was measured by letting users assign a positive or negative verdict on a website and consequently review their bidding behaviour in online auctions. Characteristics of a website, or quality can lead to influence suggesting the same may apply to blogs. This hypothesis is tested in the research model.

Below is an overview of the reviewed literature on influence, presented in tabular format:

Table 1: Overview of literature on influence

Finding and measurement (if applicable) Literature source

Influential bloggers are those in a central position in the blogging network. They become opinion leaders. Individual bloggers are more powerful than organizational blogs because they are not commercially driven. Central people influence adoption of new products. Influence depends on blogger position in the network. Influential bloggers are identified based on their position in the network around the blog. The more central, the more influence the blogger has. Measured by number of adoptions. Measurements can be done based on number of incoming and outgoing links too. Sometimes measurements are hard and a survey can be used. Research of Walsh and Elsner uses a Market Maven scale.

16 Influence increases if both information authority and credibility of the blogger are high. Influence is defined by the level of issue involvement and communication activities in terms of message production and consumption.

(Jin & Fisher Liu, 2010)

Characteristics of visited website influences buying behaviour. Often measured by surveys.

(Gregg & Walczak, 2008) (Deng & Poole, 2010) (Gefen, Karahanna, & Straub, 2003) (Parboteeah, Valacich, & Wells, 2009) (Hong, Thong, & Tam, 2004) Trust influences customer (buying) behaviour.

Often measured by surveys.

(Gefen, Karahanna, & Straub, 2003) (Liu & Goodhue, 2012) (Kim & Benbasat, 2006) (Kim, Ferrin, & Rao, 2009) (Doyle, Heslop, Ramirez, & Cray, 2012) (Chai & Kim, 2010)

Recommendations of others influences customer (buying) behaviour). Often measured by surveys.

(Sia, Lim, Leung, Lee, & Huang, 2009) (Kumar & Benbasat, 2006) (Pavlou & Dimoka, 2006) (Moe & Trusov, 2011) (Litvin, Goldsmith, & Pan, 2008)

Frequently cited posts are likely to have a bigger influence on readers than those cited less frequently. This also applies for Twitter. Citations were weighted because some instances or people citing may be more important than others. Influence is measured by number of weighted citations.

17

2.1.3 Literature on Popularity

The available research on popularity is much more limited compared to influence, and popularity is more considered something that is “taken for granted”. It is often identified in simple and objective statistic measures such as number of purchases or number of visits. The relationship between popularity and influence as presented in this research has not been research before. Some research has, however, taken both into account assuming a relationship between one another. For example, research of Aggerwal et al. suggested to define popularity in terms of subscribers and measured influence by number of readers.

Academic literature barely mentions popularity as an end-state, whereas it does play a role in blogging and micro-blogging. One could argue that a blog, or a Twitter user with a lot of followers is not necessarily a distributor or facilitator of influential messages, whereas this also seems to apply the other way around: a blog or Twitter account with a low number of followers is not necessarily be a distributor of uninfluential messages. Generally speaking, these statements are not in line with what has been presented in literature. One could also argue that the power of the message increases with the number of followers or readers (as has been done in literature). This would also be consistent with the explanation of influential people as hubs (Goldenberg, Han, Lehman, & Hong, 2009). Goldenberg et al. found that people with more linkages (defined influential) were able to better influence adoption (popularity) of products. Blog popularity is likely to affect the readership of it and also means an increase in page views (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012).

Blog popularity has also been defined as the extent to which it is able to influence others, which means a highly popular blog is also a more popular blog (Huang, Chou, & Lin, 2010). This view is consistent with the view that popular blogs are those with more citations or links, which in literature has been illustrated as influential blogs. This research takes a completely different perspective by linking popularity with influence by stating that a blog cannot be popular if it is not influential essentially states that both can be regarded as the same or very similar.

18

a message more influential. This need not necessarily be a message from a blog. However, limited research has been done on what makes a blog popular (Liu, Tsai, & Chiu, 2011).

Another view on popularity of trending topics takes the reader’s click rate into account, as calculated in the accumulated click times (Liu, Tsai, & Chiu, 2011). Popularity can quickly give a trending status to an issue. If the notion, that more popular blogs are also more influential, is applied to a micro-blog, the most popular person (the person with the largest amount of followers) is also the most influential person. Furthermore, the most popular (trending) topics on Twitter would also be most influential. That seems questionable at least.

Popularity can be related to the number of people visiting (Webster & Ahuja, 2006). This measurement was used in research on web navigation systems. The number of blog visitors is relatively easy to identify as an objective measurement and blog service providers will have access to this statistic. However, a blog page or profile may be opened without the content actually being read. Since number of people visiting is not always a representative number, popularity is sometimes defined as whether the blog is well known and well read, and people keep track of new material (Sepp, Liljander, & Gummerus, 2011) (Chau & Xu, 2012) (Lin & Chang, 2012). In other words, an active blog is not necessarily a popular blog (Armstrong & McAdams, 2009). Research of Armstrong and McAdams uses the website Popdex as a tool to measure blog popularity. This site was however sold and is no longer in use. Popdex used an algorithm to identify blog popularity. This algorithm was built on the basis that people reading the blog post would be inspired by it and therefore link to it (Tramell & Keshelashvili, 2005). This, once again, suggests ambiguity among the terms of popularity and influence as the most popular blog is the one that is visited and linked to most of all. By linking to a specific blog the blog entry is spread, therefore spreading the message. This in return means an increase in blog influence, often determined by number of citations (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012). Attributes of the blogger such as writing style play a role in determining blogger popularity, which was shown by a decline in visitors when a popular blogger was replaced by guest authors because he left for a vacation (Trusov, Bodapati, & Bucklin, 2010). This is similar to website characteristics influencing customer buying behaviour in the influence literature section.

19

available already, it was left out of the investigation. Research on other factors related to popularity is limited down to the lack of research on popularity.

One finding that again suggests overlapping definitions is that website design has an impact on a visitor’s intention to revisit (Palmer, 2002). Website design was also found to play a role in influencing consumer or visitor (buying) behaviour as discussed in the previous section. Recall that popularity has been defined as the number of visits. As such, website design seems to play a role for influence and popularity by given definitions.

Existing research has often taken the popularity and influence into one view, also when it comes to blogs. For example, it was assumed that a popular blog was cited more often on other webpages, and that a blog cited a lot would be a popular blog (Lin & Kao, 2010). The same applies for influence (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012). A blog cited more often has greater influence than those cited less; the more a post is cited, the more importance this post is. Both views show that popularity and influence are looked at in the same regard, and both are related to number of citations. The existing confusion among both terms is confirmed again. Below is an overview of reviewed literature on popularity, presented in tabular format:

Table 2: Overview of literature on popularity

Finding and measurement (if applicable) Literature source

A popular blog is a blog that is well known and well read. Measured in research among 12 blogs from Estonia. Data gathered with interviews. Researched by looking into XML data

(Sepp, Liljander, & Gummerus, 2011) (Aggerwal, Gopal, Sankaranarayanan, & Vir Singh, 2012)

Popularity can be defined as an objective measurement in terms of site visits. Sometimes depends on blogger characteristics. Consistent with (Palmer, 2002), who stated that website design determines intention to revisit.

(Webster & Ahuja, 2006) (Moe & Trusov, 2011) (Armstrong & McAdams, 2009) (Lee, Im, & Taylor, 2008) (Liu, Tsai, & Chiu, 2011) Popularity can be defined as the number of citations

or number of people linking to the blog. Popularity is therefore defined in level of influence by the blog. People discussing the issue is also a sign of

popularity. In

(Tramell & Keshelashvili, 2005) (Huang, Chou, & Lin, 2010) (Lin & Kao, 2010) (Liu, Tsai, & Chiu, 2011)

Factors influencing popularity are trust and blogger

characteristics (Palmer, 2002)

2.1.4 Conclusion on Influence and Popularity research so far

20

3 R

ESEARCH MODEL

The research model is explained below. First, the variables and their definitions are introduced. Each variable is followed by its hypotheses on the final role in the model. The research model including hypotheses is also shown in a figure at the end of this section.

3.1 M

V

Influence

The supposedly existing link between influence and popularity (of being similar) comes back in some definitions where influential people are considered those with a large number of linkages. Influential people are able to influence decisions or behaviour of others and they are therefore considered powerful. Bringing this back to micro-blogging means that influential people have an impact on those reading the micro-blog messages, therefore having an impact on decisions or behaviour of readers or followers. Influential people or blogs also lead others to spread their message, which in case of Twitter can be done by retweeting the message and thereby spreading the message throughout the Twitter network. Together with popularity, influence is a key variable in this research and in the model it is presented as an end state.

Popularity

Popularity is often measured based on numbers, for example the number of followers in the case of a micro-blog. Based on literature, a more popular person on a micro-blog would have more followers. Number of cites, in the case of Twitter retweets, is also named as a popularity measurement. However, retweeting a message leads to the message being spread which in return could lead to influence. Therefore, popularity will be measured by the number of visits, and whether the message is actually read. A simple measurement of number of visits is insufficient if the message is not read. It is therefore important to measure both number of visits and whether the messages are read. Intention to revisit is another popularity measurement, taken from the subscription on blogs. People subscribing to a specific blog intend to revisit it. Popularity, like influence, is an end state in the research model.

Heuristic and Systemic Information Processing

21

based on secondary cues (Petty, Cacioppo, & Schumann, 1983). An example of this is a person judging a message based on the expertise of the source (credibility) rather than the quality of the arguments in the message. These two concepts form key elements in the research model as one route is about core concepts, whereas the other is about the surroundings.

The research from Petty et al. discovered that persuasive messages on a topic with a high personal relevance follows a central route to attitude change, a thoughtful consideration of the arguments in the message. Persuasive messages on a topic with low personal relevance follows a peripheral route to attitude change (Petty, Cacioppo, & Schumann, 1983).

Chaiken refers to the central route towards persuasion as systemic information processing (Chaiken S. , 1980). Systemic information processing involves a significant cognitive effort in order to evaluate a message and the decision on whether to accept the message conclusion. People subject to a message with high personal involvement attach no value to communicator likeability, but they do to the number and quality of the arguments.

Chaiken refers pherical route to persuasion as heuristic information processing. Heuristic information processing, opposed to systemic information processing, requires less cognitive effort because people rely to information that is easier to access in order to decide whether a message conclusion is accepted or not. Opposed to systemic information processing, people subject to a message with low personal involvement attach value to communicator likeability, but do not attach value to the number and quality of the arguments.

22

These two concepts can help explain the possible difference between influence and popularity. A person may have a lot of followers because of giving away prizes, which suggests that people make the decision to follow based on the surroundings (the option to win a prize), rather than the quality of the message itself. On the other hand, a blogger providing high quality information is likely to have followers as a result of these high quality messages. Recall the example on restaurant choice provided in the introduction of the paper. One tourist may enter a restaurant because it has nice looks, a local may enter it because of the quality of the food. The tourist displays heuristic information processing whereas the local displays systemic information processing. The differences between popularity and influence may be explained by differences in the way information is processed. As such, all variables are related to either heuristic information processing or systemic information processing.

As mentioned, popularity is often related to the number of visits, or the number of people following. The example of gaining followers by giving away prizes, mentioned in the paragraph before, shows that popularity may be the result of surroundings rather than the core message. On the other hand, people retweeting a message do so because they agree with its content, suggesting they thought about the message, considered its content and quality and then made a decision on whether to follow or not. These two reasons to follow a person hint on a different process behind the decision to follow. The two presented different routes of information processing could explain the process behind a decision to follow as well as whether it predicts popularity or influence. Based on prior research, it is expected that both ways of information processing are related to popularity and influence as it is impossible to consider the two completely loose. Also, some variables may lead to greater popularity while also impacting influence and vice versa. This cross relation will also be tested. Therefore the following hypotheses are made:

H1: Heuristic information processing is positively related to popularity H2: Systemic information processing is positively related to influence

H3: Heuristic information processing is positively related to influence, but to a much lesser extent than popularity

23

Credibility

Credibility refers to the belief a reader has about the author and the author’s message and to the extent the reader believes this message is true. When an author expresses certainty in his message, it is more likely to have influence. However, a negative relation between high expertise authors and the level of expressed certainty in their message exists (Karmarkar & Tormola, 2010). Certainty of a message plays an important role in bringing a message across as people often forget to transfer the certainty of a message along with it, when passing it onto someone else. It explains why rumours turns into perceived facts and vice versa (Dubois, Rucker, & Tormola, 2011). Source credibility has a positive persuasive effect provided the person has a positive attitude towards the source. For example, a person using a particular brand will find advertisements by this brand to be appealing, and it may even lead to higher positive attitudes towards this brand (Tormola & Petty, 2004). The opposite applies as well (if a negative attitude towards the source was displayed, such a message could decrease the persuasive effect).

Customers with positive attitudes towards the brand are more positive and confident towards claims on brands products (Erdem & Swait, Brand Equity as a Signaling Phenomenon, 1998). Credible sources lead to persuasive outcomes (Harmon & Coney, 1982) which again suggests that popularity and influence are in fact the same. In this paper, it is hypothesized that someone following or reading a message because of the credibility of the source is taking the heuristic approach because the consideration is about a blogger’s attribute rather than the content of the message.

24

Social connectedness

Similarity between two people, such as similar attitudes, or more coincidental similarity such as a shared birthplace or birthday can have an impact on the behaviour of these people. For example, incidental similarity between customer and salesperson increases buying intention and results in a more positive attitude towards a product (Jiang, Hoegg, Dahl, & Chattopadhyay, 2010). Similarity between a person and an organization can lead to higher job satisfaction and commitment while also decreasing employee turnover (Piasentin & Chapman, 2007). People are also more inclined to comply to a request if such a shared similarity (such as a birthday or name) occurs (Burger, Messian, Patel, del Prado, & Anderson, 2004). The concept of social connectedness can return in social media or on blogs. Every blogger on Twitter has his/her own profile with information about this person for display at first sight. As the feeling of a connection towards the blogger is about a blogger or blogger attributes rather than the message itself, it is hypothesized that social connectedness leads to a heuristic way of processing information

H6: Social connectedness is positively related to heuristic information processing

Positive events

An often used promotional tool is the use of contests or sweepstakes which allows a user to win a prize by performing an action, for example by following a Twitter account or Retweeting a (marketing) message thereby spreading that specific message in the Twitter network. Prizes are often determined beforehand and winners are randomly selected. A large part of the companies use contests or sweepstakes as part of their marketing efforts to increase sales (Kalra & Shi, 2010). Unexpected surprise or rewards have positive results (Valenzuela, Mellers, & Strebel, 2010). Positive events that involve uncertainty, such as the chance to win a prize when performing a certain action, will be almost as effective as the best possible situation with certainty. In other words, offering people a chance to win a prize will almost be as effective as offering the prize to any of the people that perform the required action (Goldsmith & Amir, 2010). A variable influencing this relation is the risk a person has to take in order to be able to win a prize. However, in the case of Twitter there is virtually no risk as following a user or retweeting a message is a very simple task. Interestingly, in some cases deal proneness of customers lead to word-of-mouth thereby influencing others to also participate (Wirtz & Chew, 2002). Because positive events are not related to the quality of the message but rather to the surroundings as a user participates because of the chance to win a prize it is hypothesized that positive events lead to heuristic information processing.

25

Referrals

A referral is someone who was brought into something (e.g. a company, web shop, website, blog) through another person. A lot of companies have referral programs as a ways of acquiring new customers. An existing customer often receives a bonus or discount for bringing in new customers making referral programs a great promotional tool. They key of such programs is that existing customers bring in new customers. This can also be translated to Twitter in the broader sense, where people create an account because their friends also have account or follow a blogger just because their friends do, or because friends retweet the message. Furthermore, Twitter also shows a list of suggested users to follow based on the users a person is currently following and the interests/area of that person which could lead to new followers. The referral system is closely related to positive events (for example when a prize is awarded for following a particular user), and also to recommendations of others. But they are different: in a referral system the persons referring receives the (chance to win) prize. Also, referrals is about doing an action because of someone else (e.g. because a friend follows). As such, the message itself it not heavily considered and therefore likely to follow a heuristic way of processing information.

H8: Referrals is positively related to heuristic information processing.

Loyalty

Loyalty is something often visible in various product ranges and categories: customers stick with a particular brand for a diverse set of reasons but most likely satisfaction. If a customer is satisfied with the quality of a particular item this is likely to lead to satisfaction which in return leads to loyalty (Chang, Wang, & Yang, 2009). This could also be a reason for continuously following a blogger. Online loyalty, also referred to as e-loyalty is influenced by perceived value, which in return is the result of service quality (Fuentes-Blasco, Saura, Berenguer-Contrí, & Moliner-Velázquez, 2010). Although not hypothesized, a relationship between loyalty and satisfaction is expected because loyalty is often the result of satisfaction. Research on this relation has, however, shown that no conclusive remarks can be made on this relation as it varies among different situations. Loyalty to a product or brand may not always result in the best available outcome, other products or brands could be better in terms of quality. Loyalty could, therefore, be a reason for popularity, but not necessarily for influence. The messages itself does not really matter in terms of quality as long as they are satisfactory meaning that loyalty leads more to heuristic than systemic information processing. As such, the following is hypothesized:

26

Product effectiveness (Perceived Competence)

A customer’s belief of product effectiveness influences their buying behaviour. In general, a product that is perceived to be more effective will be used more (Lin & Chang, 2012). For example, a toothpaste showing a picture of white teeth can lead to higher perceived product effectiveness (Zhu, Billeter, & Inman, 2012). In relation to blogs, Twitter has no products posting, but people and other instances such as companies and associations. A term suiting this research variable better is perceived user competence. Competence refers to the ability of a person to do what the other person needs (McKnight, Choudhury, & Kacmar, 2002). Translated to blogs, competence refers to the ability of a blogger to post what the reader needs. If a person perceives the poster as a more competent and effective provider of information, this could have an impact on either popularity or influence. Higher perceived effectiveness or competence in providing information leads to higher usage. Because the effectiveness judgement is made on the basis of a post, the core message is evaluated rather than its surroundings. This leads to the following hypothesis:

H10: Product effectiveness is positively related to systemic information processing

Information effectiveness

Unlike product effectiveness or perceived competence of the user posting a Twitter message, information effectiveness is not about the user posting the message but about the intrinsic properties of the message itself, the perceived quality of the information. The quality of the information refers to the perception of accuracy and completeness of the information in a message (Kim, Steinfield, & Lai, 2008). This is relevant because the quality of the information available varies, especially on the internet where a lot of information is accessible right away. Part of the information may not be accurate as rumours can turn into facts once a message passes from one person to another, and facts may turn into rumours. High information quality can be a reason to follow another user and as this judgement is based on the intrinsic properties of the message, its quality. As such, a user taking information effectiveness into account is likely to follow a systemic way of processing information.

27

Recommendations

Recommendations of others have an impact on buying decision and purchasing behaviour in the broadest sense. For example, positive third-party product reviews in the movie industry increase stock returns (Chen, Liu, & Zhang, 2011) and online reviews can positively impact consumer buying behaviour, determined by the number of reviews and the quality of the review (Park, Lee, & Han, 2007). Online customers often base their buying decision on recommendations of others, which illustrates the role recommendations can have when making decisions (Smith, Menon, & Sivakumar, 2005). Use of blogs may, however, remove this effect as readers attribute positive or negative reviews to circumstances (Lee & Youn, 2009). Twitter has recommendations based on the people a blogger follows. Relatives, friends and family may lead to new followers, but retweets from other people (friends, family) can also influence people to follow a new blogger. It is not likely that a person would just follow someone at the moment the message is retweeted or the moment it is recommended by a friend or family. Contrary, the decision of whether or not to follow will be made based on the message posted by the blogger and therefore likely to follow a systemic way of processing information.

H12: Recommendations is positively related to system information processing

(Product) superstar

Product superstars are products that possess special attributes that give them an extraordinary large market share (Binken & Stremersch, 2009). This superstar phenomenon can be witnessed in several industries. An example of the power of product superstars is visible in this software industry where a single title, or a few single titles can boost hardware sales thereby having a big impact. The principle of product superstars is all about quality. They are superior to the other available alternatives. The same superstar principle, better noted as quality superstar principle, can be argued for other areas, such as newspaper and blogs. Since superstars increase sales by their unique quality, it is expected that Twitter superstars would have a positive relation to the number of people visiting, following, or reading a person’s Twitter. Product superstars (unlike real world superstars) is all about attributes that matter, core concepts and therefore likely to follow a systemic way of processing information.

28

Satisfaction

Satisfaction refers to a judgement made by user after consuming or using a product. This judgement is not a single decision point in time, but the process of reconsideration is made frequently and consists of a comparison between expectations and actual results. User satisfaction predicts continued use of a product or service (Ye, Seo, Desouza, Sangareddy, & Jha, 2008). There are also some indications that satisfaction does not necessarily lead to persuasion or influence (Wirtz & Chew, 2002). Satisfaction is also expected to have a positive relation with loyalty as satisfaction often leads to loyalty (Cyr, 2008). However, that is not a key concept in this research and therefore not hypothesized. Because satisfaction is the continued reconsideration of the message it is argued that satisfaction is more likely to lead to systemic than heuristic information processing. As such, the following is hypothesized:

H14: Satisfaction is positively related to systemic information processing.

3.2 M

H

Below the full research model is presented, based on the hypotheses and variables introduced in section 3.1. The hypotheses are also shown in the figure below. The research model with all variables is shown below:

29

4 M

ETHODOLOGY

The research consists of several stages that were performed in order to get to the conceptual model and the results as presented. Because this research aims to find unconventional variables linking to popularity or influence, the first stage was to research what was known about blogs, popularity and influence.

4.1.1 Literature research on influence and popularity

To complete the first stage, a literature research was performed. The starting point of this literature research was in two top IS journals MIS Quarterly and Information Systems Research (MIS Journal Rankings, 2012). These journals were selected because both cover top research in the field of Information Technology therefore would most likely result in conventional variables leading to popularity, influence or both.

Both journals were accessed through EBSCOhost Business Source Premier. Highwire Press Informs was used to access the last two year of Information System Research. Next to articles from these journals Business Source Premier was used to find articles from other academic journals. In order to find articles outside aforementioned journals the subject terms feature of Business Source Premier was used. This limits the number of articles to only those truly relevant to the searched term. Among others and combinations, the following subject terms were used ‘micro-blog’, ‘influence’, ‘popularity’, ‘multimedia’, ‘ website’, ‘blog (research)’, ‘ website design’, ‘ web 2.0’, ‘trust’ and ‘content’. (For micro-)blogging, a total number of 125 articles was found, of which 16 were used. For influence, a total number of 495 articles was found, of which 27 were used. For popularity a total number of 110 articles was found, of which 14 were used.

4.1.2 Literature research for model development

30

The found variables were judged on whether they would be more likely to impact influence or popularity, based on the articles they were taken from. Several variables covered the same areas but were named differently. These variables were grouped into one, taking the variable that fitted best as a name for the new grouped variable, again taking research results and definitions from the article as a starting point for this approach. As such, articles where the variables were taken from became the theoretical foundation for the conceptual model.

The end result of this stage was a list of 11 variables that could be used for the model. Based on academic literature as well as the literature the variables were taken from, hypotheses could be supported for 10 out of 11 variables. As such, one further variable was dropped from the research model. The ten remaining variables became the independent variables for the research.

The dependent variables were taken from literature that could offer an explanation to the popularity and influence based on research in popularity/influence areas (popularity as click time can be suggested as low personal involvement, influence as the consideration and spreading of the message can be high personal involvement). Finally, the relation between the dependent and independent variables was established based on the literature review on influence and popularity as well as the literature related to the area the variable was taken from.

Access to the journals was gained through the Library in the Faculty of Economics and Business (FEB) at the University of Groningen. The issues from 2012 were not all available in this library and were accessed online instead. Access was gained through EBSCOhost Business Source Premier and JSTOR Current Scholarship Program Complete. All accessed articles in both these stages were taken from academic journals. Furthermore, all articles are peer reviewed to ensure validity and reliability of used articles.

4.1.3 Survey

The conceptual model was researched with a survey among 155 people. The majority of the articles used to create the conceptual model contained survey questions for the variables used. These questions were taken and adapted from those articles. Several variables came from articles which did not use a survey as research model. For those variables, survey questions were taken and adapted from additional literature. This additional literature was searched based on references as well as subject terms, both in Business Source Premier.

31

small adjustments were made before its final launch. The survey was hosted online on the free online survey tool Qualtrics. The survey was spread through e-mail, YouTube, LinkedIn groups, Facebook, and Twitter and several online message boards.

32

5 D

ATA

C

OLLECTION

Below is an overview of the survey and the questions used in the survey. To ensure greater survey reliability questions have been taken from academic literature and were adapted to match the research context. Some questions were not directly taken from literature. In such cases, questions were drawn from previous research (methods) as well as the research methods used in respective articles they were based on. Demography questions were gathered from several of the below mentioned sources. As such, all indirect questions are backed up by academic literature. All questions were asked on a Likert-scale from 1 (strongly disagree) to 7 (strongly agree). The survey is shown in the table below:

Table 3: Overview of survey questions

Variable Questions Literature source

Product effectiveness 1. I believe the person is competent and effective in providing information 2. Overall, the person is a competent and proficient provider of information 3. In general, the person is very knowledgeable about what is going on

(McKnight, Choudhury, & Kacmar, 2002)

Information effectiveness 1. The person provides useful information 2. The person provides reliable information

3. The person provides sufficient (enough) information

(Kim, Steinfield, & Lai, Revisiting the role of web assurance seals in business-to-consumer electronic commerce, 2008)

(Wells, Valacich, & Hess, 2011) Credibility 1. The person’s claims are believable

2. The person has a name one can trust

3. The person reminds me of someone who’s competent and knows what he/she is doing

(Erdem & Swait, Brand Equity as a Signaling Phenomenon, 1998) Social connectedness 1. I share a lot in common with this person

2. I feel distant from this person

3. My personality is well suited for the personality or image of this person

(Lee & Robbins, 1995) (Piasentin & Chapman, 2007) Recommendations 1. Family, friends/relatives of mine have told me positive things about this

person’s Twitter

2. I am likely to listen to what my family, friends/relatives say about use of this person’s Twitter

3. My friends or classmates tend to retweet this person’s Tweets 4. I believe what others tell me about this person’s Twitter

(Schumann, et al., 2010)

(Ye, Seo, Desouza, Sangareddy, & Jha, 2008)

33

Positive events 1. Promotions or giveaways have led me to follow this person.

2. Promotions or giveaways have led me to be a more active follower of this person than I usually would

3. Promotions or giveaways have led me to follow this person sooner than I normally would

(Shi, Cheung, & Prendergast, 2005)

Product superstar 1. The information from this person is more reliable than information from others.

2. The quality of this person’s Twitter messages compares well to those of others

3. I am firmly convinced that this person provides high quality information in his/her Twitter messages

(Li, Wang, & Liu, 2011) (Lages, Silva, & Styles, 2009)

Referrals 1. I decided to follow this person because of the number of retweets by people I follow

2. I decided to follow this person because it was suggested to me by the Twitter website

3. I decided to follow this person because my connections follow this person as well

(Schmitt, Skiera, & Van den Bulte, 2011)

Loyalty 1. Whenever I’m looking for information on Twitter this person is my first choice.

2. To me, this is the best Twitter account around 3. This is my favourite (informational) Twitter account.

(Chang, Wang, & Yang, The impact of e-service quality, customer satisfaction and loyalty: on e-marketing: Moderating effect of perceived value, 2009)

Satisfaction 1. I am satisfied about the Twitter messages from this person 2. I am pleased with the Twitter messages from this person 3. I am happy with the Twitter messages from this person

(Ye, Seo, Desouza, Sangareddy, & Jha, 2008)

Heuristic information processing 1. A person’s expertise is important in my judgement of their Twitter messages 2. The number of retweets is important in my judgement of their Twitter

messages

3. The number of followers is important in my judgement of their Twitter messages

4. The societal position of this person is important in my judgement of their Twitter messages

(Chaiken S. , 1980)

(Petty, Cacioppo, & Schumann, 1983)

(Petty, Cacioppo, & Goldman, 1981)

Systemic information processing 1. When I read a Twitter message of this person I often spend time thinking about it

2. When I read a Twitter message of this person I often consider the quality of his/her arguments

3. When I read a Twitter message of this person I don’t care about the number of followers or retweets

4. I feel personally involved in what this person writes about

(Chaiken S. , 1980)

(Petty, Cacioppo, & Schumann, 1983)

34

Influence 1. I am often affected by the person’s Twitter messages when I think about the relevant issues.

2. I was influenced by the information gathered from the person’s Twitter messages

3. I often retweet this person’s messages because I agree with the messages

(Godes & Mayzlin, 2009) (Park, Lee, & Han, 2007)

Popularity 1. I intend to keep visiting this person’s Twitter account 2. I regularly visit this person’s Twitter

3. I often check this person’s tweets.

35

A total of 155 people participated in the survey from 20 different nationalities. Data was gathered in the months of April, May and June. All data could be used for the research. An overview of the demography results is shown below:

Table 4: Demography Results

Demography Results Gender Male 123 79.4% Female 32 20.6% Age 24 or younger 109 70.3% 25-29 25 16.1% 30-34 5 3.2% 35-39 4 2.6% 40-49 4 2.6% 50 or older 8 5.2%

Highest Education Received

Grammar/Elementary School 6 3.9%

High school or equivalent 33 21.3%

Some college/university 59 38.1% Bachelor’s Degree 31 20.0% Diploma/Master’s Degree 23 14.8% Doctoral Degree 3 1.9% Other 0 0.0% Job

Employed for wages 49 31.6%

Self-employed 12 7.7%

Out of work and looking for work 10 6.5%

Out of work but currently not looking for work 4 2.6%

Homemaker 2 1.3% Student 76 49.0% Military 1 0.6% Retired 1 0.6% Unable to work 0 0.0% Twitter Familiarity Low familiarity 38 24.5%

Familiar with Twitter 66 42.6%

Regular User 41 26.5%

Very active user 10 6.5%

36

6 R

ESULTS

The full results of the survey are shown in the appendix where each measurement is displayed along with its mean, standard deviation and outer loading. In order to ensure reliability and validity of the research all data is analysed by SmartPLS on reliability of the indicators, convergent validity and discriminant validity (Henseler, Ringle, & Sinkovics, 2009).

The outer loading is important because determines the reliability of the indicator. An indicator of 0.5 means the latent variable explains at least 50% of the variance in the indicator. In order to be a reliable indicator, the outer loading is satisfactory if it has a value of at least 0.7 with 0.5 as minimum (Hulland, 1999). All indicators with an outer loading below 0.5 have been removed from the data set. Values above 0.5 and below 0.7 can be removed if it improves the overall reliability of the variable (Henseler, Ringle, & Sinkovics, 2009). As such, two further indicators were removed from the model. A total of three indicators were removed from the data set for low reliability. The results of all questions including mean, standard deviation and outer loadings for each indicator can be found in the appendix.

37

Table 5: Convergent validity of measurements

AVE Composite Reliability Cronbach’s Alpha

Heuristic Information Processing 0.68 0.86 0.76 Influence 0.68 0.86 0.76 Information Effectiveness 0.66 0.85 0.74 Loyalty 0.70 0.87 0.86 Popularity 0.84 0.94 0.91 Positive Events 0.94 0.98 0.97 Product Effectiveness 0.77 0.91 0.85 Recommendations 0.60 0.82 0.68 Referrals 0.62 0.83 0.69 Satisfaction 0.85 0.95 0.91 Social Connectedness 0.66 0.85 0.74 Source Credibility 0.60 0.81 0.78 Superstar 0.63 0.84 0.71 Systemic Information Processing 0.53 0.77 0.56

38

In determining discriminant validity, the Fornell-Larcker criterion is used. The average variance extracted (AVE) of the construct should be greater than the highest squared correlation with any other construct (Fornell & Larcker, 1981). The result is a construct correlation matrix as shown in Table 6. The Fornell-Larcker criterion is met which means discriminant validity is ensured.

Table 6: Construct correlation matrix for discriminant validity

HEUR INFL INEF LOYA POPU POEV PREF RECO REFE SATI SOCO CRED PRSU SYST