Social media : what are the drivers behind the quality of Facebook and Twitter and how do they purchase intentions?

Amsterdam, June 30, 2014

CourseBachelor Thesis Business Studies

NameJorisHoogenboom

Studentnumber10247246

Academic year 2013-2014

Supervisor A. Krawczyk

SOCIAL MEDIA

_{What are the drivers behind thequality of Facebook and}

Abstract

Social media is used extensively these days by organisation because of the many advantages. However, the drivers behind the quality of a social networking website are still not explained in the existing literature. This study focuses on which factors influencethe purchase intentions of the consumers. Expected is that the overall quality of a social networking websites influences the perception of a brand in the eyes of the consumer, which influences the purchase intentions. Hypothesized is that the individual factors of Facebook and Twitter positively influences its quality. The quality of Facebook and Twitter combined influences the brand evaluation. And based on literature, a positive brand evaluation should lead to higher purchase intentions. These propositions are tested with a sample of 126 respondents who are familiar with Facebook and/or Twitter at least. A few elements of social media do have a significant influence on the quality of a social networking website. And evidence is found that quality influences brand evaluation and brand evaluation influences purchase intentions. Managers should take social media seriously because of these relations. Spending time and money to develop and execute an online marketing do have effects on the purchase intentions.

Table of content

1. Introduction

5

2. Literature review

7

2.1 Literature review. . . 7

2.1.1 Social media, social networking websites and current developments. . . 7

2.1.2 The quality of social networking websites . . . 9

2.1.3 Brand evaluation and purchase intentions. . . 10

2.2 Conceptual framework. . . 11

2.2.1 Influential factors of the quality of Facebook and Twitter. . . 12

2.2.2 Brand evaluation. . . .. . . .. . . .. . . .. . . .. . . 13 2.2.3 Purchase intentions. . . 13 2.3 Conclusion. . . 14

3. Methodology

15

3.4.1 Demographic variables and additional information. . . .. . . 17

3.4.2 Regression analysis . . . 17

3.4.3 The final survey construction. . . 19

4. Results

20

5. Discussion

30

5.1 The findings. . . .. . . .. . . 30

5.1.1 Facebook. . . .. . . .. . . .. . . 30

5.1.2 Twitter. . . .. . . .. . . .. . . 31

5.1.3 Brand evaluation and purchase intentions. . . 32

5.1.4 An overview of the results. . . .. . . 33

5.2 Managerial implication. . . .. . . .. . . 33

5.3 Theoretical contributions. . . .. . . 34

5.4 Limitations and suggestions for future research. . . 35

6. Conclusion

37

Appendix A:Survey. . . .. . . .. . . .. . . 40

English. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . 40

Dutch. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . 44

Appendix B:Additional data. . . .. . . .. . . 49

Part 1: Data descriptions. . . .. . . .. . . .. . . .. . . .. . . .. . . 49

Part 2: Correlations. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . 51

Part 3: Reliability statistics. . . .. . . .. . . .. . . .. . . .. . . .. . . 52

Part 4: Regression analysis. . . .. . . .. . . .. . . .. . . .. . . .. . . 55

1. Introduction

Share the status of your favourite artist on Facebook, retweet a tweet from your favourite sports club, boys and girls from fourteen years old are doing it without even thinking about it but their parents are still struggling with adding a friend on Facebook. But one thing is sure, we are all familiar with social media. Facebook achieved in less than ten years to connect more than 1 billion people with each other. In October 2012 they reached the number of 1 billion active users (Facebook.com, 2014). This is about 15 per cent of the world population (Worldbank.org, 2014). Twitter has 241 million of active users every month (Twitter.com, 2014). People update their timeline on Facebook when they wake up, check Twitter when they are stuck in traffic and are searching for videos on YouTube. Many of the users integrate the social networking websites into their daily lives (Wang, Yu & Wei, 2012). It is not surprising that organizations are trying to connect to these users because simply these people are consumers as well. The amount of organizations that are using the web for marketing and promoting increased drastically (Ranganathan & Ganapathy, 2001). Social media enables organizations to contact with the consumers directly (Mangold & Faulds, 2009; Kiang, Raghu & Shang, 1999).

It is safe to say that the relationship between the consumer and organizations have changed since the extensive use of social media in marketing activities. And we already know a lot about social media and about online marketing as well. One thing we know is that consumers became more active. They are responsible for the information available on the internet (Stewart & Pavlou, 2002). The fact that consumers share knowledge with other consumers is changing the marketing landscape (Heinonen, 2011; Mangold et al., 2009). Consumers are able to give feedback of products and

services (Mangold et al., 2009). This user-generated content (UCG) has benefits for marketers. They receive ideas from consumers at a much lower cost compared to traditional channels (Krishnamurthy & Dou, 2008). The possibility to interact with consumers strengthens the relationship between the company and the consumers (Wang et al., 2012). Previous literature describes the potential advantages and possibilities of this change extensively.

We know as well that the social media affects the perceptions of brands. Olson and Mitchell (1981) define brand attitude as a consumer’s overall evaluation of a brand. Brand attitude has a positive effect on brand purchase intentions (Schivinski & Dabrowski, 2014). And according to the theory of planned behaviour by Ajzen (1991), intention will lead to behaviour. Having a positive attitude towards a brand could eventually lead to an increase of sales. It is researched that social media can have a positive effect on the sales volume (Naylor, Lamberton & West, 2012). In the end, this is what the company is trying to accomplish.

However, little is known about the purchase intentions of consumers based on the quality of the social networking websites. In the literature, little can be found about the effect of a well-designed and executed online marketing strategy on the purchase intentions. Does a good strategy lead automatically to an increase in sales or is being active enough? Is the evaluation of a brand influenced by specific elements of social media? Which elements seem to have an impact on the quality of a social networking website? All these questions can be combined in one question and this is the research question of this study: What elements of social media influences the purchase

intentions of the consumer?

This study is trying to give more insight in the quality drivers behind a social media platform. A lot of time and money is spent to develop and execute an online marketing strategy, so it has to be done right. The aim of this research is to identify elements that directly influence the quality of Facebook and Twitter in particular. To provide an answer to the research question, a survey will be held among consumers who are familiar with Facebook and/or Twitter. With the results of this survey, it is intended to identify influential factors which could help managers to develop and optimize their online marketing strategy.

This paper will consist of a literature review to give insight in the current knowledge about social media. This study focuses on Facebook and Twitter. The next step is to develop a conceptual framework and showing the propositions. After that, the methodology is explained, followed by the results of the survey. The interpretation of the results can be found in the discussion and the paper ends with aconclusion.

2. Literature review

2.1 Literature review

In this section, existing literature will be addressed what will help to better understand the current knowledge. Many researchers have written about the concept of social media in the past and this section gives an overview of the useful literature related to the research problem described earlier. First, the concept of social media will be briefly explained as well as the current developments and a description of social networking websites. Second, current knowledge about factors that influence the quality of social networking websites will be given. Then the existing literature about brand evaluation will be discussed. After that, information about purchase intentions of consumers will be further explored. All this information has led to a visual representation of the current knowledge. The conceptual framework will be discussed after we have reviewed the literature. The expectations will be shortly discussed as well. And finally, all the relevant existing literature will be shortly discussed in a conclusion.

2.1.1 Social media, social networking websites and the current developments

Social media how the consumer nowadays knows it, is a relatively new concept. The idea behind social media however is not (Coyle & Vaughn, 2008; Kaplan & Haenlein, 2010). In the existing literature, many definitions of social media can be found and they are all similar. But maybe the easiest thing to do is to look up the word ‘media’ in the dictionary. Media is defined as ways to exchange information (Vandale, 2014). Mangold and Faulds (2009) use a similar definition, they describe social media as consumer-generated media. So in essence, social media is the exchange of information created by the users. This is the definition that will be used in this paper. It is however interesting to take a look at the origins of the phenomena. Kaplan etal. (2010) developed a more extensive definition based on two related concepts. One of those concepts clearly shows the ‘social’ element of social media. This concept is User-Generated Content (UCG), the user is responsible for the content that can be found on the World Wide Web. Individuals share, co-create, discuss and modify the content (Kietzmann, Hermkens, McCarthy & Silvestre, 2011).

Nowadays there is a diverse and rich ecology of social networking websites. Social networking websites are a type of virtual community (Dwyer, Hiltz & Passerini, 2007). Every websites has its own specialities. They vary in their scope and functionality (Kietzmann et al., 2011). By creating a profile, the users can build a personal network and create and share content with the other users (Lenhart & Madden, 2007). The amount of information shared on the profiles on the websites differs and it depends on the main function of the website (Kietzman et al., 2011). As mentioned in the introduction, the focus of this research is on Facebook and Twitter. Descriptions of

these social media can be found in the literature but perhaps the most accurate description is how these two websites describe themselves. On the website of Facebook, the mission statement of the organization is defined as:

‘Founded in 2004, Facebook’s mission is to give people the power to share and make the world more open and connected. People use Facebook to stay connected with friends and family, to discover what’s going on in the world, and to share and express what matters to them.’

– Facebook.com, 2014.

The main function of Facebook is connecting people with friends and family and share content that is relevant for these people. The user creates a profile that is supposed to show his or hers ‘real world’ identify (Keenan &Shiri, 2009). It differs from other social networking web, for example YouTube, where the users can create a profile where they can be whatever they want to be. Kietzmann et al. (2011) mentioned seven building blocks of social media (see Appendix B, figure 1a). The most important building blocks for Facebook are relationships and identity. Facebook is the largest social media with more than 1 billion users (Facebook, 2014) and is therefore valuable for this research. The mission statement of Twitter can be found on the websites as well:

‘To give everyone the power to create and share ideas and information instantly, without barriers.’ – Twitter.com, 2014.

The main function of Twitter is to create and share content as well but the fundamental difference with Facebook is the fact that the users do not necessarily have to share it with people they know. In essence, you can share ideas and information with everyone who wants to follow you. The building block identify is less important but the block ´sharing´ is more relevant for Twitter. It does not really matter who creates the content as long as other users receive the content. The mission statement describes it well, the idea is to create content ‘without barriers’. You might say the message of the Tweet is more important than the user. This difference is very useful for companies to communicate with their customers and build relationships with them (Twitter.com, 2014). For this reason, Twitter is included in this research.

Despite the differences, Facebook and Twitter fulfil similar needs for organizations. It enables them to reach their customers (Mangold et al., 2009; Kiang et al., 1999). The difference between Facebook and Twitter is smaller for organizations than for the population. Organizations do not know the people who read their posts on Facebook and the tweets on Twitter. So in essence, there is no

difference between the two websites. People do know who can read their posts on Facebook and it is possible to protect your tweets on Twitter, so not everyone can access your account.

As mentioned before, the idea behind social media is not revolutionary. But developments from the last couple of years have changed things, things that are important for organizations to recognize. Previous research discovered that social media can lead to an increase of sales (Naylor et al., 2012; Stephen & Galak, 2012). It influences the purchase intentions and the behaviour of consumers as well. Strategies well-designed and executed contribute in satisfying performance goals (Mangold et al., 2009). Social media is responsible for the changed relationship between consumers and organizations (Henning-Thurau et al, 2004; Nambisan & Baron, 2007). It enables direct communication between those two parties and this development has changed the marketing landscape (Mangold et al., 2009; Kiang, et al., 1999). The importance of the development is enormous because of the extensive use of social media. Many of the millions of users integrate social media into their daily lives (Ellison, 2008; Wang et al., 2012). Organizations can reach many customers daily.

2.1.2 The quality of social networking websites

Effective online marketing strategies can lead to an increase of sales (Naylor et al., 2012; Stephen et al., 2012). But what makes a strategy effective? What is necessary for a company to include in their marketing strategy? The focus of this paragraph is on elements of social media that can influence the quality of social networking websites. It describes tactics that a company should or should not do on their social networking websites to attract customers.

An important advantage of social media is the possibility to communicate with the customers but the possibility for customers to communicate with each other as well. Customers like to communicate with other customers. They feel more engaged to the company (Mangold et al., 2009). And people who feel more engaged are tend to interact more with each other (Tu &McIsaac, 2002). Being active on social media creates a community where this is made possible (Dwyer et al., 2007). And being in a community connects people with one another. The increased engagement is also a result of the fact that customers are able to submit feedback. And this will likely result in (electronic) word-of-mouth (Mangold et al., 2009). Interaction depends also on the level of trust (Weaver & Morrison, 2008). And there is an ongoing discussion about online trust. The fact that people are using their ‘real world’ identity increases the level of trust (Keenan et al., 2009; Weaver et al., 2008). The problem is perhaps bigger for Twitter, users are not necessarily using their own identity.

According to Ranganathan et al. (2002) the content of a website is important as well and it plays a role in influencing the purchase decisions of a consumer. Consumers are more likely to talk about an organization and even thinking about purchasing a product when they know something

about the company. People want to know facts about the products and services but about the company itself as well (Mangold et al., 2009). Content refers to the information that is available on the website. It is about what relevant information you can find about the company.

Organizations spend a great amount of time in developing the social networking websites and whole teams are responsible for the design of the websites. And this is necessary, because the design plays a crucial role in attracting, sustaining and retaining the interests of a consumer (Ranganathan, 2002). It should be easy for a visitor to navigate, people do not want to think when they visit a website (C. Spruijt, personal communication, April 28, 2014). Poorly designed websites affects the sales volume in a negative way. A websites should be entertaining as well (Mangold et al., 2009). The use of multimedia is used to capture the attention of a consumer. Multimedia includes dynamic and static texts, images and sounds (Ranganathan et al., 2002). Design is all about the look and the feeling.

The social networking websites should be up-to-date. This is the last factor that could influence the quality of a website. The social media environment is dynamic and it evolves at a fast rate. Managers spend a great amount of time to develop a social media campaign but if it is not up-to-date, it would not result in the desired outcomes (Hoffman & Fodor, 2010). Being active should reach consumers as well. A consumer will not find a Facebook-page or Twitter-account if it is inactive.

All these four factors could influence the quality of a social networking website and they correlate with each other. A company should react fast when a consumer asks a question on Facebook or complains about the service on Twitter. If something happens in the environment which is relevant for a company, they need to react fast. Social campaigns should be entertaining but provide information as well.

2.1.3 Brand evaluation and purchase intentions

The aim of this research is to discover if the quality of social networking websites influences the evaluation of a brand and the purchase intentions as well. First, existing literature of brand evaluation in general is relevant to gain some insights.

Social media has changed the marketing landscape, consumers search for information regarding brands and products on the Internet (Mangold et al., 2009; Bambauer-Sachse & Mangold, 2011). The communication of consumers on social platforms has several impacts. The interaction between consumers influences the attitudes towards the products and services of a brand (Churchill & Moschis, 1979;Naylor et al., 2012; Mukhopadhyay & Yeung, 2010). The attitude towards a brand is influenced by the characteristics of a product or service (Garvin, 1984). Social media strengthens this effect. Consumers’ attitudes towards brands are influenced by the information they have about a

product. Social media enables consumers to search for more information. And as mentioned before, the content of social media is created by the users. According to Schivinski and Dabrowski (2014), the user-generated content has a great effect on the consumers’ overall perception of a brand. The effect is larger than the content created by the firms. People trust the information provided by other individuals more and find it more credible (Pornpitakpan, 2004). The effect of firm-generated content is smaller but there still is an effect. Brand communication does positively effects the brand evaluation. The messages should, however, create a satisfactory reaction (Yoo, Donthu & Lee, 2000).

Having a positive product attitude will lead to purchase intentions (Wang et al., 2012). This is in line with the theory of planned behaviour by Azjen (1991). Attitude leads to intentions, which could lead to behaviour. Naylor et al. (2012) mentioned as well that the interaction with consumers could have a positive effect on purchase intentions.

All this information has led to the main research question: What elements of social media influences the purchase intentions of the consumer?. The next step is to develop a conceptual framework.

2.2 Conceptual framework

The conceptual framework follows from the explored literature. First, expectations about factors that influence the quality of Facebook and Twitter will be stated. Then the relation between social networking websites and brand evaluations will be given. Followed by expectations of the purchase intentions of consumers based on the brand evaluations. But first the visual representation will be given:

Figure 1: The conceptual framework.

Purchase

intentions

Brand

evaluation

The quality

of Facebook

The quality

of Twitter

2.2.1 Influential factors of the quality of Facebook and Twitter

Social media has changed the relationship between consumers and organisations (Henning-Thurau et al, 2004; Nambisan et al., 2007). The fact that consumers can communicate with other consumers is an advantage of social media, as well the possibility for organizations to communicate with their customers. More engaged consumers will lead to better results (Mangold et al., 2009). On Facebook, it is possible to leave a review on the Facebook page of a company so other users can read your experiences. The interaction between consumers is an advantage of social media. So we expect that interaction can influence the quality of a Facebook page. This will lead to proposition (P1a).

Previous research also discovered that the content on a social networking website is from great value (Ranaganathan et al., 2002). Information availability is a factor that can influence the quality as well. Information is a broad term. A description of the company should be mentioned on the Facebook page as well as information about the products (Mangold et al., 2009). We expect that a link to the original website of the company and other general information about the company should be included on the page. All this is concluded in proposition (P1b).

According to Ranaganathan et al. (2002) is the design of a website crucial and it can influence the purchase intentions of consumers. Therefore we believe design is a factor as well that can influence the quality of a Facebook page. A well-designed website attracts customers and it helps to sustain and retain the interests of consumers. A social networking should be entertaining yet informative (Mangold et al., 2009). Ranaganathan et al (2002) discovered as well that the use of dynamic texts, images and sounds captures the attention of a visitor of the website. The use of multimedia is important for a social networking website. We expect that design is important for the quality. This will lead to proposition (P1c).

A last factor that is expected to influence the quality of a Facebook-page as well is actively. In a fast changing environment is it important to stay up-to-date and actively participate on Facebook. Spending nog enough time could lead to undesirable outcomes according to Hoffman et al. (2010). This expectation has led to proposition (P1d).

Proposition 1a: Interaction of a company has a positive effect on the quality of a Facebook-page. Proposition 1b: General information about the company should be included on the Facebook-page to

improve the quality of the social networking website. Proposition 1c: The quality of a Facebook-page is influenced by its design.

Proposition 1d: A company should actively participate on Facebook to improve its quality.

Facebook and Twitter have a lot in common, most of the factors that could influence the quality of a Facebook-page could influence the quality of Twitter-account as well. There are, however, a few

differences. The contact between the organizations and the customers are more direct. So interaction is for Twitter as much as important as it is for Facebook. This difference is very useful for companies to interact with their customers (Twitter.com, 2014). This has led to proposition (P2a).

And actively is again important as well. Being up-to-date is important for every social media platform (Hoffman et al., 2010) but the fact that the contact is much more directly also indicates that reacting on questions on Twitter is an important feature of Twitter. This expectation can be found in proposition (P2b).

Design is once more important but not specifically the design of the Twitter-page but more the design of the Tweets. There are many possibilities to improve the quality of your Tweets, for example the use of multimedia. The fact that a Tweet can only contain 140 characters will demand creativity to include all the relevant and necessary information (Twitter.com, 2014). The amount of characters and the possibilities within a Tweet makes the design so important. This has led to proposition (P2c).

Proposition 2a: Active interaction with the followers on Twitter will improve the quality of an account. Proposition 2b: Actively participation on Twitter has a positive influence on the quality of an account. Proposition 2c: The design of Tweets influences the quality of a Twitter-account.

2.2.2 Brand evaluation

The evaluation of a brand is likely to be influenced by the quality of the used social media. All the factors discussed in the literature review influence the quality of a website and combining these factors will result in the overall quality of a social networking websites. Combining propositions 1a till 1d together will give valuable information about the quality of a Facebook-page. And combining the three factors that could influence a Twitter-account will result in the overall quality of Twitter. Content created by users influence other users’ perception of brands (Schivinski et al., 2014). In essence, social media influence the evaluation of brand in the eyes of the consumer. This has led to proposition (P3).

Proposition 3: The overall quality of Facebook and Twitter affects the evaluation of the brand in the eyes of the consumer.

2.2.3 Purchase intentions

The well-known psychologists Azjen (1991) have developed the theory of planned behaviour.

Consumers, who have a positive attitude towards a brand, have the intention to purchase a product or service from this specific brand. The consumers must have the idea that he or she is able to do so

as well. So for example, the consumers have the money and time to purchase the product/service. This expectation has led to proposition (P4).

Proposition 4: A positive brand evaluation will lead to higher purchase intentions.

2.3 Conclusion

The relationship between the consumer and organizations has changed and the reason for this development is the rising of social media. Social media is all about the user. The users create the content that is available on the World Wide Web. Organizations have to adapt to this change. They must be active on social media. But existing literature has also found a few factors that could influence the quality of a social networking website. All this indicates that social media is a serious development and simply being active does not seem to be enough. A social networking website should be treated carefully and with great caution. Wrong decisions could influence the brand evaluation negatively. And according to existing literature, brand evaluation influences purchase intentions.

This is however not been investigated extensively. This research combines the available information about brand evaluation and purchase intentions with factors of social networking websites that could affect one of these elements. The following question sums this up and this is the main research question of this study: What elements of social media influences the purchase intentions of the consumer?

3. Methodology

The existing literature and the conceptual framework that flows from the literature gap are discussed in the previous section. The next step is to explain the research design and the methods used to test the propositions that are stated in the literature review.

3.1 Research design

The aim of this research is to compare the answers of a large amount of people. The research is a combination of a descriptive and explanatory research. On one hand, the attitude and opinions of consumers towards social media is needed to answers the research question. On the other hand, the relationship between several variables is needed as well to see the relation between the quality of a social networking website and the purchase intentions. Knowing this, a questionnaire is considered the best for this research (Saunders, Lewis & Thornhill, 2012, p. 419). Web-based questionnaires seem to be the most efficient way to reach a significant amount of respondents for two reasons. The internet is fast and dynamic so it enables you to reach a lot of people in a short amount of time. And people on the World Wide Web will probably be familiar with Facebook and Twitter. In order to compare the answers of all the respondents, they must get the same questions. The questions must be consistent and standardised to make the comparison reliable (Saunders et al., 2009, p. 373). The fact that the questionnaires are web-based also indicates that they are self-completed questionnaires. Self-administration saves time and without an interviewer, it guarantees anonymity. This will improve the reliability because there is little to no participant or subject bias (Saunders et al., 2009, p. 156).

Every method has some limitations as well and a questionnaire is not an exception. Short questionnaires might indicate that the research is insignificant but people are more likely to complete a shorter questionnaire. People do not like to fill in larger questionnaires (Saunders et al., 2012, p. 446). To increase the response rate, the used questionnaire is made as short as possible without eliminating too much questions. Another limitation is that with a questionnaire, the questionnaire should be done properly from the beginning. Once the questionnaire is launched, it is not possible to adapt questions (Saunders et al., 2009, p. 366).

3.2 Research sample

Before going deeper into the data collection, we first describe the research sample. Knowing your respondents will help you with the analysis of the data. This study focuses on (Dutch) consumers who are active on social media. None of the propositions discussed earlier focuses on a specific

demographic characteristic. However, it still is important to have a good representation of the population.

First, the number of respondents is important to generalize the results (Saunders et al., 2009, p. 217). The larger the sample size, the better of course. But mainly due time and money constraints, a large sample was impossible to draw. According to Saunders et al., (2012), the minimum amount of respondents is at least 30 people. Having at least 30 people will enable you to analyse the data if they were normally distributed. This means the data can be used for most statistical tests. In the existing literature, demographic variables (e.g. age, gender and education level) seem to have no influence on the results. Nothing can be found about the influences of these variables on the effect of social media. However, at least 30 respondents must be familiar with Facebook and/or Twitter. Most people who are familiar with Facebook probably are familiar with Twitter as well. To sum up, approximately 50 to 60 respondents are necessary. But of course, the larger the sample size, the better. And it is unlikely that 50 à 60 respondents represent the population. It was strived to reach a sample of at least 100 fully completed questionnaires. The final sample consisted of 199 people who started the questionnaire. However, 50 people have not finished the questionnaire and 23 respondents have finished the questionnaire but did not fill in every question. Apparently a mistake is made because it was believed that people could not continue with the questionnaire if they have not filled in every question. But during the analysis of the data it became clear this was possible. Most of these 23 people forget only a few questions but some have just filled in half of the questionnaire. To avoid the discussion of which respondents should be included in the analysis and which not, only fully completed questionnaires are used during the analysis of the data. So, the final sample size consists of 126 of fully completed questionnaires which exceed the goal of 100 respondents.

3.3 Data collection

To collect the data, the website www.qualtrics.com is used to develop the questionnaire. This internet-mediated method provides useful guidelines and is easy to use for the researcher and the respondents. It saved a lot of time as well because the data had not to be entered manually (Saunders et al., 2009, p. 365). The questionnaires from qualtrics.com are easy to distribute throughout the World Wide Web as well. It has been posted several times on the personal Facebook-page and on the Twitter-account of the researcher. And many students were approached on a Facebook-page called ‘Respondentengezocht’ (translation: respondents wanted) which is specially made for people who need respondents.

Collecting data through the Internet has some limitations. People choose whether they want to fill in the questionnaire or not, it leads to self-selection (Wright, 2005). People are tend to fill in a

questionnaire if they feel connected with the topic. It is true that not everyone use the Internet. But for this research, we were looking for people who are familiar with social networking websites so with the Internet as well.

The questionnaire is available in two languages, English and Dutch. This research focussed on Dutch consumers with Facebook, Twitter and/or other social networking websites but the English is included as well because this study is written in English as well and it is therefore easier to communicate throughout the World Wide Web. And perhaps some people outside the Netherlands have filled in the questionnaire as well. The websites qualtrics.com keeps the questionnaires easy and accessible which could increase the response rate (Saunders et al., 2009, p. 387). The questionnaire starts with a short introduction where the purpose of the survey becomes clear and the fact that the opinion of the respondents is important and anonymous. It is also possible to contact the researcher with questions or comments (Saunders et al., 2009, p. 393). During the questionnaire, short definitions of Facebook and Twitter are given to ensure that people have enough knowledge about the topic to answer the questions. So, the people who are not familiar with Twitter were still able to answer the questions about Twitter.

3.4 Measures

This section describes the measures used for the variables of the survey. They appear in order of the questionnaire. It starts with demographic variables and some additional information. After that, the regression analysis is discussed. For a copy of the survey, both in English and Dutch, see appendix A.

3.4.1 Demographic variables and additional information

The questionnaire starts with a few introduction questions. These questions are asked to get to know the research sample and for respondents it is pleasant to start with some introduction questions instead of difficult in-depth questions. The questions concerning gender two answer options: what is your gender? Option 1) male or option 2) female. Age will be asked as an open-ended question. The reason for this choice is the possibility to categorize the age of the respondents after all the data is collected. The current education level as well as the highest completed level of education is questioned. Although, the English and Dutch education policy is slightly different, the options were similar. Both question had the same options, the options were: 1) primary school, 2) high school, 3) college, 4) university and 5) none. And as mentioned before, people were asked if they are familiar with Facebook and/or Twitter.

3.4.2 Regression analysis

During the literature review, we discovered several factors that influence the quality of social 17

networking websites in general. Not all these factors apply to both the social networking websites of this research. In order to test the propositions concerning these factors, the opinions of the customers had to be asked about these elements. Several questions have been asked about each of these factors and each of the social networking websites. So, the respondents had to fill in a number of questions about the interaction with companies on Facebook. They also had to answer how much they value the availability of information on the Facebook-pages. Questions about Twitter where of course included as well. The next part of the questionnaire included questions about social networking websites in general and question about the potential behaviour of consumers towards a brand.

For the question to be reliable and valid, it is important that the respondents understand the concepts in the way intended by the researcher (Saunders et al., 2009, p. 393). The respondents need to understand everything so the use of scientific terms is not preferred. A person who is not an expert on social media, brand evaluation and purchase intentions must understand the questions. In other words, lower educated people do not have to experience difficulties. To make sure every respondent know what Facebook and Twitter is all about, short descriptions are given before the questions. Similar descriptions as the once used in the literature review are used as well in the questionnaire.

All the question are asked on a Likert-scale from 1 till 7 (1=totally disagree,7=totally agree). The first step after collecting the data is checking if summated scales can be created. The goal is to create a reliable summated scale of every influential factor. After the data are transformed into these (reliable) variables, it is time to discover the most suitable method to test the propositions. A seven-point Likert-scale can be treated as quantitative data, although there is an ongoing discussion if this is appropriate. Researchers have different opinions about this. Some say it is acceptable to treat it like quantitative data, others mentions that the step from 1 to 2 is not the same as the step from 5-6 for example. In this study, the variables with a seven-point Likert-scale are treated as quantitative data. The fact that all the questions are asked on such a scale, results in quantitative data. So the predictor (independent) variable and outcome (dependent) variable are quantitative. The most suitable method to test the propositions is regression analysis. The regression analysis consists of three steps. The first step is to check how much percentage of the outcome variable can be explained by the combination of the used predictor variables. The second step is to look at the value. The F-value and the associated significance will tell us that there is a minimum chance that an F-F-value this large would happen if nothing is going on. This basically tells us that the regression model results in significantly predictions (Field, 2009, p.207). The last step is to check the influence of the individual predictor variables. The steps will be further explained along the process. If you are doing a regression analysis. It is necessary to check for multicollinearity as well (Field, 2009, p. 223).

Multicollinearity exists when there is a strong correlation between two or more predictors in a (multiple) regression model. This will be explained more extensive in the result section (chapter 4).

3.4.3 The final survey construction

Due to time and money constraints, it was not possible to pilot testing the survey as extensively as necessary (Saunders et al., 2012, p. 394). A few people were asked to read the questionnaire and give their opinion. They were asked if they understood the questions and the descriptions. This gave some information about the validity of the questionnaire (Saunders et al., 2009, p. 394). Important was the length of the questionnaire. It turned out that the question will take on average 4-8 minutes. The order of the questionnaire is important as well. It was decided to start with a few simple questions about the respondents, so that they could get used to the questionnaire. It is not preferred to start with difficult, deeper questions. The next part is introduced by a short description of Facebook. Although every respondent answered that they are familiar with Facebook, the description is preferred. After a couple question about Facebook, questions were introduced about Twitter. And this topic is introduced with a short description as well. After this part, a few questions were included about social networking websites in general. What do consumers value? What is good to do and what is discouraged? And the survey ends with a couple of question who are more focussed on the behaviour of consumers. These questions are tend to measure the willingness to purchase a product/service or not. So basically, the survey follows the conceptual framework. This makes the analysis of the data easier as well.

4. Results

This section shows the results of the questionnaire. This section starts with some sample characteristics to get to know the respondents. After that, the section continues with the reliability of scales, this is done by calculating Cronbach´s Alpha. Then, the summated scales are given with the some descriptive to get to know the variables. The propositions are tested with regression and the results are shown in the end of this section.

4.1. Sample characteristics

The total number of people who started the questionnaire was 199, but almost 25% have not finished it and a few respondents did not fill in every question. The final sample consists of 126 people, as mentioned in the methodology, chapter 3.

The average age of the research sample is almost 26 years old (25 years and 258 days old to be precise). The minimum age was 18 years old and the maximum was 60 years old. When we take a look at the box plot (see Appendix B, figure 1b), we discover that there are eleven outliers above the upper whisker. This indicates skewness to the right and the distribution is indeed positively skewed (2.636). This shows that most of the respondents are relatively young. For more information, see table 1.

This is obvious not the average age of the Dutch population. The reason for this is the fact that most of the respondents are students. 64 Respondents (50.8%) are currently studying at a university and 33 respondents (26.2%) are going to college. Only 4 people (3.2%) are going to high school and 25 people (19.8%) do not follow any kind of education at the moment (see table 2).

Of the respondents, 74 people (58.7%) were male and 52 people (41.3%) were female. In the population, there are slightly more females (50.4%) than males (49.6%) (CBS.nl, 2014). Most of the respondents are male friends from my personal network, this could be the reason why in this research there are more males than females.

Perhaps the most important descriptive to look at, is the amount of people who are familiar with Facebook and/or Twitter. Of the 126 respondents, everyone is familiar with Facebook. This is not surprisingly because the questionnaire is distributed through Facebook. Not everyone is familiar with Twitter, 108 people (85.7%) is familiar with Twitter. At least 30 respondents are familiar with Facebook and/or Twitter. This requirement is achieved.

Table 1: Age sample

< 20 20-24 25-29 30-34 35>

Sample (N=126) 9.5% 63.6% 14.3% 3.2% 9.4%

Table 2: Education distribution

Primary school High School College University None

Highest completed 1.6% 57.1% 21.4% 19.8% 0.0%

Current 0.0% 3.2% 26.2% 50.8% 19.8%

4.2 Reliability statistics

Several data derived from the questionnaire are tend to measure the same connection. Working with fewer variables in the data set is preferred. The creation of summated scales reduces the amount of data without the loss of information. The increase in reliability has to outweigh the loss of information. We have to do a reliability test to discover if it is possible to create summated scales. Cronbach´s alpha is used to make these new variables reliable. Cronbach’s alpha measures the internal consistency reliability, also known as inter-item correlation. It is the commonly used measure of scale reliability (Field, 2009, p. 674). A Cronbach’s alpha of 0.6 is the minimum value that is considered as reasonable (R. Pruppers, personal communication, 2014). It is a guideline but this thumbnail is used by many researchers.

Multiple items in the survey were relating to the same factor that could influence the quality of a social networking website. With the use of the Cronbach’s alpha, we can discover if these data are correlated. And if so, we can create a summated scale. A couple of items in the survey are related to the interaction between the organization and consumers and between consumers. Putting all the items it did result in a reliable variable. But the next step is to check if the variable can be increased in reliability by deleting one or more items. By deleting a few items, the Cronbach’s alpha can be improved to a value of 0.737. This is considered as a reliable value. The Cronbach’s alpha of this variable and the others regarding the influential factors can be found in table 3.All the correlated items regarding to the influential factors are reliable except one. Four items related to the design of Facebook are not strongly internal correlated. The measurement is not consistent. The Cronbach’s alpha is 0.537 and it can’t be increased by deleting an item. So it is not possible to create a summated scale of the variable content of Facebook. This variable is not included in the research although it might be interesting to look at. Several times, the scale had to be reversed so that the data are pointing in the same direction. A variable is created to do so.

Table 3: The reliability test

Variable Cronbach’s alpha Number of items

Interaction on Facebook 0.737 2

Actively on Facebook 0.656 2

Information availability on Facebook 0.841 2

Design of a Facebook-account 0.537 4

Actively on Twitter 0.674 3

Interaction on Twitter 0.771 3

The first part of the questionnaire was about these influential factors. The second part of the questionnaire focussed on social networking websites in general, the brand evaluation and possible purchase intentions. Six items were included in the survey about brand evaluation. They tried to measure whether the fact people valuate a brand higher if they have a good and effective online marketing strategy. Another five items were included as well about potential purchase intentions. Are consumers tend to purchase products or services from a company they value? In first instance, the Cronbach’s alpha was not consistent. Deleting one item did increase the Cronbach’s alpha to 0.617. By deleting another item, the Cronbach’s alpha increased to 0.831. The increase was relatively and absolutely important. The Cronbach’s alpha is now above 0.8 and this is reliable. And again, a few scales had to be reversed in order to point the data in the same direction. The Cronbach’s alphas can be found in table 1b below:

Table 4: The reliability test

Variable Cronbach’s alpha Number of items

Quality of Facebook 0.632 6

Quality of Twitter 0.669 2

Brand evaluation 0.790 6

Purchase intentions 0.831 3

4.3 Data descriptions

It is useful as well to get to know the data. Therefore we show some descriptive statistics regarding the valuation of social networking websites. Everything is rated on a scale from 1-7 (1 = totally disagree, 7 = totally agree). The minimum, maximum, mean and standard deviation is shown in table 5. Table 6 focuses on the brand evaluation and purchase intentions.

Table 5: Data descriptions

(N=126) Mean Std. error Min. Max.

Interaction on Facebook 4.94 0.114 1 7

Actively on Facebook 5.18 0.101 2 7

Information availability on Facebook 5.73 0.094 1 7

Actively on Twitter 4.33 0.103 1 7

Interaction on Twitter 4.38 0.115 1 7

Content of a Twitter-account 4.3 0.124 1 7

Table 6: Purchase intentions and Brand evaluation

(N=126) Mean Std. error Min. Max.

Quality of Facebook 4.54 0.078 2 6

Quality of Twitter 5.17 0.105 1 7

Brand evaluation 5.11 0.081 2 7

Purchase intentions 4.84 0.058 3 7

4.4 Correlations

Before we start testing the propositions, we first take a look at the correlations between the different variables. It will help to quantify the strength of the linear relationship between variables (Saunders, et al., 2012, p. 521). All the data are numerical and therefore we use the Pearson’s product moment correlation coefficient. A correlation of -1 indicates a perfect negative relationship and a correlation of +1 a perfect positive relationship and 0 means there is perfect independence. Furthermore, we use the following guidelines which are commonly used in business research:

Table 7: Values of the correlation coefficient

-1 – -0.8 -0.8 – -0.6 -0.6 – -0.35 -0.35 – -0.2 -0.2 – 0 Very strong negative Strong negative Moderate negative Weak negative None

1 – 0.8 0.8 – 0.6 0.6 – 0.35 0.35 – 0.2 0.2 – 0

Very strong positive Strong positive Moderate positive Weak positive None

Table 3b shows the correlations between all the variables of the influential factors of both the social networking websites. The complete table which includes the significance levels can be found in the Appendix B, figure 2a. All activities are positive correlated and it is highly significant (P < 0.05). Two correlation (see Appendix B, figure 2a) are not significantly correlated with at a significance level of 1

percent but they are significant correlate with a significance level of 5 percent (p=0.025, p=0.14). As you can see in the table below, the activities vary from weak positive to strong positive. The lowest correlation is between the information availability on Facebook and the actively on Twitter (r = 0.199) and the highest correlation is between the actively on Twitter and the interaction on Twitter (r = 0.726). It is not surprisingly that the activities between the variables of one social networking websites are in general higher than the activities between variables of both social networking websites. The correlations between the variables regarding Facebook are in general moderate positive and the same counts for the variables regarding Twitter. But the correlations between Facebook and Twitter are in general weak positive.

Table 8: The correlation between all the variables (FB = Facebook, TW = Twitter)

Interaction on FB Actively on FB Information availability on FB Interaction on TW Actively on TW Design of TW Interaction on FB _{1 0.599 0.500 0.477 0.280 0.306} Actively on FB _{0.599 1 0.562 0.367 0.331 0.218} Information availability on FB _{0.500 0.562 1 0.199 0.262 0.291} Interaction on TW _{0.477 0.367 0.199 1 0.726 0.636} Actively on TW _{0.280 0.331 0.262 0.726 1 0.648} Design of TW _{0.306 0.218 0.291 0.636 0.648 1} 4.5 Regression analysis

If we take a look and what kind of data we have, we see that the predictor (independent, PV) variables are continuous data as well as the outcome (dependent, OV) variable. In order to test the propositions, regression is the most suitable method. Regression explains the variance in one variable (OV) with several other variables (Field, 2009). The process of the analysis starts with the 𝑅𝑅2

and the adjusted𝑅𝑅2_{. The 𝑅𝑅}2 _{tells us how much of the outcome variable can be explained only by the}

predictor variables. The adjusted 𝑅𝑅2 _{keeps in mind that the 𝑅𝑅}2 _{will always improve when PVs are}

added. The adjusted 𝑅𝑅2 _{only improves when the additional PVs add more explanatory power that}

would be expected by change. We have to look at the F-value as well. The last kind of analysis we will do is to test if an individual PV is important for explaining the outcome variable. This is the individual coefficient testing. This analysis will be done a few times. We start by looking at Facebook and the factors that could influence the quality of a Facebook-account. Next, we test the propositions regarding to Twitter. Then we look at the propositions regarding brand evaluation and as last we test the propositions regarding the purchase intentions.

4.5.1 Facebook

In the literature review, it became clear that interaction, actively, design and information availability could influence the quality of a Facebook-page. Earlier in this section, we discovered that the variable ‘design of a Facebook-account’ is not reliable and that it is not recommended to develop a summated scale with the items. Therefore, we compare the model with two blocks. The first block consists of the variables that are reliable and the second block includes all the variables. Doing the regression analysis gives us the following information.

Table 9: The 𝑹𝑹𝟐𝟐_{, adjusted 𝑹𝑹}𝟐𝟐 _{and the change statistics}

Model R R Square Adjusted R Square Change statistics

R Square Change F Change Sig. F Change

1 0.451 0.204 0.184 0.204 20.852 0.000

2 0.475 0.226 0.200 0.022 0.959 0.064

Perhaps not surprisingly, the adjusted 𝑅𝑅2 _{did not increase substantially, it increased with 0.022. The}

variable ‘design of Facebook’ does not add much additional information. The change of the F-value as a response of the additional predictor variable is not significant as well (0.064). The 𝑅𝑅2 _{of model 1 is}

0.204, this indicated that the three predictor variables (interaction, actively and information availability) explains 20.4% of the evaluation of a social networking website.

Table 10: The F-value

Sum of Squares df Mean Square F Sig.

Regression 19.491 3 6.497 10.394 0.000

Residual 76.255 122 0.625

Total 95.745 125

The F-value indicates the overall probability of the relationship between the dependent variable and all the independent variables occurring by change. The F-value of this regression is 10.394 and it is significant (p=0.000<0.05). As mentioned before, the next step is to test if individual predictor variables influence the outcome variable. The method used is the t-test for the coefficient of the individual variables. The t-values of the individual PVs and the associated significant value are given in the table 11.Two t-values are significant (p<0.05), so these two variables have a significant influence on the quality of a social networking website. But which effect is the strongest? The beta values are all comparable because they are all measured in standard deviation units. They provide

insight into the importance of a predictor in the model. In other words, the variable with the highest beta has the strongest impact on the outcome variable. As shown in the table below, the interaction on Facebook has the highest beta (=0.296) and actively has the lowest (=0.003). In conclusion, propositions 1a and 1b are supported and proposition 1d is not supported.

Table 11: Impact of individual predictor variables

Beta t Sig.

Interaction on Facebook 0.296 2.847 0.005

Actively on Facebook 0.003 0.026 0.979

Information availability on Facebook 0.221 2.188 0.031

4.5.2 Twitter

The same thing we have done with Facebook, is necessary with Twitter to test the propositions. But in the literature we have only discovered three possible factors that could influence the account of an organisation on Twitter. Actively, interaction and the design of the Tweets are seemed to influence the quality of a Twitter-account. The variables ‘actively on Twitter’ and ‘interaction on Twitter’ are reliable variables. The variable ‘design of Tweets’ only consist of one item and is therefore not tested on its reliability. Table 12 and 13 will give the 𝑅𝑅2 _{and the F-value of the}

regression. The 𝑅𝑅2 _{is 0.154 and the adjusted 𝑅𝑅}2 _{is slightly lower (=0.133). This indicates that the}

three predictor variables (actively, interaction and design) explain 15.4% of the quality of Twitter. The F-value (=7.396) is significant (p=0.000<0.05).

Table 12: The R, 𝑹𝑹𝟐𝟐_{, adjusted 𝑹𝑹}𝟐𝟐

Model R R Square Adjusted R Square

1 0.392 0.154 0.133

Table 13: The F-value

Sum of Squares df Mean Square F Sig.

Regression 26.711 3 8.904 7.396 0.000

Residual 146.871 122 1.204

Total 173.581 125

When we look at the individual PVs, it becomes clear that none of the variables are significant with a significance level of 5 percent. It is clear that the variable ‘design of Twitter’ does not significantly influence the quality of a Twitter-account (p = 0.409 > 0.05). But interaction and actively have a p-value of 0.224 and 0.148 as well. So in conclusion, none of the propositions are supported.

Table 14: Impact of individual predictor variables

Beta t Sig.

Interaction on Twitter 0.156 1.223 0.224

Actively on Twitter 0.188 1.457 0.148

Content of Twitter 0.095 0.828 0.409

4.5.3 Brand evaluation

As mentioned in the methodology, the variable of brand evaluation has a lot in common with the evaluation of Twitter and Facebook. Similar factors are influencing the quality of the social networking website. Therefore, the brand evaluation of Facebook and Twitter will explain for a huge part the evaluation of a brand in general. The 𝑅𝑅2 _{is not surprisingly 0.876. A high 𝑅𝑅}2 _{usually combines}

with a high F-value as well. The F-value of this model is 433.302 and the F-value is significant (p = 0.000). And also both the predictor variables have a significant impact on the evaluation of a brand. The evaluation of Twitter has the strongest impact (Beta = 0.722). The beta of the evaluation of Facebook is 0.324. All the data can be found in the tables below:

Table 15: The R, 𝑹𝑹𝟐𝟐_{, adjusted 𝑹𝑹}𝟐𝟐

Model R R Square Adjusted R Square

1 0.936 0.876 0.874

Table 16: The F-value

Sum of Squares df Mean Square F Sig.

Regression 90.928 2 45.464 433.02 0.000

Residual 12.906 123 0.105

Total 103.833 125

Table 17: Impact of individual predictor variables

Beta t Sig.

Quality of Twitter 0.722 19.199 0.000

Quality of Facebook 0.324 8.631 0.000

4.5.4 Purchase intentions

The last regression analysis will be done to test the relation between brand evaluation and potential purchase intentions. In the end, marketing should lead to an increase in sales and therefore the intentions to purchase a product or service must be increased. We follow the same process as we have done the entire research. The 𝑅𝑅2 _{of this model is 0.07 and the adjusted 𝑅𝑅}2 _{is again slightly}

lower (=0.062). So, 7% of the purchase intentions can be explained by the evaluation of the specific brand. The F-value of this model is 9.326 and the p-value of this F-value is 0.003(<0.05), it is significant. The beta is equal to the R, the beta-value is 0.264. The t-value (=3.054) is significant, the p-value is 0.003(<0.05).

Table 18: The R, 𝑹𝑹𝟐𝟐_{, adjusted 𝑹𝑹}𝟐𝟐

Model R R Square Adjusted R Square

1 0.264 0.070 0.062

Table 19: The F-value

Sum of Squares df Mean Square F Sig.

Regression 3.704 1 3.704 9.326 0.003

Residual 49.247 124 0.397

Total 52.951 125

Table 20: Impact of individual predictor variables

Beta t Sig.

Brand evaluation 0.264 3.054 0.003

4.6 Multicollinearity

Multicollinearity exists when there is a strong correlation between two or more predictors in a (multiple) regression model (Field, 2009, p.223). A strong presence of correlation among the independent variables has the effect that the PVs are not totally independent. It affects calculations

regarding individual PVs. And this is a statistical problem. SPSS produces two collinearity diagnostics, variance inflation factor (VIF) and related to the VIF is the tolerance statistic. The last step of the analysis is checking the tolerance and VIF. A tolerance between 0 and 0.1 means serious collinearity, a value between 0.1 and 0.2 means there is some collinearity but not too severe and a value of 0.2 or higher means there is no collinearity problem (Field, 2009, p. 224; R. Pruppers, personal communication, 2014; Saunders et al., 2012, p. 525). If we take a look at the table, we see that there is no problem with collinearity if we look at the tolerance value. Not surprisingly, there is no problem with collinearity if we look at the variance inflation factor. The two values are related. A VIF between 0 and 5 means there is no collinearity, a value between 5 and 10 means there is some collinearity and a value above 10 means there is serious collinearity. Fortunately, this is not the case. The values of the multiple regression models can be found in the table below:

Table 21: Multicollinearity

Dependent variable Independent variable Tolerance VIF

Quality of Facebook Interaction on Facebook 0.602 1.660

Actively on Facebook 0.549 1.822

Information available on Facebook 0.642 1.557 Quality of Twitter Interaction on Twitter 0.426 2.347

Actively on Twitter 0.415 2.408

Design of a Twitter-account 0.522 1.917

Brand evaluation Quality of Facebook 0.715 1.398

Quality of Twitter 0.715 1.398

5. Discussion

This section will go deeper into the results, we will interpret the results in the context of the existing literature. It includes the supported propositions, the propositions that were not supported and additional findings. The section starts with linking these findings with the existing literature. The theoretical contributions and managerial implications of the research are given after that. And the section ends with the limitations of the study and suggestions for future research.

5.1 The findings

5.1.1Facebook

The existing literature has discussed a few elements a company should include in their online marketing strategy. Analysing the literature resulted in four elements of Facebook. These factors seem to influence the quality of a Facebook-page of a company. Paying attention to these factors should improve the quality. And an effective online marketing strategy can lead to an increase of sales (Naylor et al., 2012; Stephan et al., 2012).

An important advantage of social media in general is the fact that customers can communicate and interaction with the organization and with other customers as well (Dwyer et al., 2007). Consumers who are more engaged in the company are more profitable consumers (Mangold et al., 2009). Companies should interaction actively on Facebook to create more loyal consumers. Based on this information, expected is that the interaction of a company on Facebook has a positive effect on the quality of the Facebook-page. And this is supported by the results of this study as well. Interaction does have a significant effect on the quality of Facebook. Significant evidence has proved that interaction on Facebook is preferred by consumers. Interaction improves the quality of a Facebook-page of a company.

The content on a social networking websites is from great value as well (Ranaganathan et al., 2012). A customer prefers to know something about a company before they consider purchasing a product or service. Therefore a short description of the company should be included on the Facebook-page and information about the products/service is preferred as well. Expected is that the information availability is an influential factor of the quality as well. If consumers know something about the company, they are more willing to talk about the company and perhaps purchase an item. This proposition is supported by the results of this study as well. The impact on the quality is less strong that the impact of interaction on the quality, but there is a significant effect. A company should include general information on their Facebook-page to improve its quality.

The third element of Facebook that might have an impact on its quality is the actively of the organization on the social networking website. The existing literature mentioned that spending not

enough time on social media in general could lead to undesirable outcomes (Hofmann et al., 2010). Actively participation on Facebook seems to improve the quality of a company’s Facebook-page. This indicates that having a Facebook-account is not enough for a company, they should be active on the social networking websites. The social media environment is dynamic and fast as well and it evolves at a fast rate. If a company is not participating actively, they fall behind and this is considered negatively. A Facebook-account should be up to date. This is formulated in a proposition. This expectation is not supported by the results of this study. Believed is that the existing literature is not incorrect but there might be a methodological problem. Perhaps the questionnaire is not perfect and that therefore we could not find support for this proposition. Only the data of this study does not supported the proposition. It seems incorrect to assume that having a Facebook-account is enough for an organization. There is a possibility that this is true. This is against the grain of the existing literature. Hoffman et al. (2010) discovered that spending not enough time on social media leads to undesirable outcomes.

The last factor that was expected to influence the quality of a Facebook-page was the design of the page. Unfortunately, with the existing data from the survey, it was impossible to create a reliable variable who measures the design correctly. We exclude this variable from the analysis and therefore the proposition is not supported. The existing literature does mention that multimedia helps to sustain and retain the interests of consumers (Ranaganathan et al., 2002).

In this study, evidence is found for two propositions. Interaction and information availability significantly influence the quality of a Facebook-page. The combination of the three variables explain 20.4 (𝑅𝑅2 _{= 0.204) percent of the quality of Facebook. This perhaps seems not so much, but about one}

fifth of the overall quality is explained by only three factors. Many other factors could influence the quality as well.

5.1.2 Twitter

As mentioned in the literature review, Facebook and Twitter have a lot in common. Similar factors influence the quality of the social networking website. Companies use Facebook and Twitter both to communicate with their customers. The contact between the organisation and the customers is perhaps more direct on Twitter than on Facebook. On Twitter you receive a personal reaction on your question or your complain. Nevertheless, the interaction is an advantage of social media. According to Twitter (2014) is it ideal to interact with people. And actively is important for a Twitter-account as well. Being up-to-date is crucial for every social media platform (Hoffman et al., 2010). The third factor that is expected to influence the quality of a Twitter-account is the design of the Tweets. This is found in the existing literature. All the proposition can’t be supported with the results of this study. The probability that we’ll find equally or more extreme values is at least more than 14%